primops: Run iwyu and update includes

Change-Id: Ic3f48d6fca6ca1ae444475f7ccdee6220da946de

src/libexpr/meson.build

@@ -86,11 +86,25 @@ libexpr_sources = files(
   'flake/flake.cc',
   'flake/flakeref.cc',
   'flake/lockfile.cc',
+  'primops/arithmetic.cc',
+  'primops/attrset.cc',
   'primops/context.cc',
+  'primops/control.cc',
+  'primops/debug.cc',
+  'primops/derivation.cc',
+  'primops/hash.cc',
   'primops/fetchClosure.cc',
   'primops/fetchMercurial.cc',
   'primops/fetchTree.cc',
   'primops/fromTOML.cc',
+  'primops/import.cc',
+  'primops/json.cc',
+  'primops/list.cc',
+  'primops/path.cc',
+  'primops/string.cc',
+  'primops/system.cc',
+  'primops/toXML.cc',
+  'primops/types.cc',
   'value/context.cc',
 )
 

src/libexpr/primops.hh

@@ -1,9 +1,10 @@
 #pragma once
 ///@file
 
+#include "derivations.hh"
 #include "eval.hh"
 
-#include <tuple>
+#include <regex>
 #include <vector>
 
 namespace nix {
@@ -44,13 +45,150 @@ struct RegisterPrimOp
 /**
  * Load a ValueInitializer from a DSO and return whatever it initializes
  */
-void prim_importNative(EvalState & state, const PosIdx pos, Value * * args, Value & v);
+void prim_importNative(EvalState & state, const PosIdx pos, Value ** args, Value & v);
 
 /**
  * Execute a program and parse its output
  */
-void prim_exec(EvalState & state, const PosIdx pos, Value * * args, Value & v);
+void prim_exec(EvalState & state, const PosIdx pos, Value ** args, Value & v);
+
+void prim_lessThan(EvalState & state, const PosIdx pos, Value ** args, Value & v);
+
+void prim_importNative(EvalState & state, const PosIdx pos, Value ** args, Value & v);
+
+void prim_trace(EvalState & state, const PosIdx pos, Value ** args, Value & v);
+
+void prim_second(EvalState & state, const PosIdx pos, Value ** args, Value & v);
 
 void makePositionThunks(EvalState & state, const PosIdx pos, Value & line, Value & column);
 
+void mkOutputString(
+    EvalState & state,
+    BindingsBuilder & attrs,
+    const StorePath & drvPath,
+    const std::pair<std::string, DerivationOutput> & o
+);
+
+#if HAVE_BOEHMGC
+typedef std::list<Value *, gc_allocator<Value *>> ValueList;
+#else
+typedef std::list<Value *> ValueList;
+#endif
+
+/**
+ * getAttr wrapper
+ */
+
+Bindings::iterator
+getAttr(EvalState & state, Symbol attrSym, Bindings * attrSet, std::string_view errorCtx);
+
+/**
+ * Struct definitions
+ */
+
+struct RegexCache
+{
+    // TODO use C++20 transparent comparison when available
+    std::unordered_map<std::string_view, std::regex> cache;
+    std::list<std::string> keys;
+
+    std::regex get(std::string_view re)
+    {
+        auto it = cache.find(re);
+        if (it != cache.end()) {
+            return it->second;
+        }
+        keys.emplace_back(re);
+        return cache.emplace(keys.back(), std::regex(keys.back(), std::regex::extended))
+            .first->second;
+    }
+};
+
+struct CompareValues
+{
+    EvalState & state;
+    const PosIdx pos;
+    const std::string_view errorCtx;
+
+    CompareValues(EvalState & state, const PosIdx pos, const std::string_view && errorCtx)
+        : state(state)
+        , pos(pos)
+        , errorCtx(errorCtx){};
+
+    bool operator()(Value * v1, Value * v2) const
+    {
+        return (*this)(v1, v2, errorCtx);
+    }
+
+    bool operator()(Value * v1, Value * v2, std::string_view errorCtx) const
+    {
+        try {
+            if (v1->type() == nFloat && v2->type() == nInt) {
+                return v1->fpoint < v2->integer;
+            }
+            if (v1->type() == nInt && v2->type() == nFloat) {
+                return v1->integer < v2->fpoint;
+            }
+            if (v1->type() != v2->type()) {
+                state.error<EvalError>("cannot compare %s with %s", showType(*v1), showType(*v2))
+                    .debugThrow();
+            }
+// Allow selecting a subset of enum values
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wswitch-enum"
+            switch (v1->type()) {
+            case nInt:
+                return v1->integer < v2->integer;
+            case nFloat:
+                return v1->fpoint < v2->fpoint;
+            case nString:
+                return strcmp(v1->string.s, v2->string.s) < 0;
+            case nPath:
+                return strcmp(v1->_path, v2->_path) < 0;
+            case nList:
+                // Lexicographic comparison
+                for (size_t i = 0;; i++) {
+                    if (i == v2->listSize()) {
+                        return false;
+                    } else if (i == v1->listSize()) {
+                        return true;
+                    } else if (!state.eqValues(
+                                   *v1->listElems()[i], *v2->listElems()[i], pos, errorCtx
+                               ))
+                    {
+                        return (*this)(
+                            v1->listElems()[i],
+                            v2->listElems()[i],
+                            "while comparing two list elements"
+                        );
+                    }
+                }
+            default:
+                state
+                    .error<EvalError>(
+                        "cannot compare %s with %s; values of that type are incomparable",
+                        showType(*v1),
+                        showType(*v2)
+                    )
+                    .debugThrow();
+#pragma GCC diagnostic pop
+            }
+        } catch (Error & e) {
+            if (!errorCtx.empty()) {
+                e.addTrace(nullptr, errorCtx);
+            }
+            throw;
+        }
+    }
+};
+
+struct RealisePathFlags
+{
+    // Whether to check that the path is allowed in pure eval mode
+    bool checkForPureEval = true;
+};
+
+SourcePath
+realisePath(EvalState & state, const PosIdx pos, Value & v, const RealisePathFlags flags = {});
+
 }

src/libexpr/primops/arithmetic.cc

@@ -0,0 +1,312 @@
+#include <math.h>        // for ceil, floor
+#include <limits>        // for numeric_limits
+#include "eval-error.hh" // for EvalError, EvalErrorBuilder
+#include "eval.hh"       // for EvalState, PrimOp
+#include "pos-idx.hh"    // for PosIdx, noPos
+#include "primops.hh"    // for CompareValues, prim_lessThan
+#include "value.hh"      // for Value, nFloat, NixInt, NixFloat
+
+namespace nix {
+
+/**
+ * builtins.add
+ */
+
+static void prim_add(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    state.forceValue(*args[1], pos);
+    if (args[0]->type() == nFloat || args[1]->type() == nFloat) {
+        v.mkFloat(
+            state.forceFloat(*args[0], pos, "while evaluating the first argument of the addition")
+            + state.forceFloat(
+                *args[1], pos, "while evaluating the second argument of the addition"
+            )
+        );
+    } else {
+        v.mkInt(
+            state.forceInt(*args[0], pos, "while evaluating the first argument of the addition")
+            + state.forceInt(*args[1], pos, "while evaluating the second argument of the addition")
+        );
+    }
+}
+
+PrimOp primop_add({
+    .name = "__add",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return the sum of the numbers *e1* and *e2*.
+      Return a float if either *e1* or *e2* is a float, otherwise
+      return an integer.
+    )",
+    .fun = prim_add,
+});
+
+/**
+ * builtins.bitAnd
+ */
+
+static void prim_bitAnd(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    v.mkInt(
+        state.forceInt(
+            *args[0], pos, "while evaluating the first argument passed to builtins.bitAnd"
+        )
+        & state.forceInt(
+            *args[1], pos, "while evaluating the second argument passed to builtins.bitAnd"
+        )
+    );
+}
+
+PrimOp primop_bitAnd({
+    .name = "__bitAnd",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return the bitwise AND of the integers *e1* and *e2*.
+    )",
+    .fun = prim_bitAnd,
+});
+
+/**
+ * builtins.bitOr
+ */
+
+static void prim_bitOr(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    v.mkInt(
+        state.forceInt(
+            *args[0], pos, "while evaluating the first argument passed to builtins.bitOr"
+        )
+        | state.forceInt(
+            *args[1], pos, "while evaluating the second argument passed to builtins.bitOr"
+        )
+    );
+}
+
+PrimOp primop_bitOr({
+    .name = "__bitOr",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return the bitwise OR of the integers *e1* and *e2*.
+    )",
+    .fun = prim_bitOr,
+});
+
+/**
+ * builtins.bitXor
+ */
+
+static void prim_bitXor(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    v.mkInt(
+        state.forceInt(
+            *args[0], pos, "while evaluating the first argument passed to builtins.bitXor"
+        )
+        ^ state.forceInt(
+            *args[1], pos, "while evaluating the second argument passed to builtins.bitXor"
+        )
+    );
+}
+
+PrimOp primop_bitXor({
+    .name = "__bitXor",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return the bitwise XOR of the integers *e1* and *e2*.
+    )",
+    .fun = prim_bitXor,
+});
+
+/**
+ * builtins.ceil
+ */
+
+static void prim_ceil(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto value = state.forceFloat(
+        *args[0],
+        args[0]->determinePos(pos),
+        "while evaluating the first argument passed to builtins.ceil"
+    );
+    v.mkInt(ceil(value));
+}
+
+PrimOp primop_ceil({
+    .name = "__ceil",
+    .args = {"double"},
+    .doc = R"(
+        Converts an IEEE-754 double-precision floating-point number (*double*) to
+        the next higher integer.
+
+        If the datatype is neither an integer nor a "float", an evaluation error will be
+        thrown.
+    )",
+    .fun = prim_ceil,
+});
+
+/**
+ * builtins.div
+ */
+
+static void prim_div(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    state.forceValue(*args[1], pos);
+
+    NixFloat f2 =
+        state.forceFloat(*args[1], pos, "while evaluating the second operand of the division");
+    if (f2 == 0) {
+        state.error<EvalError>("division by zero").atPos(pos).debugThrow();
+    }
+
+    if (args[0]->type() == nFloat || args[1]->type() == nFloat) {
+        v.mkFloat(
+            state.forceFloat(*args[0], pos, "while evaluating the first operand of the division")
+            / f2
+        );
+    } else {
+        NixInt i1 =
+            state.forceInt(*args[0], pos, "while evaluating the first operand of the division");
+        NixInt i2 =
+            state.forceInt(*args[1], pos, "while evaluating the second operand of the division");
+        /* Avoid division overflow as it might raise SIGFPE. */
+        if (i1 == std::numeric_limits<NixInt>::min() && i2 == -1) {
+            state.error<EvalError>("overflow in integer division").atPos(pos).debugThrow();
+        }
+
+        v.mkInt(i1 / i2);
+    }
+}
+
+PrimOp primop_div({
+    .name = "__div",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return the quotient of the numbers *e1* and *e2*.
+    )",
+    .fun = prim_div,
+});
+
+/**
+ * builtins.floor
+ */
+
+static void prim_floor(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto value = state.forceFloat(
+        *args[0],
+        args[0]->determinePos(pos),
+        "while evaluating the first argument passed to builtins.floor"
+    );
+    v.mkInt(floor(value));
+}
+
+PrimOp primop_floor({
+    .name = "__floor",
+    .args = {"double"},
+    .doc = R"(
+        Converts an IEEE-754 double-precision floating-point number (*double*) to
+        the next lower integer.
+
+        If the datatype is neither an integer nor a "float", an evaluation error will be
+        thrown.
+    )",
+    .fun = prim_floor,
+});
+
+/**
+ * builtins.lessThan
+ */
+
+void prim_lessThan(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    state.forceValue(*args[1], pos);
+    // pos is exact here, no need for a message.
+    CompareValues comp(state, noPos, "");
+    v.mkBool(comp(args[0], args[1]));
+}
+
+PrimOp primop_lessThan({
+    .name = "__lessThan",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return `true` if the number *e1* is less than the number *e2*, and
+      `false` otherwise. Evaluation aborts if either *e1* or *e2* does not
+      evaluate to a number.
+    )",
+    .fun = prim_lessThan,
+});
+
+/**
+ * builtins.mul
+ */
+
+static void prim_mul(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    state.forceValue(*args[1], pos);
+    if (args[0]->type() == nFloat || args[1]->type() == nFloat) {
+        v.mkFloat(
+            state.forceFloat(*args[0], pos, "while evaluating the first of the multiplication")
+            * state.forceFloat(
+                *args[1], pos, "while evaluating the second argument of the multiplication"
+            )
+        );
+    } else {
+        v.mkInt(
+            state.forceInt(
+                *args[0], pos, "while evaluating the first argument of the multiplication"
+            )
+            * state.forceInt(
+                *args[1], pos, "while evaluating the second argument of the multiplication"
+            )
+        );
+    }
+}
+
+PrimOp primop_mul({
+    .name = "__mul",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return the product of the numbers *e1* and *e2*.
+    )",
+    .fun = prim_mul,
+});
+
+/**
+ * builtins.sub
+ */
+
+static void prim_sub(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    state.forceValue(*args[1], pos);
+    if (args[0]->type() == nFloat || args[1]->type() == nFloat) {
+        v.mkFloat(
+            state.forceFloat(
+                *args[0], pos, "while evaluating the first argument of the subtraction"
+            )
+            - state.forceFloat(
+                *args[1], pos, "while evaluating the second argument of the subtraction"
+            )
+        );
+    } else {
+        v.mkInt(
+            state.forceInt(*args[0], pos, "while evaluating the first argument of the subtraction")
+            - state.forceInt(
+                *args[1], pos, "while evaluating the second argument of the subtraction"
+            )
+        );
+    }
+}
+
+PrimOp primop_sub({
+    .name = "__sub",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return the difference between the numbers *e1* and *e2*.
+    )",
+    .fun = prim_sub,
+});
+}

src/libexpr/primops/attrset.cc

@@ -0,0 +1,698 @@
+#include <string.h>                         // for size_t, memcpy, strcmp
+#include <algorithm>                        // for sort, min, set_difference
+#include <boost/container/small_vector.hpp> // for small_vector
+#include <boost/container/vector.hpp>       // for operator+, operator-
+#include <iterator>                         // for back_insert_iterator
+#include <map>                              // for map, _Rb_tree_iterator
+#include <set>                              // for set
+#include <string_view>                      // for string_view, operator<=>
+#include <utility>                          // for pair
+#include <vector>                           // for vector
+#include "attr-set.hh"                      // for Attr, Bindings, Bindings...
+#include "eval-error.hh"                    // for TypeError, EvalErrorBuilder
+#include "eval.hh"                          // for EvalState, PrimOp, prim_...
+#include "gc-small-vector.hh"               // for SmallValueVector
+#include "pos-idx.hh"                       // for PosIdx, noPos
+#include "primops.hh"                       // for CompareValues, ValueList
+#include "symbol-table.hh"                  // for Symbol, SymbolTable, Sym...
+#include "value.hh"                         // for Value, ValueVector, Valu...
+
+namespace nix {
+
+Bindings::iterator
+getAttr(EvalState & state, Symbol attrSym, Bindings * attrSet, std::string_view errorCtx)
+{
+    Bindings::iterator value = attrSet->find(attrSym);
+    if (value == attrSet->end()) {
+        state.error<TypeError>("attribute '%s' missing", state.symbols[attrSym])
+            .withTrace(noPos, errorCtx)
+            .debugThrow();
+    }
+    return value;
+}
+
+/**
+ * builtins.attrNames
+ */
+
+static void prim_attrNames(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceAttrs(*args[0], pos, "while evaluating the argument passed to builtins.attrNames");
+
+    state.mkList(v, args[0]->attrs->size());
+
+    size_t n = 0;
+    for (auto & i : *args[0]->attrs) {
+        (v.listElems()[n++] = state.allocValue())->mkString(state.symbols[i.name]);
+    }
+
+    std::sort(v.listElems(), v.listElems() + n, [](Value * v1, Value * v2) {
+        return strcmp(v1->string.s, v2->string.s) < 0;
+    });
+}
+
+PrimOp primop_attrNames({
+    .name = "__attrNames",
+    .args = {"set"},
+    .doc = R"(
+      Return the names of the attributes in the set *set* in an
+      alphabetically sorted list. For instance, `builtins.attrNames { y
+      = 1; x = "foo"; }` evaluates to `[ "x" "y" ]`.
+    )",
+    .fun = prim_attrNames,
+});
+
+/**
+ * builtins.attrNames
+ */
+
+static void prim_attrValues(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceAttrs(*args[0], pos, "while evaluating the argument passed to builtins.attrValues");
+
+    state.mkList(v, args[0]->attrs->size());
+
+    unsigned int n = 0;
+    for (auto & i : *args[0]->attrs) {
+        v.listElems()[n++] = (Value *) &i;
+    }
+
+    std::sort(v.listElems(), v.listElems() + n, [&](Value * v1, Value * v2) {
+        std::string_view s1 = state.symbols[((Attr *) v1)->name],
+                         s2 = state.symbols[((Attr *) v2)->name];
+        return s1 < s2;
+    });
+
+    for (unsigned int i = 0; i < n; ++i) {
+        v.listElems()[i] = ((Attr *) v.listElems()[i])->value;
+    }
+}
+
+PrimOp primop_attrValues({
+    .name = "__attrValues",
+    .args = {"set"},
+    .doc = R"(
+      Return the values of the attributes in the set *set* in the order
+      corresponding to the sorted attribute names.
+    )",
+    .fun = prim_attrValues,
+});
+
+/**
+ * builtins.catAttrs
+ */
+
+static void prim_catAttrs(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto attrName = state.symbols.create(state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.catAttrs"
+    ));
+    state.forceList(
+        *args[1], pos, "while evaluating the second argument passed to builtins.catAttrs"
+    );
+
+    SmallValueVector<nonRecursiveStackReservation> res(args[1]->listSize());
+    size_t found = 0;
+
+    for (auto v2 : args[1]->listItems()) {
+        state.forceAttrs(
+            *v2,
+            pos,
+            "while evaluating an element in the list passed as second argument to builtins.catAttrs"
+        );
+        Bindings::iterator i = v2->attrs->find(attrName);
+        if (i != v2->attrs->end()) {
+            res[found++] = i->value;
+        }
+    }
+
+    state.mkList(v, found);
+    for (unsigned int n = 0; n < found; ++n) {
+        v.listElems()[n] = res[n];
+    }
+}
+
+PrimOp primop_catAttrs({
+    .name = "__catAttrs",
+    .args = {"attr", "list"},
+    .doc = R"(
+      Collect each attribute named *attr* from a list of attribute
+      sets.  Attrsets that don't contain the named attribute are
+      ignored. For example,
+
+      ```nix
+      builtins.catAttrs "a" [{a = 1;} {b = 0;} {a = 2;}]
+      ```
+
+      evaluates to `[1 2]`.
+    )",
+    .fun = prim_catAttrs,
+});
+
+/**
+ * builtins.genericClosure
+ */
+
+static void prim_genericClosure(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceAttrs(
+        *args[0], noPos, "while evaluating the first argument passed to builtins.genericClosure"
+    );
+
+    /* Get the start set. */
+    Bindings::iterator startSet = getAttr(
+        state,
+        state.sStartSet,
+        args[0]->attrs,
+        "in the attrset passed as argument to builtins.genericClosure"
+    );
+
+    state.forceList(
+        *startSet->value,
+        noPos,
+        "while evaluating the 'startSet' attribute passed as argument to builtins.genericClosure"
+    );
+
+    ValueList workSet;
+    for (auto elem : startSet->value->listItems()) {
+        workSet.push_back(elem);
+    }
+
+    if (startSet->value->listSize() == 0) {
+        v = *startSet->value;
+        return;
+    }
+
+    /* Get the operator. */
+    Bindings::iterator op = getAttr(
+        state,
+        state.sOperator,
+        args[0]->attrs,
+        "in the attrset passed as argument to builtins.genericClosure"
+    );
+    state.forceFunction(
+        *op->value,
+        noPos,
+        "while evaluating the 'operator' attribute passed as argument to builtins.genericClosure"
+    );
+
+    /* Construct the closure by applying the operator to elements of
+       `workSet', adding the result to `workSet', continuing until
+       no new elements are found. */
+    ValueList res;
+    // `doneKeys' doesn't need to be a GC root, because its values are
+    // reachable from res.
+    auto cmp = CompareValues(
+        state, noPos, "while comparing the `key` attributes of two genericClosure elements"
+    );
+    std::set<Value *, decltype(cmp)> doneKeys(cmp);
+    while (!workSet.empty()) {
+        Value * e = *(workSet.begin());
+        workSet.pop_front();
+
+        state.forceAttrs(
+            *e,
+            noPos,
+            "while evaluating one of the elements generated by (or initially passed to) "
+            "builtins.genericClosure"
+        );
+
+        Bindings::iterator key = getAttr(
+            state,
+            state.sKey,
+            e->attrs,
+            "in one of the attrsets generated by (or initially passed to) builtins.genericClosure"
+        );
+        state.forceValue(*key->value, noPos);
+
+        if (!doneKeys.insert(key->value).second) {
+            continue;
+        }
+        res.push_back(e);
+
+        /* Call the `operator' function with `e' as argument. */
+        Value newElements;
+        state.callFunction(*op->value, 1, &e, newElements, noPos);
+        state.forceList(
+            newElements,
+            noPos,
+            "while evaluating the return value of the `operator` passed to builtins.genericClosure"
+        );
+
+        /* Add the values returned by the operator to the work set. */
+        for (auto elem : newElements.listItems()) {
+            state.forceValue(*elem, noPos); // "while evaluating one one of the elements returned by
+                                            // the `operator` passed to builtins.genericClosure");
+            workSet.push_back(elem);
+        }
+    }
+
+    /* Create the result list. */
+    state.mkList(v, res.size());
+    unsigned int n = 0;
+    for (auto & i : res) {
+        v.listElems()[n++] = i;
+    }
+}
+
+PrimOp primop_genericClosure(PrimOp{
+    .name = "__genericClosure",
+    .args = {"attrset"},
+    .arity = 1,
+    .doc = R"(
+      Take an *attrset* with values named `startSet` and `operator` in order to
+      return a *list of attrsets* by starting with the `startSet` and recursively
+      applying the `operator` function to each `item`. The *attrsets* in the
+      `startSet` and the *attrsets* produced by `operator` must contain a value
+      named `key` which is comparable. The result is produced by calling `operator`
+      for each `item` with a value for `key` that has not been called yet including
+      newly produced `item`s. The function terminates when no new `item`s are
+      produced. The resulting *list of attrsets* contains only *attrsets* with a
+      unique key. For example,
+
+      ```
+      builtins.genericClosure {
+        startSet = [ {key = 5;} ];
+        operator = item: [{
+          key = if (item.key / 2 ) * 2 == item.key
+               then item.key / 2
+               else 3 * item.key + 1;
+        }];
+      }
+      ```
+      evaluates to
+      ```
+      [ { key = 5; } { key = 16; } { key = 8; } { key = 4; } { key = 2; } { key = 1; } ]
+      ```
+      )",
+    .fun = prim_genericClosure,
+});
+
+/**
+ * builtins.getAttr
+ */
+
+void prim_getAttr(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto attr = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.getAttr"
+    );
+    state.forceAttrs(
+        *args[1], pos, "while evaluating the second argument passed to builtins.getAttr"
+    );
+    Bindings::iterator i = getAttr(
+        state,
+        state.symbols.create(attr),
+        args[1]->attrs,
+        "in the attribute set under consideration"
+    );
+    // !!! add to stack trace?
+    if (state.countCalls && i->pos) {
+        state.attrSelects[i->pos]++;
+    }
+    state.forceValue(*i->value, pos);
+    v = *i->value;
+}
+
+PrimOp primop_getAttr({
+    .name = "__getAttr",
+    .args = {"s", "set"},
+    .doc = R"(
+      `getAttr` returns the attribute named *s* from *set*. Evaluation
+      aborts if the attribute doesn’t exist. This is a dynamic version of
+      the `.` operator, since *s* is an expression rather than an
+      identifier.
+    )",
+    .fun = prim_getAttr,
+});
+
+/**
+ * builtins.groupBy
+ */
+
+static void prim_groupBy(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.groupBy"
+    );
+    state.forceList(
+        *args[1], pos, "while evaluating the second argument passed to builtins.groupBy"
+    );
+
+    ValueVectorMap attrs;
+
+    for (auto vElem : args[1]->listItems()) {
+        Value res;
+        state.callFunction(*args[0], *vElem, res, pos);
+        auto name = state.forceStringNoCtx(
+            res,
+            pos,
+            "while evaluating the return value of the grouping function passed to builtins.groupBy"
+        );
+        auto sym = state.symbols.create(name);
+        auto vector = attrs.try_emplace(sym, ValueVector()).first;
+        vector->second.push_back(vElem);
+    }
+
+    auto attrs2 = state.buildBindings(attrs.size());
+
+    for (auto & i : attrs) {
+        auto & list = attrs2.alloc(i.first);
+        auto size = i.second.size();
+        state.mkList(list, size);
+        memcpy(list.listElems(), i.second.data(), sizeof(Value *) * size);
+    }
+
+    v.mkAttrs(attrs2.alreadySorted());
+}
+
+PrimOp primop_groupBy({
+    .name = "__groupBy",
+    .args = {"f", "list"},
+    .doc = R"(
+      Groups elements of *list* together by the string returned from the
+      function *f* called on each element. It returns an attribute set
+      where each attribute value contains the elements of *list* that are
+      mapped to the same corresponding attribute name returned by *f*.
+
+      For example,
+
+      ```nix
+      builtins.groupBy (builtins.substring 0 1) ["foo" "bar" "baz"]
+      ```
+
+      evaluates to
+
+      ```nix
+      { b = [ "bar" "baz" ]; f = [ "foo" ]; }
+      ```
+    )",
+    .fun = prim_groupBy,
+});
+
+/**
+ * builtins.hasAttr
+ */
+
+static void prim_hasAttr(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto attr = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.hasAttr"
+    );
+    state.forceAttrs(
+        *args[1], pos, "while evaluating the second argument passed to builtins.hasAttr"
+    );
+    v.mkBool(args[1]->attrs->find(state.symbols.create(attr)) != args[1]->attrs->end());
+}
+
+PrimOp primop_hasAttr({
+    .name = "__hasAttr",
+    .args = {"s", "set"},
+    .doc = R"(
+      `hasAttr` returns `true` if *set* has an attribute named *s*, and
+      `false` otherwise. This is a dynamic version of the `?` operator,
+      since *s* is an expression rather than an identifier.
+    )",
+    .fun = prim_hasAttr,
+});
+
+/**
+ * builtins.intersectAttrs
+ */
+
+static void prim_intersectAttrs(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceAttrs(
+        *args[0], pos, "while evaluating the first argument passed to builtins.intersectAttrs"
+    );
+    state.forceAttrs(
+        *args[1], pos, "while evaluating the second argument passed to builtins.intersectAttrs"
+    );
+
+    Bindings & left = *args[0]->attrs;
+    Bindings & right = *args[1]->attrs;
+
+    auto attrs = state.buildBindings(std::min(left.size(), right.size()));
+
+    // The current implementation has good asymptotic complexity and is reasonably
+    // simple. Further optimization may be possible, but does not seem productive,
+    // considering the state of eval performance in 2022.
+    //
+    // I have looked for reusable and/or standard solutions and these are my
+    // findings:
+    //
+    // STL
+    // ===
+    // std::set_intersection is not suitable, as it only performs a simultaneous
+    // linear scan; not taking advantage of random access. This is O(n + m), so
+    // linear in the largest set, which is not acceptable for callPackage in Nixpkgs.
+    //
+    // Simultaneous scan, with alternating simple binary search
+    // ===
+    // One alternative algorithm scans the attrsets simultaneously, jumping
+    // forward using `lower_bound` in case of inequality. This should perform
+    // well on very similar sets, having a local and predictable access pattern.
+    // On dissimilar sets, it seems to need more comparisons than the current
+    // algorithm, as few consecutive attrs match. `lower_bound` could take
+    // advantage of the decreasing remaining search space, but this causes
+    // the medians to move, which can mean that they don't stay in the cache
+    // like they would with the current naive `find`.
+    //
+    // Double binary search
+    // ===
+    // The optimal algorithm may be "Double binary search", which doesn't
+    // scan at all, but rather divides both sets simultaneously.
+    // See "Fast Intersection Algorithms for Sorted Sequences" by Baeza-Yates et al.
+    // https://cs.uwaterloo.ca/~ajsaling/papers/intersection_alg_app10.pdf
+    // The only downsides I can think of are not having a linear access pattern
+    // for similar sets, and having to maintain a more intricate algorithm.
+    //
+    // Adaptive
+    // ===
+    // Finally one could run try a simultaneous scan, count misses and fall back
+    // to double binary search when the counter hit some threshold and/or ratio.
+
+    if (left.size() < right.size()) {
+        for (auto & l : left) {
+            Bindings::iterator r = right.find(l.name);
+            if (r != right.end()) {
+                attrs.insert(*r);
+            }
+        }
+    } else {
+        for (auto & r : right) {
+            Bindings::iterator l = left.find(r.name);
+            if (l != left.end()) {
+                attrs.insert(r);
+            }
+        }
+    }
+
+    v.mkAttrs(attrs.alreadySorted());
+}
+
+PrimOp primop_intersectAttrs({
+    .name = "__intersectAttrs",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Return a set consisting of the attributes in the set *e2* which have the
+      same name as some attribute in *e1*.
+
+      Performs in O(*n* log *m*) where *n* is the size of the smaller set and *m* the larger set's size.
+    )",
+    .fun = prim_intersectAttrs,
+});
+
+/**
+ * builtins.mapAttrs
+ */
+
+static void prim_mapAttrs(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceAttrs(
+        *args[1], pos, "while evaluating the second argument passed to builtins.mapAttrs"
+    );
+
+    auto attrs = state.buildBindings(args[1]->attrs->size());
+
+    for (auto & i : *args[1]->attrs) {
+        Value * vName = state.allocValue();
+        Value * vFun2 = state.allocValue();
+        vName->mkString(state.symbols[i.name]);
+        vFun2->mkApp(args[0], vName);
+        attrs.alloc(i.name).mkApp(vFun2, i.value);
+    }
+
+    v.mkAttrs(attrs.alreadySorted());
+}
+
+PrimOp primop_mapAttrs({
+    .name = "__mapAttrs",
+    .args = {"f", "attrset"},
+    .doc = R"(
+      Apply function *f* to every element of *attrset*. For example,
+
+      ```nix
+      builtins.mapAttrs (name: value: value * 10) { a = 1; b = 2; }
+      ```
+
+      evaluates to `{ a = 10; b = 20; }`.
+    )",
+    .fun = prim_mapAttrs,
+});
+
+/**
+ * builtins.mapAttrs
+ */
+
+static void prim_removeAttrs(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceAttrs(
+        *args[0], pos, "while evaluating the first argument passed to builtins.removeAttrs"
+    );
+    state.forceList(
+        *args[1], pos, "while evaluating the second argument passed to builtins.removeAttrs"
+    );
+
+    /* Get the attribute names to be removed.
+       We keep them as Attrs instead of Symbols so std::set_difference
+       can be used to remove them from attrs[0]. */
+    // 64: large enough to fit the attributes of a derivation
+    boost::container::small_vector<Attr, 64> names;
+    names.reserve(args[1]->listSize());
+    for (auto elem : args[1]->listItems()) {
+        state.forceStringNoCtx(
+            *elem,
+            pos,
+            "while evaluating the values of the second argument passed to builtins.removeAttrs"
+        );
+        names.emplace_back(state.symbols.create(elem->string.s), nullptr);
+    }
+    std::sort(names.begin(), names.end());
+
+    /* Copy all attributes not in that set.  Note that we don't need
+       to sort v.attrs because it's a subset of an already sorted
+       vector. */
+    auto attrs = state.buildBindings(args[0]->attrs->size());
+    std::set_difference(
+        args[0]->attrs->begin(),
+        args[0]->attrs->end(),
+        names.begin(),
+        names.end(),
+        std::back_inserter(attrs)
+    );
+    v.mkAttrs(attrs.alreadySorted());
+}
+
+PrimOp primop_removeAttrs({
+    .name = "removeAttrs",
+    .args = {"set", "list"},
+    .doc = R"(
+      Remove the attributes listed in *list* from *set*. The attributes
+      don’t have to exist in *set*. For instance,
+
+      ```nix
+      removeAttrs { x = 1; y = 2; z = 3; } [ "a" "x" "z" ]
+      ```
+
+      evaluates to `{ y = 2; }`.
+    )",
+    .fun = prim_removeAttrs,
+});
+
+/**
+ * builtins.zipAttrsWith
+ */
+
+static void prim_zipAttrsWith(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    // we will first count how many values are present for each given key.
+    // we then allocate a single attrset and pre-populate it with lists of
+    // appropriate sizes, stash the pointers to the list elements of each,
+    // and populate the lists. after that we replace the list in the every
+    // attribute with the merge function application. this way we need not
+    // use (slightly slower) temporary storage the GC does not know about.
+
+    std::map<Symbol, std::pair<size_t, Value **>> attrsSeen;
+
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.zipAttrsWith"
+    );
+    state.forceList(
+        *args[1], pos, "while evaluating the second argument passed to builtins.zipAttrsWith"
+    );
+    const auto listSize = args[1]->listSize();
+    const auto listElems = args[1]->listElems();
+
+    for (unsigned int n = 0; n < listSize; ++n) {
+        Value * vElem = listElems[n];
+        state.forceAttrs(
+            *vElem,
+            noPos,
+            "while evaluating a value of the list passed as second argument to "
+            "builtins.zipAttrsWith"
+        );
+        for (auto & attr : *vElem->attrs) {
+            attrsSeen[attr.name].first++;
+        }
+    }
+
+    auto attrs = state.buildBindings(attrsSeen.size());
+    for (auto & [sym, elem] : attrsSeen) {
+        auto & list = attrs.alloc(sym);
+        state.mkList(list, elem.first);
+        elem.second = list.listElems();
+    }
+    v.mkAttrs(attrs.alreadySorted());
+
+    for (unsigned int n = 0; n < listSize; ++n) {
+        Value * vElem = listElems[n];
+        for (auto & attr : *vElem->attrs) {
+            *attrsSeen[attr.name].second++ = attr.value;
+        }
+    }
+
+    for (auto & attr : *v.attrs) {
+        auto name = state.allocValue();
+        name->mkString(state.symbols[attr.name]);
+        auto call1 = state.allocValue();
+        call1->mkApp(args[0], name);
+        auto call2 = state.allocValue();
+        call2->mkApp(call1, attr.value);
+        attr.value = call2;
+    }
+}
+
+PrimOp primop_zipAttrsWith({
+    .name = "__zipAttrsWith",
+    .args = {"f", "list"},
+    .doc = R"(
+      Transpose a list of attribute sets into an attribute set of lists,
+      then apply `mapAttrs`.
+
+      `f` receives two arguments: the attribute name and a non-empty
+      list of all values encountered for that attribute name.
+
+      The result is an attribute set where the attribute names are the
+      union of the attribute names in each element of `list`. The attribute
+      values are the return values of `f`.
+
+      ```nix
+      builtins.zipAttrsWith
+        (name: values: { inherit name values; })
+        [ { a = "x"; } { a = "y"; b = "z"; } ]
+      ```
+
+      evaluates to
+
+      ```
+      {
+        a = { name = "a"; values = [ "x" "y" ]; };
+        b = { name = "b"; values = [ "z" ]; };
+      }
+      ```
+    )",
+    .fun = prim_zipAttrsWith,
+});
+
+}

src/libexpr/primops/context.cc

@@ -1,142 +1,191 @@
-#include "primops.hh"
-#include "eval-inline.hh"
-#include "derivations.hh"
-#include "store-api.hh"
+#include <map>                        // for map, _Rb_tree_const_iterator
+#include <memory>                     // for shared_ptr
+#include <string>                     // for basic_string, operator<=>, char...
+#include <utility>                    // for move, pair
+#include <variant>                    // for visit, get_if
+#include <vector>                     // for vector
+#include "attr-set.hh"                // for Attr, Bindings, BindingsBuilder
+#include "config.hh"                  // for Setting, ExperimentalFeatureSet...
+#include "derivations.hh"             // for isDerivation
+#include "derived-path.hh"            // for DerivedPath, makeConstantStoreP...
+#include "downstream-placeholder.hh"  // for DownstreamPlaceholder
+#include "error.hh"                   // for Error
+#include "eval-error.hh"              // for EvalError, EvalErrorBuilder
+#include "eval-settings.hh"           // for EvalSettings, evalSettings
+#include "eval.hh"                    // for EvalState, PrimOp
+#include "experimental-features.hh"   // for ExperimentalFeature, Xp
+#include "fmt.hh"                     // for HintFmt
+#include "globals.hh"                 // for Settings, settings
+#include "outputs-spec.hh"            // for OutputsSpec
+#include "path.hh"                    // for StorePath, StorePathSet
+#include "pos-idx.hh"                 // for PosIdx, noPos
+#include "ref.hh"                     // for ref
+#include "store-api.hh"               // for Store, resolveDerivedPath, copy...
+#include "symbol-table.hh"            // for SymbolStr, Symbol, SymbolTable
+#include "types.hh"                   // for StringMap, BackedStringView
+#include "util.hh"                    // for overloaded, enumerate
+#include "value.hh"                   // for Value
+#include "value/context.hh"           // for NixStringContextElem, NixString...
 
 namespace nix {
 
-static void prim_unsafeDiscardStringContext(EvalState & state, const PosIdx pos, Value * * args, Value & v)
+StringMap EvalState::realiseContext(const NixStringContext & context)
 {
-    NixStringContext context;
-    auto s = state.coerceToString(pos, *args[0], context, "while evaluating the argument passed to builtins.unsafeDiscardStringContext");
-    v.mkString(*s);
-}
+    std::vector<DerivedPath::Built> drvs;
+    StringMap res;
 
-static RegisterPrimOp primop_unsafeDiscardStringContext({
-    .name = "__unsafeDiscardStringContext",
-    .arity = 1,
-    .fun = prim_unsafeDiscardStringContext
-});
+    for (auto & c : context) {
+        auto ensureValid = [&](const StorePath & p) {
+            if (!store->isValidPath(p)) {
+                error<InvalidPathError>(store->printStorePath(p)).debugThrow();
+            }
+        };
+        std::visit(
+            overloaded{
+                [&](const NixStringContextElem::Built & b) {
+                    drvs.push_back(DerivedPath::Built{
+                        .drvPath = b.drvPath,
+                        .outputs = OutputsSpec::Names{b.output},
+                    });
+                    ensureValid(b.drvPath->getBaseStorePath());
+                },
+                [&](const NixStringContextElem::Opaque & o) {
+                    auto ctxS = store->printStorePath(o.path);
+                    res.insert_or_assign(ctxS, ctxS);
+                    ensureValid(o.path);
+                },
+                [&](const NixStringContextElem::DrvDeep & d) {
+                    /* Treat same as Opaque */
+                    auto ctxS = store->printStorePath(d.drvPath);
+                    res.insert_or_assign(ctxS, ctxS);
+                    ensureValid(d.drvPath);
+                },
+            },
+            c.raw
+        );
+    }
 
+    if (drvs.empty()) {
+        return {};
+    }
 
-static void prim_hasContext(EvalState & state, const PosIdx pos, Value * * args, Value & v)
-{
-    NixStringContext context;
-    state.forceString(*args[0], context, pos, "while evaluating the argument passed to builtins.hasContext");
-    v.mkBool(!context.empty());
-}
+    if (!evalSettings.enableImportFromDerivation) {
+        error<EvalError>(
+            "cannot build '%1%' during evaluation because the option "
+            "'allow-import-from-derivation' is disabled",
+            drvs.begin()->to_string(*store)
+        )
+            .debugThrow();
+    }
 
-static RegisterPrimOp primop_hasContext({
-    .name = "__hasContext",
-    .args = {"s"},
-    .doc = R"(
-      Return `true` if string *s* has a non-empty context.
-      The context can be obtained with
-      [`getContext`](#builtins-getContext).
+    /* Build/substitute the context. */
+    std::vector<DerivedPath> buildReqs;
+    for (auto & d : drvs) {
+        buildReqs.emplace_back(DerivedPath{d});
+    }
+    buildStore->buildPaths(buildReqs, bmNormal, store);
 
-      > **Example**
-      >
-      > Many operations require a string context to be empty because they are intended only to work with "regular" strings, and also to help users avoid unintentionally losing track of string context elements.
-      > `builtins.hasContext` can help create better domain-specific errors in those case.
-      >
-      > ```nix
-      > name: meta:
-      >
-      > if builtins.hasContext name
-      > then throw "package name cannot contain string context"
-      > else { ${name} = meta; }
-      > ```
-    )",
-    .fun = prim_hasContext
-});
+    StorePathSet outputsToCopyAndAllow;
 
+    for (auto & drv : drvs) {
+        auto outputs = resolveDerivedPath(*buildStore, drv, &*store);
+        for (auto & [outputName, outputPath] : outputs) {
+            outputsToCopyAndAllow.insert(outputPath);
 
-static void prim_unsafeDiscardOutputDependency(EvalState & state, const PosIdx pos, Value * * args, Value & v)
-{
-    NixStringContext context;
-    auto s = state.coerceToString(pos, *args[0], context, "while evaluating the argument passed to builtins.unsafeDiscardOutputDependency");
-
-    NixStringContext context2;
-    for (auto && c : context) {
-        if (auto * ptr = std::get_if<NixStringContextElem::DrvDeep>(&c.raw)) {
-            context2.emplace(NixStringContextElem::Opaque {
-                .path = ptr->drvPath
-            });
-        } else {
-            /* Can reuse original item */
-            context2.emplace(std::move(c).raw);
+            /* Get all the output paths corresponding to the placeholders we had */
+            if (experimentalFeatureSettings.isEnabled(Xp::CaDerivations)) {
+                res.insert_or_assign(
+                    DownstreamPlaceholder::fromSingleDerivedPathBuilt(SingleDerivedPath::Built{
+                                                                          .drvPath = drv.drvPath,
+                                                                          .output = outputName,
+                                                                      })
+                        .render(),
+                    buildStore->printStorePath(outputPath)
+                );
+            }
         }
     }
 
-    v.mkString(*s, context2);
+    if (store != buildStore) {
+        copyClosure(*buildStore, *store, outputsToCopyAndAllow);
+    }
+    if (allowedPaths) {
+        for (auto & outputPath : outputsToCopyAndAllow) {
+            /* Add the output of this derivations to the allowed
+               paths. */
+            allowPath(outputPath);
+        }
+    }
+
+    return res;
 }
 
-static RegisterPrimOp primop_unsafeDiscardOutputDependency({
-    .name = "__unsafeDiscardOutputDependency",
-    .args = {"s"},
-    .doc = R"(
-      Create a copy of the given string where every "derivation deep" string context element is turned into a constant string context element.
+/**
+ * builtins.addDrvOutputDependencies
+ */
 
-      This is the opposite of [`builtins.addDrvOutputDependencies`](#builtins-addDrvOutputDependencies).
-
-      This is unsafe because it allows us to "forget" store objects we would have otherwise refered to with the string context,
-      whereas Nix normally tracks all dependencies consistently.
-      Safe operations "grow" but never "shrink" string contexts.
-      [`builtins.addDrvOutputDependencies`] in contrast is safe because "derivation deep" string context element always refers to the underlying derivation (among many more things).
-      Replacing a constant string context element with a "derivation deep" element is a safe operation that just enlargens the string context without forgetting anything.
-
-      [`builtins.addDrvOutputDependencies`]: #builtins-addDrvOutputDependencies
-    )",
-    .fun = prim_unsafeDiscardOutputDependency
-});
-
-
-static void prim_addDrvOutputDependencies(EvalState & state, const PosIdx pos, Value * * args, Value & v)
+static void
+prim_addDrvOutputDependencies(EvalState & state, const PosIdx pos, Value ** args, Value & v)
 {
     NixStringContext context;
-    auto s = state.coerceToString(pos, *args[0], context, "while evaluating the argument passed to builtins.addDrvOutputDependencies");
+    auto s = state.coerceToString(
+        pos,
+        *args[0],
+        context,
+        "while evaluating the argument passed to builtins.addDrvOutputDependencies"
+    );
 
-	auto contextSize = context.size();
+    auto contextSize = context.size();
     if (contextSize != 1) {
-        state.error<EvalError>(
-            "context of string '%s' must have exactly one element, but has %d",
-            *s,
-            contextSize
-        ).atPos(pos).debugThrow();
+        state
+            .error<EvalError>(
+                "context of string '%s' must have exactly one element, but has %d", *s, contextSize
+            )
+            .atPos(pos)
+            .debugThrow();
     }
-    NixStringContext context2 {
-        (NixStringContextElem { std::visit(overloaded {
-            [&](const NixStringContextElem::Opaque & c) -> NixStringContextElem::DrvDeep {
-                if (!c.path.isDerivation()) {
-                    state.error<EvalError>(
-                        "path '%s' is not a derivation",
-                        state.store->printStorePath(c.path)
-                    ).atPos(pos).debugThrow();
-                }
-                return NixStringContextElem::DrvDeep {
-                    .drvPath = c.path,
-                };
+    NixStringContext context2{
+        (NixStringContextElem{std::visit(
+            overloaded{
+                [&](const NixStringContextElem::Opaque & c) -> NixStringContextElem::DrvDeep {
+                    if (!c.path.isDerivation()) {
+                        state
+                            .error<EvalError>(
+                                "path '%s' is not a derivation", state.store->printStorePath(c.path)
+                            )
+                            .atPos(pos)
+                            .debugThrow();
+                    }
+                    return NixStringContextElem::DrvDeep{
+                        .drvPath = c.path,
+                    };
+                },
+                [&](const NixStringContextElem::Built & c) -> NixStringContextElem::DrvDeep {
+                    state
+                        .error<EvalError>(
+                            "`addDrvOutputDependencies` can only act on derivations, not on a "
+                            "derivation output such as '%1%'",
+                            c.output
+                        )
+                        .atPos(pos)
+                        .debugThrow();
+                },
+                [&](const NixStringContextElem::DrvDeep & c) -> NixStringContextElem::DrvDeep {
+                    /* Reuse original item because we want this to be idempotent. */
+                    return std::move(c);
+                },
             },
-            [&](const NixStringContextElem::Built & c) -> NixStringContextElem::DrvDeep {
-                state.error<EvalError>(
-                    "`addDrvOutputDependencies` can only act on derivations, not on a derivation output such as '%1%'",
-                    c.output
-                ).atPos(pos).debugThrow();
-            },
-            [&](const NixStringContextElem::DrvDeep & c) -> NixStringContextElem::DrvDeep {
-                /* Reuse original item because we want this to be idempotent. */
-                return std::move(c);
-            },
-        }, context.begin()->raw) }),
+            context.begin()->raw
+        )}),
     };
 
     v.mkString(*s, context2);
 }
 
-static RegisterPrimOp primop_addDrvOutputDependencies({
-    .name = "__addDrvOutputDependencies",
-    .args = {"s"},
-    .doc = R"(
+PrimOp primop_addDrvOutputDependencies(
+    {.name = "__addDrvOutputDependencies",
+     .args = {"s"},
+     .doc = R"(
       Create a copy of the given string where a single constant string context element is turned into a "derivation deep" string context element.
 
       The store path that is the constant string context element should point to a valid derivation, and end in `.drv`.
@@ -146,9 +195,153 @@ static RegisterPrimOp primop_addDrvOutputDependencies({
 
       This is the opposite of [`builtins.unsafeDiscardOutputDependency`](#builtins-unsafeDiscardOutputDependency).
     )",
-    .fun = prim_addDrvOutputDependencies
+     .fun = prim_addDrvOutputDependencies}
+);
+
+/**
+ * builtins.addErrorContext
+ */
+
+static void prim_addErrorContext(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    try {
+        state.forceValue(*args[1], pos);
+        v = *args[1];
+    } catch (Error & e) {
+        NixStringContext context;
+        auto message =
+            state
+                .coerceToString(
+                    pos,
+                    *args[0],
+                    context,
+                    "while evaluating the error message passed to builtins.addErrorContext",
+                    false,
+                    false
+                )
+                .toOwned();
+        e.addTrace(nullptr, HintFmt(message));
+        throw;
+    }
+}
+
+PrimOp primop_addErrorContext(PrimOp{
+    .name = "__addErrorContext",
+    .arity = 2,
+    .fun = prim_addErrorContext,
 });
 
+/**
+ * builtins.appendContext
+ */
+
+static void prim_appendContext(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto orig = state.forceString(
+        *args[0],
+        context,
+        noPos,
+        "while evaluating the first argument passed to builtins.appendContext"
+    );
+
+    state.forceAttrs(
+        *args[1], pos, "while evaluating the second argument passed to builtins.appendContext"
+    );
+
+    auto sPath = state.symbols.create("path");
+    auto sAllOutputs = state.symbols.create("allOutputs");
+    for (auto & i : *args[1]->attrs) {
+        const auto & name = state.symbols[i.name];
+        if (!state.store->isStorePath(name)) {
+            state.error<EvalError>("context key '%s' is not a store path", name)
+                .atPos(i.pos)
+                .debugThrow();
+        }
+        auto namePath = state.store->parseStorePath(name);
+        if (!settings.readOnlyMode) {
+            state.store->ensurePath(namePath);
+        }
+        state.forceAttrs(*i.value, i.pos, "while evaluating the value of a string context");
+        auto iter = i.value->attrs->find(sPath);
+        if (iter != i.value->attrs->end()) {
+            if (state.forceBool(
+                    *iter->value,
+                    iter->pos,
+                    "while evaluating the `path` attribute of a string context"
+                ))
+            {
+                context.emplace(NixStringContextElem::Opaque{
+                    .path = namePath,
+                });
+            }
+        }
+
+        iter = i.value->attrs->find(sAllOutputs);
+        if (iter != i.value->attrs->end()) {
+            if (state.forceBool(
+                    *iter->value,
+                    iter->pos,
+                    "while evaluating the `allOutputs` attribute of a string context"
+                ))
+            {
+                if (!isDerivation(name)) {
+                    state
+                        .error<EvalError>(
+                            "tried to add all-outputs context of %s, which is not a derivation, to "
+                            "a string",
+                            name
+                        )
+                        .atPos(i.pos)
+                        .debugThrow();
+                }
+                context.emplace(NixStringContextElem::DrvDeep{
+                    .drvPath = namePath,
+                });
+            }
+        }
+
+        iter = i.value->attrs->find(state.sOutputs);
+        if (iter != i.value->attrs->end()) {
+            state.forceList(
+                *iter->value,
+                iter->pos,
+                "while evaluating the `outputs` attribute of a string context"
+            );
+            if (iter->value->listSize() && !isDerivation(name)) {
+                state
+                    .error<EvalError>(
+                        "tried to add derivation output context of %s, which is not a derivation, "
+                        "to a string",
+                        name
+                    )
+                    .atPos(i.pos)
+                    .debugThrow();
+            }
+            for (auto elem : iter->value->listItems()) {
+                auto outputName = state.forceStringNoCtx(
+                    *elem, iter->pos, "while evaluating an output name within a string context"
+                );
+                context.emplace(NixStringContextElem::Built{
+                    .drvPath = makeConstantStorePathRef(namePath),
+                    .output = std::string{outputName},
+                });
+            }
+        }
+    }
+
+    v.mkString(orig, context);
+}
+
+PrimOp primop_appendContext({
+    .name = "__appendContext",
+    .arity = 2,
+    .fun = prim_appendContext,
+});
+
+/**
+ * builtins.getContext
+ */
 
 /* Extract the context of a string as a structured Nix value.
 
@@ -169,31 +362,37 @@ static RegisterPrimOp primop_addDrvOutputDependencies({
    Note that for a given path any combination of the above attributes
    may be present.
 */
-static void prim_getContext(EvalState & state, const PosIdx pos, Value * * args, Value & v)
+static void prim_getContext(EvalState & state, const PosIdx pos, Value ** args, Value & v)
 {
-    struct ContextInfo {
+    struct ContextInfo
+    {
         bool path = false;
         bool allOutputs = false;
         Strings outputs;
     };
     NixStringContext context;
-    state.forceString(*args[0], context, pos, "while evaluating the argument passed to builtins.getContext");
+    state.forceString(
+        *args[0], context, pos, "while evaluating the argument passed to builtins.getContext"
+    );
     auto contextInfos = std::map<StorePath, ContextInfo>();
     for (auto && i : context) {
-        std::visit(overloaded {
-            [&](NixStringContextElem::DrvDeep && d) {
-                contextInfos[std::move(d.drvPath)].allOutputs = true;
+        std::visit(
+            overloaded{
+                [&](NixStringContextElem::DrvDeep && d) {
+                    contextInfos[std::move(d.drvPath)].allOutputs = true;
+                },
+                [&](NixStringContextElem::Built && b) {
+                    // FIXME should eventually show string context as is, no
+                    // resolving here.
+                    auto drvPath = resolveDerivedPath(*state.store, *b.drvPath);
+                    contextInfos[std::move(drvPath)].outputs.emplace_back(std::move(b.output));
+                },
+                [&](NixStringContextElem::Opaque && o) {
+                    contextInfos[std::move(o.path)].path = true;
+                },
             },
-            [&](NixStringContextElem::Built && b) {
-                // FIXME should eventually show string context as is, no
-                // resolving here.
-                auto drvPath = resolveDerivedPath(*state.store, *b.drvPath);
-                contextInfos[std::move(drvPath)].outputs.emplace_back(std::move(b.output));
-            },
-            [&](NixStringContextElem::Opaque && o) {
-                contextInfos[std::move(o.path)].path = true;
-            },
-        }, ((NixStringContextElem &&) i).raw);
+            ((NixStringContextElem &&) i).raw
+        );
     }
 
     auto attrs = state.buildBindings(contextInfos.size());
@@ -202,15 +401,18 @@ static void prim_getContext(EvalState & state, const PosIdx pos, Value * * args,
     auto sAllOutputs = state.symbols.create("allOutputs");
     for (const auto & info : contextInfos) {
         auto infoAttrs = state.buildBindings(3);
-        if (info.second.path)
+        if (info.second.path) {
             infoAttrs.alloc(sPath).mkBool(true);
-        if (info.second.allOutputs)
+        }
+        if (info.second.allOutputs) {
             infoAttrs.alloc(sAllOutputs).mkBool(true);
+        }
         if (!info.second.outputs.empty()) {
             auto & outputsVal = infoAttrs.alloc(state.sOutputs);
             state.mkList(outputsVal, info.second.outputs.size());
-            for (const auto & [i, output] : enumerate(info.second.outputs))
+            for (const auto & [i, output] : enumerate(info.second.outputs)) {
                 (outputsVal.listElems()[i] = state.allocValue())->mkString(output);
+            }
         }
         attrs.alloc(state.store->printStorePath(info.first)).mkAttrs(infoAttrs);
     }
@@ -218,7 +420,7 @@ static void prim_getContext(EvalState & state, const PosIdx pos, Value * * args,
     v.mkAttrs(attrs);
 }
 
-static RegisterPrimOp primop_getContext({
+PrimOp primop_getContext({
     .name = "__getContext",
     .args = {"s"},
     .doc = R"(
@@ -240,84 +442,114 @@ static RegisterPrimOp primop_getContext({
       { "/nix/store/arhvjaf6zmlyn8vh8fgn55rpwnxq0n7l-a.drv" = { outputs = [ "out" ]; }; }
       ```
     )",
-    .fun = prim_getContext
+    .fun = prim_getContext,
 });
 
+/**
+ * builtins.hasContext
+ */
 
-/* Append the given context to a given string.
-
-   See the commentary above unsafeGetContext for details of the
-   context representation.
-*/
-static void prim_appendContext(EvalState & state, const PosIdx pos, Value * * args, Value & v)
+static void prim_hasContext(EvalState & state, const PosIdx pos, Value ** args, Value & v)
 {
     NixStringContext context;
-    auto orig = state.forceString(*args[0], context, noPos, "while evaluating the first argument passed to builtins.appendContext");
+    state.forceString(
+        *args[0], context, pos, "while evaluating the argument passed to builtins.hasContext"
+    );
+    v.mkBool(!context.empty());
+}
 
-    state.forceAttrs(*args[1], pos, "while evaluating the second argument passed to builtins.appendContext");
+PrimOp primop_hasContext(
+    {.name = "__hasContext",
+     .args = {"s"},
+     .doc = R"(
+      Return `true` if string *s* has a non-empty context.
+      The context can be obtained with
+      [`getContext`](#builtins-getContext).
 
-    auto sPath = state.symbols.create("path");
-    auto sAllOutputs = state.symbols.create("allOutputs");
-    for (auto & i : *args[1]->attrs) {
-        const auto & name = state.symbols[i.name];
-        if (!state.store->isStorePath(name))
-            state.error<EvalError>(
-                "context key '%s' is not a store path",
-                name
-            ).atPos(i.pos).debugThrow();
-        auto namePath = state.store->parseStorePath(name);
-        if (!settings.readOnlyMode)
-            state.store->ensurePath(namePath);
-        state.forceAttrs(*i.value, i.pos, "while evaluating the value of a string context");
-        auto iter = i.value->attrs->find(sPath);
-        if (iter != i.value->attrs->end()) {
-            if (state.forceBool(*iter->value, iter->pos, "while evaluating the `path` attribute of a string context"))
-                context.emplace(NixStringContextElem::Opaque {
-                    .path = namePath,
-                });
-        }
+      > **Example**
+      >
+      > Many operations require a string context to be empty because they are intended only to work with "regular" strings, and also to help users avoid unintentionally losing track of string context elements.
+      > `builtins.hasContext` can help create better domain-specific errors in those case.
+      >
+      > ```nix
+      > name: meta:
+      >
+      > if builtins.hasContext name
+      > then throw "package name cannot contain string context"
+      > else { ${name} = meta; }
+      > ```
+    )",
+     .fun = prim_hasContext}
+);
 
-        iter = i.value->attrs->find(sAllOutputs);
-        if (iter != i.value->attrs->end()) {
-            if (state.forceBool(*iter->value, iter->pos, "while evaluating the `allOutputs` attribute of a string context")) {
-                if (!isDerivation(name)) {
-                    state.error<EvalError>(
-                        "tried to add all-outputs context of %s, which is not a derivation, to a string",
-                        name
-                    ).atPos(i.pos).debugThrow();
-                }
-                context.emplace(NixStringContextElem::DrvDeep {
-                    .drvPath = namePath,
-                });
-            }
-        }
+/**
+ * builtins.unsafeDiscardOutputDependency
+ */
 
-        iter = i.value->attrs->find(state.sOutputs);
-        if (iter != i.value->attrs->end()) {
-            state.forceList(*iter->value, iter->pos, "while evaluating the `outputs` attribute of a string context");
-            if (iter->value->listSize() && !isDerivation(name)) {
-                state.error<EvalError>(
-                    "tried to add derivation output context of %s, which is not a derivation, to a string",
-                    name
-                ).atPos(i.pos).debugThrow();
-            }
-            for (auto elem : iter->value->listItems()) {
-                auto outputName = state.forceStringNoCtx(*elem, iter->pos, "while evaluating an output name within a string context");
-                context.emplace(NixStringContextElem::Built {
-                    .drvPath = makeConstantStorePathRef(namePath),
-                    .output = std::string { outputName },
-                });
-            }
+static void
+prim_unsafeDiscardOutputDependency(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto s = state.coerceToString(
+        pos,
+        *args[0],
+        context,
+        "while evaluating the argument passed to builtins.unsafeDiscardOutputDependency"
+    );
+
+    NixStringContext context2;
+    for (auto && c : context) {
+        if (auto * ptr = std::get_if<NixStringContextElem::DrvDeep>(&c.raw)) {
+            context2.emplace(NixStringContextElem::Opaque{.path = ptr->drvPath});
+        } else {
+            /* Can reuse original item */
+            context2.emplace(std::move(c).raw);
         }
     }
 
-    v.mkString(orig, context);
+    v.mkString(*s, context2);
 }
 
-static RegisterPrimOp primop_appendContext({
-    .name = "__appendContext",
-    .arity = 2,
-    .fun = prim_appendContext
+PrimOp primop_unsafeDiscardOutputDependency(
+    {.name = "__unsafeDiscardOutputDependency",
+     .args = {"s"},
+     .doc = R"(
+      Create a copy of the given string where every "derivation deep" string context element is turned into a constant string context element.
+
+      This is the opposite of [`builtins.addDrvOutputDependencies`](#builtins-addDrvOutputDependencies).
+
+      This is unsafe because it allows us to "forget" store objects we would have otherwise refered to with the string context,
+      whereas Nix normally tracks all dependencies consistently.
+      Safe operations "grow" but never "shrink" string contexts.
+      [`builtins.addDrvOutputDependencies`] in contrast is safe because "derivation deep" string context element always refers to the underlying derivation (among many more things).
+      Replacing a constant string context element with a "derivation deep" element is a safe operation that just enlargens the string context without forgetting anything.
+
+      [`builtins.addDrvOutputDependencies`]: #builtins-addDrvOutputDependencies
+    )",
+     .fun = prim_unsafeDiscardOutputDependency}
+);
+
+/**
+ * builtins. unsafeDiscardStringContext
+ */
+
+static void
+prim_unsafeDiscardStringContext(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto s = state.coerceToString(
+        pos,
+        *args[0],
+        context,
+        "while evaluating the argument passed to builtins.unsafeDiscardStringContext"
+    );
+    v.mkString(*s);
+}
+
+PrimOp primop_unsafeDiscardStringContext({
+    .name = "__unsafeDiscardStringContext",
+    .arity = 1,
+    .fun = prim_unsafeDiscardStringContext,
 });
 
 }

src/libexpr/primops/control.cc

@@ -0,0 +1,165 @@
+#include <list>             // for list
+#include <string>           // for basic_string
+#include "attr-set.hh"      // for BindingsBuilder
+#include "error.hh"         // for Trace, Error, lvlInfo, ErrorInfo
+#include "eval-error.hh"    // for Abort, ThrownError, EvalErrorBuilder
+#include "eval-settings.hh"
+#include "eval.hh"          // for EvalState, PrimOp, DebugTrace, ValMap
+#include "fmt.hh"           // for HintFmt
+#include "pos-idx.hh"       // for PosIdx
+#include "pos-table.hh"     // for PosTable
+#include "ref.hh"           // for ref
+#include "symbol-table.hh"  // for Symbol
+#include "types.hh"         // for BackedStringView
+#include "util.hh"          // for MaintainCount
+#include "value.hh"         // for Value
+#include "value/context.hh" // for NixStringContext
+
+namespace nix {
+
+enum class ReplExitStatus;
+
+/**
+ * builtins.abort
+ */
+
+static void prim_abort(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto s = state
+                 .coerceToString(
+                     pos,
+                     *args[0],
+                     context,
+                     "while evaluating the error message passed to builtins.abort"
+                 )
+                 .toOwned();
+    state.error<Abort>("evaluation aborted with the following error message: '%1%'", s)
+        .debugThrow();
+}
+
+PrimOp primop_abort({
+    .name = "abort",
+    .args = {"s"},
+    .doc = R"(
+      Abort Nix expression evaluation and print the error message *s*.
+    )",
+    .fun = prim_abort,
+});
+
+/**
+ * builtins.break
+ */
+
+static void prim_break(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    if (state.debugRepl && !state.debugTraces.empty()) {
+        auto error = Error(ErrorInfo{
+            .level = lvlInfo,
+            .msg = HintFmt("breakpoint reached"),
+            .pos = state.positions[pos],
+        });
+
+        auto & dt = state.debugTraces.front();
+        state.runDebugRepl(&error, dt.env, dt.expr);
+    }
+
+    // Return the value we were passed.
+    v = *args[0];
+}
+
+PrimOp primop_break({
+    .name = "break",
+    .args = {"v"},
+    .doc = R"(
+      In debug mode (enabled using `--debugger`), pause Nix expression evaluation and enter the REPL.
+      Otherwise, return the argument `v`.
+    )",
+    .fun = prim_break,
+});
+
+/**
+ * builtins.throw
+ */
+
+static void prim_throw(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto s =
+        state
+            .coerceToString(
+                pos, *args[0], context, "while evaluating the error message passed to builtin.throw"
+            )
+            .toOwned();
+    state.error<ThrownError>(s).debugThrow();
+}
+
+PrimOp primop_throw({
+    .name = "throw",
+    .args = {"s"},
+    .doc = R"(
+      Throw an error message *s*. This usually aborts Nix expression
+      evaluation, but in `nix-env -qa` and other commands that try to
+      evaluate a set of derivations to get information about those
+      derivations, a derivation that throws an error is silently skipped
+      (which is not the case for `abort`).
+    )",
+    .fun = prim_throw,
+});
+
+/**
+ * builtins.tryEval
+ */
+
+static void prim_tryEval(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto attrs = state.buildBindings(2);
+
+    /* increment state.trylevel, and decrement it when this function returns. */
+    MaintainCount trylevel(state.trylevel);
+
+    ReplExitStatus (*savedDebugRepl)(ref<EvalState> es, const ValMap & extraEnv) = nullptr;
+    if (state.debugRepl && evalSettings.ignoreExceptionsDuringTry) {
+        /* to prevent starting the repl from exceptions withing a tryEval, null it. */
+        savedDebugRepl = state.debugRepl;
+        state.debugRepl = nullptr;
+    }
+
+    try {
+        state.forceValue(*args[0], pos);
+        attrs.insert(state.sValue, args[0]);
+        attrs.alloc("success").mkBool(true);
+    } catch (AssertionError & e) {
+        attrs.alloc(state.sValue).mkBool(false);
+        attrs.alloc("success").mkBool(false);
+    }
+
+    // restore the debugRepl pointer if we saved it earlier.
+    if (savedDebugRepl) {
+        state.debugRepl = savedDebugRepl;
+    }
+
+    v.mkAttrs(attrs);
+}
+
+PrimOp primop_tryEval({
+    .name = "__tryEval",
+    .args = {"e"},
+    .doc = R"(
+      Try to shallowly evaluate *e*. Return a set containing the
+      attributes `success` (`true` if *e* evaluated successfully,
+      `false` if an error was thrown) and `value`, equalling *e* if
+      successful and `false` otherwise. `tryEval` will only prevent
+      errors created by `throw` or `assert` from being thrown.
+      Errors `tryEval` will not catch are for example those created
+      by `abort` and type errors generated by builtins. Also note that
+      this doesn't evaluate *e* deeply, so `let e = { x = throw ""; };
+      in (builtins.tryEval e).success` will be `true`. Using
+      `builtins.deepSeq` one can get the expected result:
+      `let e = { x = throw ""; }; in
+      (builtins.tryEval (builtins.deepSeq e e)).success` will be
+      `false`.
+    )",
+    .fun = prim_tryEval,
+});
+}

src/libexpr/primops/debug.cc

@@ -0,0 +1,181 @@
+#include <list>             // for list
+#include <string>           // for basic_string
+#include "attr-set.hh"      // for Bindings
+#include "eval-settings.hh" // for EvalSettings, evalSettings
+#include "eval.hh"          // for EvalState, PrimOp, DebugTrace
+#include "logging.hh"       // for Logger, printError
+#include "pos-idx.hh"       // for PosIdx
+#include "pos-table.hh"     // for PosTable
+#include "position.hh"      // for Pos
+#include "primops.hh"       // for makePositionThunks, prim_second, prim_trace
+#include "print.hh"         // for ValuePrinter
+#include "symbol-table.hh"  // for SymbolTable
+#include "value.hh"         // for Value, nString
+
+namespace nix {
+
+// access to exact position information (ie, line and colum numbers) is deferred
+// due to the cost associated with calculating that information and how rarely
+// it is used in practice. this is achieved by creating thunks to otherwise
+// inaccessible primops that are not exposed as __op or under builtins to turn
+// the internal PosIdx back into a line and column number, respectively. exposing
+// these primops in any way would at best be not useful and at worst create wildly
+// indeterministic eval results depending on parse order of files.
+//
+// in a simpler world this would instead be implemented as another kind of thunk,
+// but each type of thunk has an associated runtime cost in the current evaluator.
+// as with black holes this cost is too high to justify another thunk type to check
+// for in the very hot path that is forceValue.
+static struct LazyPosAcessors
+{
+    PrimOp primop_lineOfPos{
+        .arity = 1,
+        .fun = [](EvalState & state, PosIdx pos, Value ** args, Value & v
+               ) { v.mkInt(state.positions[PosIdx(args[0]->integer)].line); }
+    };
+    PrimOp primop_columnOfPos{
+        .arity = 1,
+        .fun = [](EvalState & state, PosIdx pos, Value ** args, Value & v
+               ) { v.mkInt(state.positions[PosIdx(args[0]->integer)].column); }
+    };
+
+    Value lineOfPos, columnOfPos;
+
+    LazyPosAcessors()
+    {
+        lineOfPos.mkPrimOp(&primop_lineOfPos);
+        columnOfPos.mkPrimOp(&primop_columnOfPos);
+    }
+
+    void operator()(EvalState & state, const PosIdx pos, Value & line, Value & column)
+    {
+        Value * posV = state.allocValue();
+        posV->mkInt(pos.id);
+        line.mkApp(&lineOfPos, posV);
+        column.mkApp(&columnOfPos, posV);
+    }
+} makeLazyPosAccessors;
+
+void makePositionThunks(EvalState & state, const PosIdx pos, Value & line, Value & column)
+{
+    makeLazyPosAccessors(state, pos, line, column);
+}
+
+/**
+ * Takes two arguments and evaluates to the second one. Used as the
+ * builtins.traceVerbose implementation when --trace-verbose is not enabled
+ */
+
+void prim_second(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[1], pos);
+    v = *args[1];
+}
+
+/**
+ * builtins.deepSeq
+ */
+
+static void prim_deepSeq(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValueDeep(*args[0]);
+    state.forceValue(*args[1], pos);
+    v = *args[1];
+}
+
+PrimOp primop_deepSeq({
+    .name = "__deepSeq",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      This is like `seq e1 e2`, except that *e1* is evaluated *deeply*:
+      if it’s a list or set, its elements or attributes are also
+      evaluated recursively.
+    )",
+    .fun = prim_deepSeq,
+});
+
+/**
+ * builtins.seq
+ */
+
+static void prim_seq(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    state.forceValue(*args[1], pos);
+    v = *args[1];
+}
+
+PrimOp primop_seq({
+    .name = "__seq",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Evaluate *e1*, then evaluate and return *e2*. This ensures that a
+      computation is strict in the value of *e1*.
+    )",
+    .fun = prim_seq,
+});
+
+/**
+ * builtins.trace
+ */
+
+void prim_trace(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    if (args[0]->type() == nString) {
+        printError("trace: %1%", args[0]->string.s);
+    } else {
+        printError("trace: %1%", ValuePrinter(state, *args[0]));
+    }
+    if (evalSettings.builtinsTraceDebugger && state.debugRepl && !state.debugTraces.empty()) {
+        const DebugTrace & last = state.debugTraces.front();
+        state.runDebugRepl(nullptr, last.env, last.expr);
+    }
+    state.forceValue(*args[1], pos);
+    v = *args[1];
+}
+
+PrimOp primop_trace({
+    .name = "__trace",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      Evaluate *e1* and print its abstract syntax representation on
+      standard error. Then return *e2*. This function is useful for
+      debugging.
+
+      If the
+      [`debugger-on-trace`](@docroot@/command-ref/conf-file.md#conf-debugger-on-trace)
+      option is set to `true` and the `--debugger` flag is given, the
+      interactive debugger will be started when `trace` is called (like
+      [`break`](@docroot@/language/builtins.md#builtins-break)).
+    )",
+    .fun = prim_trace,
+});
+
+/**
+ * builtins.unsafeGetAttrPos
+ */
+
+static void prim_unsafeGetAttrPos(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto attr = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.unsafeGetAttrPos"
+    );
+    state.forceAttrs(
+        *args[1], pos, "while evaluating the second argument passed to builtins.unsafeGetAttrPos"
+    );
+    Bindings::iterator i = args[1]->attrs->find(state.symbols.create(attr));
+    if (i == args[1]->attrs->end()) {
+        v.mkNull();
+    } else {
+        state.mkPos(v, i->pos);
+    }
+}
+
+PrimOp primop_unsafeGetAttrPos(PrimOp{
+    .name = "__unsafeGetAttrPos",
+    .arity = 2,
+    .fun = prim_unsafeGetAttrPos,
+});
+
+}

src/libexpr/primops/derivation.cc

@@ -0,0 +1,429 @@
+#include <map>                                     // for _Rb_tree_const_ite...
+#include <memory>                                  // for allocator, shared_ptr
+#include <nlohmann/json.hpp>                       // for basic_json
+#include <nlohmann/json_fwd.hpp>                   // for json
+#include <optional>                                // for optional, operator==
+#include <set>                                     // for set
+#include <string>                                  // for char_traits, basic...
+#include <string_view>                             // for operator==, basic_...
+#include <utility>                                 // for move, pair
+#include <variant>                                 // for visit
+#include "attr-set.hh"                             // for Attr, Bindings
+#include "config.hh"                               // for ExperimentalFeatur...
+#include "content-address.hh"                      // for ContentAddressMethod
+#include "derivations.hh"                          // for Derivation, Deriva...
+#include "derived-path-map.hh"                     // for DerivedPathMap
+#include "error.hh"                                // for Error, lvlChatty
+#include "eval-error.hh"                           // for EvalError, EvalErr...
+#include "eval.hh"                                 // for EvalState, PrimOp
+#include "experimental-features.hh"                // for ExperimentalFeature
+#include "fmt.hh"                                  // for HintFmt
+#include "hash.hh"                                 // for parseHashTypeOpt
+#include "logging.hh"                              // for Logger, printMsg
+#include "path.hh"                                 // for StorePath, StorePa...
+#include "pos-idx.hh"                              // for PosIdx, noPos
+#include "pos-table.hh"                            // for PosTable
+#include "position.hh"                             // for Pos, operator<<
+#include "primops.hh"                              // for getAttr, mkOutputS...
+#include "ref.hh"                                  // for ref
+#include "store-api.hh"                            // for Store
+#include "symbol-table.hh"                         // for Symbol, SymbolStr
+#include "sync.hh"                                 // for Sync
+#include "types.hh"                                // for Strings, BackedStr...
+#include "util.hh"                                 // for get, tokenizeString
+#include "value-to-json.hh"                        // for printValueAsJSON
+#include "value.hh"                                // for Value, nNull
+#include "value/context.hh"                        // for NixStringContextElem
+
+namespace nix {
+
+static void derivationStrictInternal(EvalState & state, const std::string &
+drvName, Bindings * attrs, Value & v)
+{
+    /* Check whether attributes should be passed as a JSON file. */
+    using nlohmann::json;
+    std::optional<json> jsonObject;
+    auto pos = v.determinePos(noPos);
+    auto attr = attrs->find(state.sStructuredAttrs);
+    if (attr != attrs->end() &&
+        state.forceBool(*attr->value, pos,
+                        "while evaluating the `__structuredAttrs` "
+                        "attribute passed to builtins.derivationStrict"))
+        jsonObject = json::object();
+
+    /* Check whether null attributes should be ignored. */
+    bool ignoreNulls = false;
+    attr = attrs->find(state.sIgnoreNulls);
+    if (attr != attrs->end())
+        ignoreNulls = state.forceBool(*attr->value, pos, "while evaluating the `__ignoreNulls` attribute " "passed to builtins.derivationStrict");
+
+    /* Build the derivation expression by processing the attributes. */
+    Derivation drv;
+    drv.name = drvName;
+
+    NixStringContext context;
+
+    bool contentAddressed = false;
+    bool isImpure = false;
+    std::optional<std::string> outputHash;
+    std::string outputHashAlgo;
+    std::optional<ContentAddressMethod> ingestionMethod;
+
+    StringSet outputs;
+    outputs.insert("out");
+
+    for (auto & i : attrs->lexicographicOrder(state.symbols)) {
+        if (i->name == state.sIgnoreNulls) continue;
+        const std::string & key = state.symbols[i->name];
+        vomit("processing attribute '%1%'", key);
+
+        auto handleHashMode = [&](const std::string_view s) {
+            if (s == "recursive") ingestionMethod = FileIngestionMethod::Recursive;
+            else if (s == "flat") ingestionMethod = FileIngestionMethod::Flat;
+            else if (s == "text") {
+                experimentalFeatureSettings.require(Xp::DynamicDerivations);
+                ingestionMethod = TextIngestionMethod {};
+            } else
+                state.error<EvalError>(
+                    "invalid value '%s' for 'outputHashMode' attribute", s
+                ).atPos(v).debugThrow();
+        };
+
+        auto handleOutputs = [&](const Strings & ss) {
+            outputs.clear();
+            for (auto & j : ss) {
+                if (outputs.find(j) != outputs.end())
+                    state.error<EvalError>("duplicate derivation output '%1%'", j)
+                        .atPos(v)
+                        .debugThrow();
+                /* !!! Check whether j is a valid attribute
+                   name. */
+                /* Derivations cannot be named ‘drv’, because
+                   then we'd have an attribute ‘drvPath’ in
+                   the resulting set. */
+                if (j == "drv")
+                    state.error<EvalError>("invalid derivation output name 'drv'")
+                        .atPos(v)
+                        .debugThrow();
+                outputs.insert(j);
+            }
+            if (outputs.empty())
+                state.error<EvalError>("derivation cannot have an empty set of outputs")
+                    .atPos(v)
+                    .debugThrow();
+        };
+
+        try {
+            // This try-catch block adds context for most errors.
+            // Use this empty error context to signify that we defer to it.
+            const std::string_view context_below("");
+
+            if (ignoreNulls) {
+                state.forceValue(*i->value, pos);
+                if (i->value->type() == nNull) continue;
+            }
+
+            if (i->name == state.sContentAddressed && state.forceBool(*i->value, pos, context_below)) {
+                contentAddressed = true;
+                experimentalFeatureSettings.require(Xp::CaDerivations);
+            }
+
+            else if (i->name == state.sImpure && state.forceBool(*i->value, pos, context_below)) {
+                isImpure = true;
+                experimentalFeatureSettings.require(Xp::ImpureDerivations);
+            }
+
+            /* The `args' attribute is special: it supplies the
+               command-line arguments to the builder. */
+            else if (i->name == state.sArgs) {
+                state.forceList(*i->value, pos, context_below);
+                for (auto elem : i->value->listItems()) {
+                    auto s = state.coerceToString(pos, *elem, context,
+                            "while evaluating an element of the argument list",
+                            true).toOwned();
+                    drv.args.push_back(s);
+                }
+            }
+
+            /* All other attributes are passed to the builder through
+               the environment. */
+            else {
+
+                if (jsonObject) {
+
+                    if (i->name == state.sStructuredAttrs) continue;
+
+                    (*jsonObject)[key] = printValueAsJSON(state, true, *i->value, pos, context);
+
+                    if (i->name == state.sBuilder)
+                        drv.builder = state.forceString(*i->value, context, pos, context_below);
+                    else if (i->name == state.sSystem)
+                        drv.platform = state.forceStringNoCtx(*i->value, pos, context_below);
+                    else if (i->name == state.sOutputHash)
+                        outputHash = state.forceStringNoCtx(*i->value, pos, context_below);
+                    else if (i->name == state.sOutputHashAlgo)
+                        outputHashAlgo = state.forceStringNoCtx(*i->value, pos, context_below);
+                    else if (i->name == state.sOutputHashMode)
+                        handleHashMode(state.forceStringNoCtx(*i->value, pos, context_below));
+                    else if (i->name == state.sOutputs) {
+                        /* Require ‘outputs’ to be a list of strings. */
+                        state.forceList(*i->value, pos, context_below);
+                        Strings ss;
+                        for (auto elem : i->value->listItems())
+                            ss.emplace_back(state.forceStringNoCtx(*elem, pos, context_below));
+                        handleOutputs(ss);
+                    }
+
+                } else {
+                    auto s = state.coerceToString(pos, *i->value, context, context_below, true).toOwned();
+                    drv.env.emplace(key, s);
+                    if (i->name == state.sBuilder) drv.builder = std::move(s);
+                    else if (i->name == state.sSystem) drv.platform = std::move(s);
+                    else if (i->name == state.sOutputHash) outputHash = std::move(s);
+                    else if (i->name == state.sOutputHashAlgo) outputHashAlgo = std::move(s);
+                    else if (i->name == state.sOutputHashMode) handleHashMode(s);
+                    else if (i->name == state.sOutputs)
+                        handleOutputs(tokenizeString<Strings>(s));
+                }
+
+            }
+
+        } catch (Error & e) {
+            e.addTrace(state.positions[i->pos],
+                HintFmt("while evaluating attribute '%1%' of derivation '%2%'", key, drvName));
+            throw;
+        }
+    }
+
+    if (jsonObject) {
+        drv.env.emplace("__json", jsonObject->dump());
+        jsonObject.reset();
+    }
+
+    /* Everything in the context of the strings in the derivation
+       attributes should be added as dependencies of the resulting
+       derivation. */
+    for (auto & c : context) {
+        std::visit(overloaded {
+            /* Since this allows the builder to gain access to every
+               path in the dependency graph of the derivation (including
+               all outputs), all paths in the graph must be added to
+               this derivation's list of inputs to ensure that they are
+               available when the builder runs. */
+            [&](const NixStringContextElem::DrvDeep & d) {
+                /* !!! This doesn't work if readOnlyMode is set. */
+                StorePathSet refs;
+                state.store->computeFSClosure(d.drvPath, refs);
+                for (auto & j : refs) {
+                    drv.inputSrcs.insert(j);
+                    if (j.isDerivation()) {
+                        drv.inputDrvs.map[j].value = state.store->readDerivation(j).outputNames();
+                    }
+                }
+            },
+            [&](const NixStringContextElem::Built & b) {
+                drv.inputDrvs.ensureSlot(*b.drvPath).value.insert(b.output);
+            },
+            [&](const NixStringContextElem::Opaque & o) {
+                drv.inputSrcs.insert(o.path);
+            },
+        }, c.raw);
+    }
+
+    /* Do we have all required attributes? */
+    if (drv.builder == "")
+        state.error<EvalError>("required attribute 'builder' missing")
+            .atPos(v)
+            .debugThrow();
+
+    if (drv.platform == "")
+        state.error<EvalError>("required attribute 'system' missing")
+            .atPos(v)
+            .debugThrow();
+
+    /* Check whether the derivation name is valid. */
+    if (isDerivation(drvName) &&
+        !(ingestionMethod == ContentAddressMethod { TextIngestionMethod { } } &&
+          outputs.size() == 1 &&
+          *(outputs.begin()) == "out"))
+    {
+        state.error<EvalError>(
+            "derivation names are allowed to end in '%s' only if they produce a single derivation file",
+            drvExtension
+        ).atPos(v).debugThrow();
+    }
+
+    if (outputHash) {
+        /* Handle fixed-output derivations.
+
+           Ignore `__contentAddressed` because fixed output derivations are
+           already content addressed. */
+        if (outputs.size() != 1 || *(outputs.begin()) != "out")
+            state.error<EvalError>(
+                "multiple outputs are not supported in fixed-output derivations"
+            ).atPos(v).debugThrow();
+
+        auto h = newHashAllowEmpty(*outputHash, parseHashTypeOpt(outputHashAlgo));
+
+        auto method = ingestionMethod.value_or(FileIngestionMethod::Flat);
+
+        DerivationOutput::CAFixed dof {
+            .ca = ContentAddress {
+                .method = std::move(method),
+                .hash = std::move(h),
+            },
+        };
+
+        drv.env["out"] = state.store->printStorePath(dof.path(*state.store, drvName, "out"));
+        drv.outputs.insert_or_assign("out", std::move(dof));
+    }
+
+    else if (contentAddressed || isImpure) {
+        if (contentAddressed && isImpure)
+            state.error<EvalError>("derivation cannot be both content-addressed and impure")
+                .atPos(v).debugThrow();
+
+        auto ht = parseHashTypeOpt(outputHashAlgo).value_or(htSHA256);
+        auto method = ingestionMethod.value_or(FileIngestionMethod::Recursive);
+
+        for (auto & i : outputs) {
+            drv.env[i] = hashPlaceholder(i);
+            if (isImpure)
+                drv.outputs.insert_or_assign(i,
+                    DerivationOutput::Impure {
+                        .method = method,
+                        .hashType = ht,
+                    });
+            else
+                drv.outputs.insert_or_assign(i,
+                    DerivationOutput::CAFloating {
+                        .method = method,
+                        .hashType = ht,
+                    });
+        }
+    }
+
+    else {
+        /* Compute a hash over the "masked" store derivation, which is
+           the final one except that in the list of outputs, the
+           output paths are empty strings, and the corresponding
+           environment variables have an empty value.  This ensures
+           that changes in the set of output names do get reflected in
+           the hash. */
+        for (auto & i : outputs) {
+            drv.env[i] = "";
+            drv.outputs.insert_or_assign(i,
+                DerivationOutput::Deferred { });
+        }
+
+        auto hashModulo = hashDerivationModulo(*state.store, Derivation(drv), true);
+        switch (hashModulo.kind) {
+        case DrvHash::Kind::Regular:
+            for (auto & i : outputs) {
+                auto h = get(hashModulo.hashes, i);
+                if (!h)
+                    state.error<AssertionError>(
+                        "derivation produced no hash for output '%s'",
+                        i
+                    ).atPos(v).debugThrow();
+                auto outPath = state.store->makeOutputPath(i, *h, drvName);
+                drv.env[i] = state.store->printStorePath(outPath);
+                drv.outputs.insert_or_assign(
+                    i,
+                    DerivationOutput::InputAddressed {
+                        .path = std::move(outPath),
+                    });
+            }
+            break;
+            ;
+        case DrvHash::Kind::Deferred:
+            for (auto & i : outputs) {
+                drv.outputs.insert_or_assign(i, DerivationOutput::Deferred {});
+            }
+        }
+    }
+
+    /* Write the resulting term into the Nix store directory. */
+    auto drvPath = writeDerivation(*state.store, drv, state.repair);
+    auto drvPathS = state.store->printStorePath(drvPath);
+
+    printMsg(lvlChatty, "instantiated '%1%' -> '%2%'", drvName, drvPathS);
+
+    /* Optimisation, but required in read-only mode! because in that
+       case we don't actually write store derivations, so we can't
+       read them later. */
+    {
+        auto h = hashDerivationModulo(*state.store, drv, false);
+        drvHashes.lock()->insert_or_assign(drvPath, h);
+    }
+
+    auto result = state.buildBindings(1 + drv.outputs.size());
+    result.alloc(state.sDrvPath).mkString(drvPathS, {
+        NixStringContextElem::DrvDeep { .drvPath = drvPath },
+    });
+    for (auto & i : drv.outputs)
+        mkOutputString(state, result, drvPath, i);
+
+    v.mkAttrs(result);
+}
+
+/* Construct (as a unobservable side effect) a Nix derivation
+   expression that performs the derivation described by the argument
+   set.  Returns the original set extended with the following
+   attributes: `outPath' containing the primary output path of the
+   derivation; `drvPath' containing the path of the Nix expression;
+   and `type' set to `derivation' to indicate that this is a
+   derivation. */
+static void prim_derivationStrict(EvalState & state, const PosIdx pos, Value * * args, Value & v)
+{
+    state.forceAttrs(*args[0], pos, "while evaluating the argument passed to builtins.derivationStrict");
+
+    Bindings * attrs = args[0]->attrs;
+
+    /* Figure out the name first (for stack backtraces). */
+    Bindings::iterator nameAttr = getAttr(state, state.sName, attrs, "in the attrset passed as argument to builtins.derivationStrict");
+
+    std::string drvName;
+    try {
+        drvName = state.forceStringNoCtx(*nameAttr->value, pos, "while evaluating the `name` attribute passed to builtins.derivationStrict");
+    } catch (Error & e) {
+        e.addTrace(state.positions[nameAttr->pos], "while evaluating the derivation attribute 'name'");
+        throw;
+    }
+
+    try {
+        derivationStrictInternal(state, drvName, attrs, v);
+    } catch (Error & e) {
+        Pos pos = state.positions[nameAttr->pos];
+        /*
+         * Here we make two abuses of the error system
+         *
+         * 1. We print the location as a string to avoid a code snippet being
+         * printed. While the location of the name attribute is a good hint, the
+         * exact code there is irrelevant.
+         *
+         * 2. We mark this trace as a frame trace, meaning that we stop printing
+         * less important traces from now on. In particular, this prevents the
+         * display of the automatic "while calling builtins.derivationStrict"
+         * trace, which is of little use for the public we target here.
+         *
+         * Please keep in mind that error reporting is done on a best-effort
+         * basis in nix. There is no accurate location for a derivation, as it
+         * often results from the composition of several functions
+         * (derivationStrict, derivation, mkDerivation, mkPythonModule, etc.)
+         */
+        e.addTrace(nullptr, HintFmt(
+                "while evaluating derivation '%s'\n"
+                "  whose name attribute is located at %s",
+                drvName, pos));
+        throw;
+    }
+}
+
+PrimOp primop_derivationStrict(PrimOp {
+    .name = "derivationStrict",
+    .arity = 1,
+    .fun = prim_derivationStrict,
+});
+
+}

src/libexpr/primops/fetchClosure.cc

@@ -1,7 +1,20 @@
-#include "primops.hh"
-#include "store-api.hh"
-#include "make-content-addressed.hh"
-#include "url.hh"
+#include <optional>                                // for optional
+#include <string>                                  // for allocator, char_tr...
+#include "attr-set.hh"                             // for Attr, Bindings
+#include "error.hh"                                // for Error, Trace
+#include "eval.hh"                                 // for EvalState, PrimOp
+#include "experimental-features.hh"                // for ExperimentalFeature
+#include "fmt.hh"                                  // for HintFmt
+#include "make-content-addressed.hh"               // for makeContentAddressed
+#include "path-info.hh"                            // for ValidPathInfo
+#include "path.hh"                                 // for StorePath
+#include "pos-idx.hh"                              // for PosIdx
+#include "realisation.hh"                          // for RealisedPath
+#include "store-api.hh"                            // for Store, copyClosure
+#include "url.hh"                                  // for ParsedURL, parseURL
+#include "util.hh"                                 // for getEnv, operator+
+#include "value.hh"                                // for Value, nString
+#include "value/context.hh"                        // for NixStringContext
 
 namespace nix {
 
@@ -207,7 +220,7 @@ static void prim_fetchClosure(EvalState & state, const PosIdx pos, Value * * arg
         runFetchClosureWithContentAddressedPath(state, pos, *fromStore, *fromPath, v);
 }
 
-static RegisterPrimOp primop_fetchClosure({
+PrimOp primop_fetchClosure({
     .name = "__fetchClosure",
     .args = {"args"},
     .doc = R"(

src/libexpr/primops/fetchMercurial.cc

@@ -1,10 +1,23 @@
-#include "primops.hh"
-#include "eval-inline.hh"
-#include "eval-settings.hh"
-#include "store-api.hh"
-#include "fetchers.hh"
-#include "url.hh"
-#include "url-parts.hh"
+#include <optional>          // for optional
+#include <regex>             // for regex_match
+#include <string>            // for allocator, char_traits, basic_string
+#include <string_view>       // for operator==, basic_string_view, string_view
+#include <utility>           // for move
+#include "attr-set.hh"       // for Attr, BindingsBuilder, Bindings
+#include "attrs.hh"          // for Attrs
+#include "error.hh"          // for Error
+#include "eval-error.hh"     // for EvalError, EvalErrorBuilder
+#include "eval-settings.hh"  // for EvalSettings, evalSettings
+#include "eval.hh"           // for EvalState, PrimOp
+#include "fetchers.hh"       // for Input, Tree
+#include "hash.hh"           // for Hash, HashType
+#include "pos-idx.hh"        // for PosIdx
+#include "ref.hh"            // for ref
+#include "symbol-table.hh"   // for Symbol, SymbolStr, SymbolTable
+#include "types.hh"          // for BackedStringView
+#include "url-parts.hh"      // for revRegex
+#include "value.hh"          // for Value, nAttrs
+#include "value/context.hh"  // for NixStringContext
 
 namespace nix {
 
@@ -83,7 +96,7 @@ static void prim_fetchMercurial(EvalState & state, const PosIdx pos, Value * * a
     state.allowPath(tree.storePath);
 }
 
-static RegisterPrimOp r_fetchMercurial({
+PrimOp primop_fetchMercurial({
     .name = "fetchMercurial",
     .arity = 1,
     .fun = prim_fetchMercurial

src/libexpr/primops/fetchTree.cc

@@ -1,15 +1,36 @@
-#include "primops.hh"
-#include "eval-inline.hh"
-#include "eval-settings.hh"
-#include "store-api.hh"
-#include "fetchers.hh"
-#include "filetransfer.hh"
-#include "registry.hh"
-#include "url.hh"
-
-#include <ctime>
-#include <iomanip>
-#include <regex>
+#include <assert.h>                                // for assert
+#include <stdint.h>                                // for uint64_t
+#include <ctime>                                   // for gmtime
+#include <iomanip>                                 // for put_time, operator<<
+#include <map>                                     // for operator==, map
+#include <optional>                                // for optional, operator==
+#include <regex>                                   // for regex_match, regex...
+#include <string>                                  // for char_traits, basic...
+#include <string_view>                             // for operator==, basic_...
+#include <utility>                                 // for move, pair
+#include "attr-set.hh"                             // for Attr, BindingsBuilder
+#include "attrs.hh"                                // for maybeGetStrAttr
+#include "config.hh"                               // for Setting, Experimen...
+#include "content-address.hh"                      // for FileIngestionMethod
+#include "eval-error.hh"                           // for EvalError, EvalErr...
+#include "eval-settings.hh"                        // for EvalSettings, eval...
+#include "eval.hh"                                 // for EvalState, PrimOp
+#include "experimental-features.hh"                // for ExperimentalFeature
+#include "fetchers.hh"                             // for Input, Tree, downl...
+#include "fmt.hh"                                  // for fmt
+#include "hash.hh"                                 // for Hash, HashType, Base
+#include "path-info.hh"                            // for ValidPathInfo
+#include "path.hh"                                 // for StorePath
+#include "pos-idx.hh"                              // for PosIdx
+#include "ref.hh"                                  // for ref
+#include "registry.hh"                             // for lookupInRegistries
+#include "store-api.hh"                            // for Store
+#include "symbol-table.hh"                         // for Symbol, SymbolTable
+#include "types.hh"                                // for BackedStringView
+#include "url.hh"                                  // for ParsedURL
+#include "util.hh"                                 // for baseNameOf
+#include "value.hh"                                // for Value, nAttrs, nBool
+#include "value/context.hh"                        // for NixStringContext
 
 namespace nix {
 
@@ -198,7 +219,7 @@ static void prim_fetchTree(EvalState & state, const PosIdx pos, Value * * args,
 }
 
 // FIXME: document
-static RegisterPrimOp primop_fetchTree({
+PrimOp primop_fetchTree({
     .name = "fetchTree",
     .arity = 1,
     .fun = prim_fetchTree
@@ -290,7 +311,7 @@ static void prim_fetchurl(EvalState & state, const PosIdx pos, Value * * args, V
     fetch(state, pos, args, v, "fetchurl", false, "");
 }
 
-static RegisterPrimOp primop_fetchurl({
+PrimOp primop_fetchurl({
     .name = "__fetchurl",
     .args = {"url"},
     .doc = R"(
@@ -306,7 +327,7 @@ static void prim_fetchTarball(EvalState & state, const PosIdx pos, Value * * arg
     fetch(state, pos, args, v, "fetchTarball", true, "source");
 }
 
-static RegisterPrimOp primop_fetchTarball({
+PrimOp primop_fetchTarball({
     .name = "fetchTarball",
     .args = {"args"},
     .doc = R"(
@@ -356,7 +377,7 @@ static void prim_fetchGit(EvalState & state, const PosIdx pos, Value * * args, V
     fetchTree(state, pos, args, v, "git", FetchTreeParams { .emptyRevFallback = true, .allowNameArgument = true });
 }
 
-static RegisterPrimOp primop_fetchGit({
+PrimOp primop_fetchGit({
     .name = "fetchGit",
     .args = {"args"},
     .doc = R"(

src/libexpr/primops/fromTOML.cc

@@ -1,8 +1,27 @@
-#include "primops.hh"
-#include "eval-inline.hh"
-
-#include <sstream>
-#include <toml.hpp>
+#include <stddef.h>                  // for size_t
+#include <stdint.h>                  // for int64_t
+#include <exception>                 // for exception
+#include <functional>                // for function
+#include <sstream>                   // for char_traits, basic_ostringstream
+#include <stdexcept>                 // for runtime_error
+#include <string>                    // for basic_string, operator<<, operat...
+#include <toml/comments.hpp>         // for operator<<
+#include <toml/datetime.hpp>         // for operator<<
+#include <toml/get.hpp>              // for get
+#include <toml/parser.hpp>           // for parse
+#include <toml/serializer.hpp>       // for operator<<
+#include <toml/types.hpp>            // for value_t
+#include <toml/value.hpp>            // for value, table
+#include <unordered_map>             // for unordered_map, _Node_iterator
+#include <utility>                   // for pair
+#include <vector>                    // for vector
+#include "attr-set.hh"               // for BindingsBuilder
+#include "config.hh"                 // for ExperimentalFeatureSettings, exp...
+#include "eval-error.hh"             // for EvalError, EvalErrorBuilder
+#include "eval.hh"                   // for EvalState, PrimOp
+#include "experimental-features.hh"  // for ExperimentalFeature, Xp
+#include "pos-idx.hh"                // for PosIdx
+#include "value.hh"                  // for Value, NixFloat
 
 namespace nix {
 
@@ -86,7 +105,7 @@ static void prim_fromTOML(EvalState & state, const PosIdx pos, Value * * args, V
     }
 }
 
-static RegisterPrimOp primop_fromTOML({
+PrimOp primop_fromTOML({
     .name = "fromTOML",
     .args = {"e"},
     .doc = R"(

src/libexpr/primops/hash.cc

@@ -0,0 +1,75 @@
+#include "hash.hh"
+#include "eval-error.hh"    // for EvalError, EvalErrorBuilder
+#include "eval.hh"          // for EvalState, PrimOp
+#include "pos-idx.hh"       // for PosIdx
+#include "primops.hh"       // for realisePath
+#include "source-path.hh"   // for SourcePath
+#include "value.hh"         // for Value
+#include "value/context.hh" // for NixStringContext
+
+namespace nix {
+
+/**
+ * builtins.hashFile
+ */
+
+static void prim_hashFile(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto type = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.hashFile"
+    );
+    std::optional<HashType> ht = parseHashType(type);
+    if (!ht) {
+        state.error<EvalError>("unknown hash type '%1%'", type).atPos(pos).debugThrow();
+    }
+
+    auto path = realisePath(state, pos, *args[1]);
+
+    v.mkString(hashString(*ht, path.readFile()).to_string(Base16, false));
+}
+
+PrimOp primop_hashFile({
+    .name = "__hashFile",
+    .args = {"type", "p"},
+    .doc = R"(
+      Return a base-16 representation of the cryptographic hash of the
+      file at path *p*. The hash algorithm specified by *type* must be one
+      of `"md5"`, `"sha1"`, `"sha256"` or `"sha512"`.
+    )",
+    .fun = prim_hashFile,
+});
+
+/**
+ * builtins.hashString
+ */
+
+static void prim_hashString(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto type = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.hashString"
+    );
+    std::optional<HashType> ht = parseHashType(type);
+    if (!ht) {
+        state.error<EvalError>("unknown hash algorithm '%1%'", type).atPos(pos).debugThrow();
+    }
+
+    NixStringContext context; // discarded
+    auto s = state.forceString(
+        *args[1], context, pos, "while evaluating the second argument passed to builtins.hashString"
+    );
+
+    v.mkString(hashString(*ht, s).to_string(Base16, false));
+}
+
+PrimOp primop_hashString({
+    .name = "__hashString",
+    .args = {"type", "s"},
+    .doc = R"(
+      Return a base-16 representation of the cryptographic hash of string
+      *s*. The hash algorithm specified by *type* must be one of `"md5"`,
+      `"sha1"`, `"sha256"` or `"sha512"`.
+    )",
+    .fun = prim_hashString,
+});
+
+}

src/libexpr/primops/import.cc

@@ -0,0 +1,290 @@
+#include <dlfcn.h>           // for dlerror, dlopen, dlsym, RTLD_LAZY, RTLD_...
+#include <memory>            // for allocator, __shared_ptr_access, shared_ptr
+#include <optional>          // for optional, nullopt, nullopt_t
+#include <string>            // for basic_string, char_traits, string, opera...
+#include <utility>           // for pair
+#include "attr-set.hh"       // for BindingsBuilder, Bindings, Attr
+#include "canon-path.hh"     // for CanonPath
+#include "derivations.hh"    // for Derivation, DerivationOutput, isDerivation
+#include "derived-path.hh"   // for makeConstantStorePathRef, SingleDerivedPath
+#include "eval-error.hh"     // for EvalError, EvalErrorBuilder
+#include "eval.hh"           // for EvalState, PrimOp, Env, resolveExprPath
+#include "logging.hh"        // for Logger, debug
+#include "nixexpr.hh"        // for StaticEnv, Expr
+#include "path.hh"           // for StorePath
+#include "pos-idx.hh"        // for PosIdx
+#include "primops.hh"        // for realisePath, prim_importNative, mkOutput...
+#include "ref.hh"            // for ref
+#include "source-path.hh"    // for SourcePath
+#include "store-api.hh"      // for Store
+#include "symbol-table.hh"   // for Symbol
+#include "util.hh"           // for enumerate
+#include "value.hh"          // for Value, allocRootValue
+#include "value/context.hh"  // for NixStringContextElem
+
+namespace nix {
+
+/* Want reasonable symbol names, so extern C */
+/* !!! Should we pass the Pos or the file name too? */
+extern "C" typedef void (*ValueInitializer)(EvalState & state, Value & v);
+
+/**
+ * Add and attribute to the given attribute map from the output name to
+ * the output path, or a placeholder.
+ *
+ * Where possible the path is used, but for floating CA derivations we
+ * may not know it. For sake of determinism we always assume we don't
+ * and instead put in a place holder. In either case, however, the
+ * string context will contain the drv path and output name, so
+ * downstream derivations will have the proper dependency, and in
+ * addition, before building, the placeholder will be rewritten to be
+ * the actual path.
+ *
+ * The 'drv' and 'drvPath' outputs must correspond.
+ */
+
+void mkOutputString(
+    EvalState & state,
+    BindingsBuilder & attrs,
+    const StorePath & drvPath,
+    const std::pair<std::string, DerivationOutput> & o
+)
+{
+    state.mkOutputString(
+        attrs.alloc(o.first),
+        SingleDerivedPath::Built{
+            .drvPath = makeConstantStorePathRef(drvPath),
+            .output = o.first,
+        },
+        o.second.path(*state.store, Derivation::nameFromPath(drvPath), o.first)
+    );
+}
+
+/* Load and evaluate an expression from path specified by the
+   argument. */
+static void import(EvalState & state, const PosIdx pos, Value & vPath, Value * vScope, Value & v)
+{
+    auto path = realisePath(state, pos, vPath);
+    auto path2 = path.path.abs();
+
+    // FIXME
+    auto isValidDerivationInStore = [&]() -> std::optional<StorePath> {
+        if (!state.store->isStorePath(path2)) {
+            return std::nullopt;
+        }
+        auto storePath = state.store->parseStorePath(path2);
+        if (!(state.store->isValidPath(storePath) && isDerivation(path2))) {
+            return std::nullopt;
+        }
+        return storePath;
+    };
+
+    if (auto storePath = isValidDerivationInStore()) {
+        Derivation drv = state.store->readDerivation(*storePath);
+        auto attrs = state.buildBindings(3 + drv.outputs.size());
+        attrs.alloc(state.sDrvPath)
+            .mkString(
+                path2,
+                {
+                    NixStringContextElem::DrvDeep{.drvPath = *storePath},
+                }
+            );
+        attrs.alloc(state.sName).mkString(drv.env["name"]);
+        auto & outputsVal = attrs.alloc(state.sOutputs);
+        state.mkList(outputsVal, drv.outputs.size());
+
+        for (const auto & [i, o] : enumerate(drv.outputs)) {
+            mkOutputString(state, attrs, *storePath, o);
+            (outputsVal.listElems()[i] = state.allocValue())->mkString(o.first);
+        }
+
+        auto w = state.allocValue();
+        w->mkAttrs(attrs);
+
+        if (!state.vImportedDrvToDerivation) {
+            state.vImportedDrvToDerivation = allocRootValue(state.allocValue());
+            state.eval(
+                state.parseExprFromString(
+#include "imported-drv-to-derivation.nix.gen.hh"
+                    , CanonPath::root
+                ),
+                **state.vImportedDrvToDerivation
+            );
+        }
+
+        state.forceFunction(
+            **state.vImportedDrvToDerivation,
+            pos,
+            "while evaluating imported-drv-to-derivation.nix.gen.hh"
+        );
+        v.mkApp(*state.vImportedDrvToDerivation, w);
+        state.forceAttrs(v, pos, "while calling imported-drv-to-derivation.nix.gen.hh");
+    }
+
+    else if (path2 == corepkgsPrefix + "fetchurl.nix")
+    {
+        state.eval(
+            state.parseExprFromString(
+#include "fetchurl.nix.gen.hh"
+                , CanonPath::root
+            ),
+            v
+        );
+    }
+
+    else
+    {
+        if (!vScope) {
+            state.evalFile(path, v);
+        } else {
+            state.forceAttrs(
+                *vScope, pos, "while evaluating the first argument passed to builtins.scopedImport"
+            );
+
+            Env * env = &state.allocEnv(vScope->attrs->size());
+            env->up = &state.baseEnv;
+
+            auto staticEnv = std::make_shared<StaticEnv>(
+                nullptr, state.staticBaseEnv.get(), vScope->attrs->size()
+            );
+
+            unsigned int displ = 0;
+            for (auto & attr : *vScope->attrs) {
+                staticEnv->vars.emplace_back(attr.name, displ);
+                env->values[displ++] = attr.value;
+            }
+
+            // No need to call staticEnv.sort(), because
+            // args[0]->attrs is already sorted.
+
+            debug("evaluating file '%1%'", path);
+            Expr * e = state.parseExprFromFile(resolveExprPath(path), staticEnv);
+
+            e->eval(state, *env, v);
+        }
+    }
+}
+
+static void prim_import(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    import(state, pos, *args[0], nullptr, v);
+}
+
+PrimOp primop_import({
+    .name = "import",
+    .args = {"path"},
+    // TODO turn "normal path values" into link below
+    .doc = R"(
+      Load, parse and return the Nix expression in the file *path*.
+
+      The value *path* can be a path, a string, or an attribute set with an
+      `__toString` attribute or a `outPath` attribute (as derivations or flake
+      inputs typically have).
+
+      If *path* is a directory, the file `default.nix` in that directory
+      is loaded.
+
+      Evaluation aborts if the file doesn’t exist or contains
+      an incorrect Nix expression. `import` implements Nix’s module
+      system: you can put any Nix expression (such as a set or a
+      function) in a separate file, and use it from Nix expressions in
+      other files.
+
+      > **Note**
+      >
+      > Unlike some languages, `import` is a regular function in Nix.
+      > Paths using the angle bracket syntax (e.g., `import` *\<foo\>*)
+      > are normal [path values](@docroot@/language/values.md#type-path).
+
+      A Nix expression loaded by `import` must not contain any *free
+      variables* (identifiers that are not defined in the Nix expression
+      itself and are not built-in). Therefore, it cannot refer to
+      variables that are in scope at the call site. For instance, if you
+      have a calling expression
+
+      ```nix
+      rec {
+        x = 123;
+        y = import ./foo.nix;
+      }
+      ```
+
+      then the following `foo.nix` will give an error:
+
+      ```nix
+      x + 456
+      ```
+
+      since `x` is not in scope in `foo.nix`. If you want `x` to be
+      available in `foo.nix`, you should pass it as a function argument:
+
+      ```nix
+      rec {
+        x = 123;
+        y = import ./foo.nix x;
+      }
+      ```
+
+      and
+
+      ```nix
+      x: x + 456
+      ```
+
+      (The function argument doesn’t have to be called `x` in `foo.nix`;
+      any name would work.)
+    )",
+    .fun = prim_import,
+});
+
+/* Load a ValueInitializer from a DSO and return whatever it initializes */
+void prim_importNative(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto path = realisePath(state, pos, *args[0]);
+
+    std::string sym(state.forceStringNoCtx(
+        *args[1], pos, "while evaluating the second argument passed to builtins.importNative"
+    ));
+
+    void * handle = dlopen(path.path.c_str(), RTLD_LAZY | RTLD_LOCAL);
+    if (!handle) {
+        state.error<EvalError>("could not open '%1%': %2%", path, dlerror()).debugThrow();
+    }
+
+    dlerror();
+    ValueInitializer func = (ValueInitializer) dlsym(handle, sym.c_str());
+    if (!func) {
+        char * message = dlerror();
+        if (message) {
+            state
+                .error<EvalError>("could not load symbol '%1%' from '%2%': %3%", sym, path, message)
+                .debugThrow();
+        } else {
+            state
+                .error<EvalError>(
+                    "symbol '%1%' from '%2%' resolved to NULL when a function pointer was expected",
+                    sym,
+                    path
+                )
+                .debugThrow();
+        }
+    }
+
+    (func)(state, v);
+
+    /* We don't dlclose because v may be a primop referencing a function in the shared object file
+     */
+}
+
+static void prim_scopedImport(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    import(state, pos, *args[1], args[0], v);
+}
+
+PrimOp primop_scopedImport(PrimOp{
+    .name = "scopedImport",
+    .arity = 2,
+    .fun = prim_scopedImport,
+});
+
+}

src/libexpr/primops/json.cc

@@ -0,0 +1,70 @@
+#include <sstream>          // for basic_ostringstream, basic_ios, basic_os...
+#include "eval.hh"          // for PrimOp, EvalState
+#include "json-to-value.hh" // for JSONParseError, parseJSON
+#include "pos-idx.hh"       // for PosIdx
+#include "pos-table.hh"     // for PosTable
+#include "value-to-json.hh" // for printValueAsJSON
+#include "value.hh"         // for Value
+#include "value/context.hh" // for NixStringContext
+
+namespace nix {
+
+/**
+ * builtins.fromJSON
+ */
+
+static void prim_fromJSON(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto s = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.fromJSON"
+    );
+    try {
+        parseJSON(state, s, v);
+    } catch (JSONParseError & e) {
+        e.addTrace(state.positions[pos], "while decoding a JSON string");
+        throw;
+    }
+}
+
+PrimOp primop_fromJSON({
+    .name = "__fromJSON",
+    .args = {"e"},
+    .doc = R"(
+      Convert a JSON string to a Nix value. For example,
+
+      ```nix
+      builtins.fromJSON ''{"x": [1, 2, 3], "y": null}''
+      ```
+
+      returns the value `{ x = [ 1 2 3 ]; y = null; }`.
+    )",
+    .fun = prim_fromJSON,
+});
+
+/**
+ * builtins.toJSON
+ */
+
+static void prim_toJSON(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    std::ostringstream out;
+    NixStringContext context;
+    printValueAsJSON(state, true, *args[0], pos, out, context);
+    v.mkString(out.str(), context);
+}
+
+PrimOp primop_toJSON({
+    .name = "__toJSON",
+    .args = {"e"},
+    .doc = R"(
+      Return a string containing a JSON representation of *e*. Strings,
+      integers, floats, booleans, nulls and lists are mapped to their JSON
+      equivalents. Sets (except derivations) are represented as objects.
+      Derivations are translated to a JSON string containing the
+      derivation’s output path. Paths are copied to the store and
+      represented as a JSON string of the resulting store path.
+    )",
+    .fun = prim_toJSON,
+});
+
+}

src/libexpr/primops/list.cc

@@ -0,0 +1,694 @@
+#include <string.h>                   // for memcpy, size_t
+#include <algorithm>                  // for stable_sort
+#include <boost/container/vector.hpp> // for vector
+#include <functional>                 // for function
+#include <set>                        // for set
+#include <string>                     // for basic_string, allocator, opera...
+#include <utility>                    // for pair
+#include "attr-set.hh"                // for BindingsBuilder, Bindings
+#include "eval-error.hh"              // for EvalError, EvalErrorBuilder
+#include "eval.hh"                    // for EvalState, PrimOp
+#include "gc-small-vector.hh"         // for SmallTemporaryValueVector, Sma...
+#include "pos-idx.hh"                 // for PosIdx, noPos
+#include "primops.hh"                 // for getAttr, CompareValues, prim_l...
+#include "symbol-table.hh"            // for Symbol, SymbolTable
+#include "util.hh"                    // for enumerate
+#include "value.hh"                   // for Value, ValueVector
+
+namespace nix {
+
+static void anyOrAll(bool any, EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    const std::string fun = any ? "builtins.any" : "builtins.all";
+    state.forceFunction(
+        *args[0], pos, std::string("while evaluating the first argument passed to " + fun)
+    );
+    state.forceList(
+        *args[1], pos, std::string("while evaluating the second argument passed to " + fun)
+    );
+    Value vTmp;
+    for (auto elem : args[1]->listItems()) {
+        state.callFunction(*args[0], *elem, vTmp, pos);
+        bool res = state.forceBool(
+            vTmp, pos, "while evaluating the return value of the function passed to " + fun
+        );
+        if (res == any) {
+            v.mkBool(any);
+            return;
+        }
+    }
+    v.mkBool(!any);
+}
+
+static void elemAt(EvalState & state, const PosIdx pos, Value & list, int n, Value & v)
+{
+    state.forceList(list, pos, "while evaluating the first argument passed to builtins.elemAt");
+    if (n < 0 || (unsigned int) n >= list.listSize()) {
+        state.error<EvalError>("list index %1% is out of bounds", n).atPos(pos).debugThrow();
+    }
+    state.forceValue(*list.listElems()[n], pos);
+    v = *list.listElems()[n];
+}
+
+/**
+ * builtins.all
+ */
+
+static void prim_all(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    anyOrAll(false, state, pos, args, v);
+}
+
+PrimOp primop_all({
+    .name = "__all",
+    .args = {"pred", "list"},
+    .doc = R"(
+      Return `true` if the function *pred* returns `true` for all elements
+      of *list*, and `false` otherwise.
+    )",
+    .fun = prim_all,
+});
+
+/**
+ * builtins.any
+ */
+
+static void prim_any(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    anyOrAll(true, state, pos, args, v);
+}
+
+PrimOp primop_any({
+    .name = "__any",
+    .args = {"pred", "list"},
+    .doc = R"(
+      Return `true` if the function *pred* returns `true` for at least one
+      element of *list*, and `false` otherwise.
+    )",
+    .fun = prim_any,
+});
+
+/**
+ * builtins.concatLists
+ */
+
+static void prim_concatLists(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(
+        *args[0], pos, "while evaluating the first argument passed to builtins.concatLists"
+    );
+    state.concatLists(
+        v,
+        args[0]->listSize(),
+        args[0]->listElems(),
+        pos,
+        "while evaluating a value of the list passed to builtins.concatLists"
+    );
+}
+
+PrimOp primop_concatLists({
+    .name = "__concatLists",
+    .args = {"lists"},
+    .doc = R"(
+      Concatenate a list of lists into a single list.
+    )",
+    .fun = prim_concatLists,
+});
+
+/**
+ * builtins.concatMap
+ */
+
+static void prim_concatMap(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.concatMap"
+    );
+    state.forceList(
+        *args[1], pos, "while evaluating the second argument passed to builtins.concatMap"
+    );
+    auto nrLists = args[1]->listSize();
+
+    // List of returned lists before concatenation. References to these Values must NOT be
+    // persisted.
+    SmallTemporaryValueVector<conservativeStackReservation> lists(nrLists);
+    size_t len = 0;
+
+    for (unsigned int n = 0; n < nrLists; ++n) {
+        Value * vElem = args[1]->listElems()[n];
+        state.callFunction(*args[0], *vElem, lists[n], pos);
+        state.forceList(
+            lists[n],
+            lists[n].determinePos(args[0]->determinePos(pos)),
+            "while evaluating the return value of the function passed to builtins.concatMap"
+        );
+        len += lists[n].listSize();
+    }
+
+    state.mkList(v, len);
+    auto out = v.listElems();
+    for (unsigned int n = 0, pos = 0; n < nrLists; ++n) {
+        auto l = lists[n].listSize();
+        if (l) {
+            memcpy(out + pos, lists[n].listElems(), l * sizeof(Value *));
+        }
+        pos += l;
+    }
+}
+
+PrimOp primop_concatMap({
+    .name = "__concatMap",
+    .args = {"f", "list"},
+    .doc = R"(
+      This function is equivalent to `builtins.concatLists (map f list)`
+      but is more efficient.
+    )",
+    .fun = prim_concatMap,
+});
+
+/**
+ * builtins.elem
+ */
+
+static void prim_elem(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    bool res = false;
+    state.forceList(*args[1], pos, "while evaluating the second argument passed to builtins.elem");
+    for (auto elem : args[1]->listItems()) {
+        if (state.eqValues(
+                *args[0],
+                *elem,
+                pos,
+                "while searching for the presence of the given element in the list"
+            ))
+        {
+            res = true;
+            break;
+        }
+    }
+    v.mkBool(res);
+}
+
+PrimOp primop_elem({
+    .name = "__elem",
+    .args = {"x", "xs"},
+    .doc = R"(
+      Return `true` if a value equal to *x* occurs in the list *xs*, and
+      `false` otherwise.
+    )",
+    .fun = prim_elem,
+});
+
+/**
+ * builtins.elemAt
+ */
+
+static void prim_elemAt(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    elemAt(
+        state,
+        pos,
+        *args[0],
+        state.forceInt(
+            *args[1], pos, "while evaluating the second argument passed to builtins.elemAt"
+        ),
+        v
+    );
+}
+
+PrimOp primop_elemAt({
+    .name = "__elemAt",
+    .args = {"xs", "n"},
+    .doc = R"(
+      Return element *n* from the list *xs*. Elements are counted starting
+      from 0. A fatal error occurs if the index is out of bounds.
+    )",
+    .fun = prim_elemAt,
+});
+
+/**
+ * builtins.filter
+ */
+
+static void prim_filter(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(
+        *args[1], pos, "while evaluating the second argument passed to builtins.filter"
+    );
+
+    if (args[1]->listSize() == 0) {
+        v = *args[1];
+        return;
+    }
+
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.filter"
+    );
+
+    SmallValueVector<nonRecursiveStackReservation> vs(args[1]->listSize());
+    size_t k = 0;
+
+    bool same = true;
+    for (unsigned int n = 0; n < args[1]->listSize(); ++n) {
+        Value res;
+        state.callFunction(*args[0], *args[1]->listElems()[n], res, noPos);
+        if (state.forceBool(
+                res,
+                pos,
+                "while evaluating the return value of the filtering function passed to "
+                "builtins.filter"
+            ))
+        {
+            vs[k++] = args[1]->listElems()[n];
+        } else {
+            same = false;
+        }
+    }
+
+    if (same) {
+        v = *args[1];
+    } else {
+        state.mkList(v, k);
+        for (unsigned int n = 0; n < k; ++n) {
+            v.listElems()[n] = vs[n];
+        }
+    }
+}
+
+/**
+ * builtins.foldl'
+ */
+
+static void prim_foldlStrict(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.foldlStrict"
+    );
+    state.forceList(
+        *args[2], pos, "while evaluating the third argument passed to builtins.foldlStrict"
+    );
+    if (args[2]->listSize()) {
+        Value * vCur = args[1];
+        for (auto [n, elem] : enumerate(args[2]->listItems())) {
+            Value * vs[]{vCur, elem};
+            vCur = n == args[2]->listSize() - 1 ? &v : state.allocValue();
+            state.callFunction(*args[0], 2, vs, *vCur, pos);
+        }
+        state.forceValue(v, pos);
+    } else {
+        state.forceValue(*args[1], pos);
+        v = *args[1];
+    }
+}
+
+PrimOp primop_foldlStrict({
+    .name = "__foldl'",
+    .args = {"op", "nul", "list"},
+    .doc = R"(
+      Reduce a list by applying a binary operator, from left to right,
+      e.g. `foldl' op nul [x0 x1 x2 ...] = op (op (op nul x0) x1) x2)
+      ...`. For example, `foldl' (x: y: x + y) 0 [1 2 3]` evaluates to 6.
+      The return value of each application of `op` is evaluated immediately,
+      even for intermediate values.
+    )",
+    .fun = prim_foldlStrict,
+});
+
+/**
+ * builtins.genList
+ */
+
+static void prim_genList(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto len = state.forceInt(
+        *args[1], pos, "while evaluating the second argument passed to builtins.genList"
+    );
+    if (len < 0) {
+        state.error<EvalError>("cannot create list of size %1%", len).atPos(pos).debugThrow();
+    }
+    // More strict than striclty (!) necessary, but acceptable
+    // as evaluating map without accessing any values makes little sense.
+    state.forceFunction(
+        *args[0], noPos, "while evaluating the first argument passed to builtins.genList"
+    );
+    state.mkList(v, len);
+    for (unsigned int n = 0; n < (unsigned int) len; ++n) {
+        auto arg = state.allocValue();
+        arg->mkInt(n);
+        (v.listElems()[n] = state.allocValue())->mkApp(args[0], arg);
+    }
+}
+
+PrimOp primop_genList({
+    .name = "__genList",
+    .args = {"generator", "length"},
+    .doc = R"(
+      Generate list of size *length*, with each element *i* equal to the
+      value returned by *generator* `i`. For example,
+      ```nix
+      builtins.genList (x: x * x) 5
+      ```
+      returns the list `[ 0 1 4 9 16 ]`.
+    )",
+    .fun = prim_genList,
+});
+
+/**
+ * builtins.head
+ */
+
+static void prim_head(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    elemAt(state, pos, *args[0], 0, v);
+}
+
+PrimOp primop_head({
+    .name = "__head",
+    .args = {"list"},
+    .doc = R"(
+      Return the first element of a list; abort evaluation if the argument
+      isn’t a list or is an empty list. You can test whether a list is
+      empty by comparing it with `[]`.
+    )",
+    .fun = prim_head,
+});
+
+/**
+ * builtins.length
+ */
+
+static void prim_length(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(*args[0], pos, "while evaluating the first argument passed to builtins.length");
+    v.mkInt(args[0]->listSize());
+}
+PrimOp primop_length({
+    .name = "__length",
+    .args = {"e"},
+    .doc = R"(
+      Return the length of the list *e*.
+    )",
+    .fun = prim_length,
+});
+
+PrimOp primop_filter({
+    .name = "__filter",
+    .args = {"f", "list"},
+    .doc = R"(
+      Return a list consisting of the elements of *list* for which the
+      function *f* returns `true`.
+    )",
+    .fun = prim_filter,
+});
+
+/**
+ * builtins.listToAttrs
+ */
+
+static void prim_listToAttrs(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(*args[0], pos, "while evaluating the argument passed to builtins.listToAttrs");
+
+    auto attrs = state.buildBindings(args[0]->listSize());
+
+    std::set<Symbol> seen;
+
+    for (auto v2 : args[0]->listItems()) {
+        state.forceAttrs(
+            *v2, pos, "while evaluating an element of the list passed to builtins.listToAttrs"
+        );
+
+        Bindings::iterator j =
+            getAttr(state, state.sName, v2->attrs, "in a {name=...; value=...;} pair");
+
+        auto name = state.forceStringNoCtx(
+            *j->value,
+            j->pos,
+            "while evaluating the `name` attribute of an element of the list passed to "
+            "builtins.listToAttrs"
+        );
+
+        auto sym = state.symbols.create(name);
+        if (seen.insert(sym).second) {
+            Bindings::iterator j2 =
+                getAttr(state, state.sValue, v2->attrs, "in a {name=...; value=...;} pair");
+            attrs.insert(sym, j2->value, j2->pos);
+        }
+    }
+
+    v.mkAttrs(attrs);
+}
+
+PrimOp primop_listToAttrs({
+    .name = "__listToAttrs",
+    .args = {"e"},
+    .doc = R"(
+      Construct a set from a list specifying the names and values of each
+      attribute. Each element of the list should be a set consisting of a
+      string-valued attribute `name` specifying the name of the attribute,
+      and an attribute `value` specifying its value.
+
+      In case of duplicate occurrences of the same name, the first
+      takes precedence.
+
+      Example:
+
+      ```nix
+      builtins.listToAttrs
+        [ { name = "foo"; value = 123; }
+          { name = "bar"; value = 456; }
+          { name = "bar"; value = 420; }
+        ]
+      ```
+
+      evaluates to
+
+      ```nix
+      { foo = 123; bar = 456; }
+      ```
+    )",
+    .fun = prim_listToAttrs,
+});
+
+/**
+ * builtins.map
+ */
+
+static void prim_map(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(*args[1], pos, "while evaluating the second argument passed to builtins.map");
+
+    if (args[1]->listSize() == 0) {
+        v = *args[1];
+        return;
+    }
+
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.map"
+    );
+
+    state.mkList(v, args[1]->listSize());
+    for (unsigned int n = 0; n < v.listSize(); ++n) {
+        (v.listElems()[n] = state.allocValue())->mkApp(args[0], args[1]->listElems()[n]);
+    }
+}
+
+PrimOp primop_map({
+    .name = "map",
+    .args = {"f", "list"},
+    .doc = R"(
+      Apply the function *f* to each element in the list *list*. For
+      example,
+
+      ```nix
+      map (x: "foo" + x) [ "bar" "bla" "abc" ]
+      ```
+
+      evaluates to `[ "foobar" "foobla" "fooabc" ]`.
+    )",
+    .fun = prim_map,
+});
+
+/**
+ * builtins.partition
+ */
+
+static void prim_partition(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.partition"
+    );
+    state.forceList(
+        *args[1], pos, "while evaluating the second argument passed to builtins.partition"
+    );
+
+    auto len = args[1]->listSize();
+
+    ValueVector right, wrong;
+
+    for (unsigned int n = 0; n < len; ++n) {
+        auto vElem = args[1]->listElems()[n];
+        state.forceValue(*vElem, pos);
+        Value res;
+        state.callFunction(*args[0], *vElem, res, pos);
+        if (state.forceBool(
+                res,
+                pos,
+                "while evaluating the return value of the partition function passed to "
+                "builtins.partition"
+            ))
+        {
+            right.push_back(vElem);
+        } else {
+            wrong.push_back(vElem);
+        }
+    }
+
+    auto attrs = state.buildBindings(2);
+
+    auto & vRight = attrs.alloc(state.sRight);
+    auto rsize = right.size();
+    state.mkList(vRight, rsize);
+    if (rsize) {
+        memcpy(vRight.listElems(), right.data(), sizeof(Value *) * rsize);
+    }
+
+    auto & vWrong = attrs.alloc(state.sWrong);
+    auto wsize = wrong.size();
+    state.mkList(vWrong, wsize);
+    if (wsize) {
+        memcpy(vWrong.listElems(), wrong.data(), sizeof(Value *) * wsize);
+    }
+
+    v.mkAttrs(attrs);
+}
+
+PrimOp primop_partition({
+    .name = "__partition",
+    .args = {"pred", "list"},
+    .doc = R"(
+      Given a predicate function *pred*, this function returns an
+      attrset containing a list named `right`, containing the elements
+      in *list* for which *pred* returned `true`, and a list named
+      `wrong`, containing the elements for which it returned
+      `false`. For example,
+
+      ```nix
+      builtins.partition (x: x > 10) [1 23 9 3 42]
+      ```
+
+      evaluates to
+
+      ```nix
+      { right = [ 23 42 ]; wrong = [ 1 9 3 ]; }
+      ```
+    )",
+    .fun = prim_partition,
+});
+
+/**
+ * builtins.sort
+ */
+
+static void prim_sort(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(*args[1], pos, "while evaluating the second argument passed to builtins.sort");
+
+    auto len = args[1]->listSize();
+    if (len == 0) {
+        v = *args[1];
+        return;
+    }
+
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.sort"
+    );
+
+    state.mkList(v, len);
+    for (unsigned int n = 0; n < len; ++n) {
+        state.forceValue(*args[1]->listElems()[n], pos);
+        v.listElems()[n] = args[1]->listElems()[n];
+    }
+
+    auto comparator = [&](Value * a, Value * b) {
+        /* Optimization: if the comparator is lessThan, bypass
+           callFunction. */
+        /* TODO: (layus) this is absurd. An optimisation like this
+           should be outside the lambda creation */
+        if (args[0]->isPrimOp()) {
+            auto ptr = args[0]->primOp->fun.target<decltype(&prim_lessThan)>();
+            if (ptr && *ptr == prim_lessThan) {
+                return CompareValues(state, noPos, "while evaluating the ordering function passed to builtins.sort")(a, b);
+            }
+        }
+
+        Value * vs[] = {a, b};
+        Value vBool;
+        state.callFunction(*args[0], 2, vs, vBool, noPos);
+        return state.forceBool(
+            vBool,
+            pos,
+            "while evaluating the return value of the sorting function passed to builtins.sort"
+        );
+    };
+
+    /* FIXME: std::sort can segfault if the comparator is not a strict
+       weak ordering. What to do? std::stable_sort() seems more
+       resilient, but no guarantees... */
+    std::stable_sort(v.listElems(), v.listElems() + len, comparator);
+}
+
+PrimOp primop_sort({
+    .name = "__sort",
+    .args = {"comparator", "list"},
+    .doc = R"(
+      Return *list* in sorted order. It repeatedly calls the function
+      *comparator* with two elements. The comparator should return `true`
+      if the first element is less than the second, and `false` otherwise.
+      For example,
+
+      ```nix
+      builtins.sort builtins.lessThan [ 483 249 526 147 42 77 ]
+      ```
+
+      produces the list `[ 42 77 147 249 483 526 ]`.
+
+      This is a stable sort: it preserves the relative order of elements
+      deemed equal by the comparator.
+    )",
+    .fun = prim_sort,
+});
+
+/**
+ * builtins.tail
+ */
+
+static void prim_tail(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(*args[0], pos, "while evaluating the first argument passed to builtins.tail");
+    if (args[0]->listSize() == 0) {
+        state.error<EvalError>("'tail' called on an empty list").atPos(pos).debugThrow();
+    }
+
+    state.mkList(v, args[0]->listSize() - 1);
+    for (unsigned int n = 0; n < v.listSize(); ++n) {
+        v.listElems()[n] = args[0]->listElems()[n + 1];
+    }
+}
+
+PrimOp primop_tail({
+    .name = "__tail",
+    .args = {"list"},
+    .doc = R"(
+      Return the second to last elements of a list; abort evaluation if
+      the argument isn’t a list or is an empty list.
+
+      > **Warning**
+      >
+      > This function should generally be avoided since it's inefficient:
+      > unlike Haskell's `tail`, it takes O(n) time, so recursing over a
+      > list by repeatedly calling `tail` takes O(n^2) time.
+    )",
+    .fun = prim_tail,
+});
+}

src/libexpr/primops/path.cc

@@ -0,0 +1,966 @@
+#include <sys/stat.h>                              // for stat, S_ISDIR, S_I...
+#include <list>                                    // for list
+#include <optional>                                // for optional, operator!=
+#include <set>                                     // for _Rb_tree_const_ite...
+#include <string>                                  // for basic_string, char...
+#include <string_view>                             // for string_view
+#include <utility>                                 // for move, pair
+#include <variant>                                 // for get_if
+#include "attr-set.hh"                             // for Attr, Bindings
+#include "canon-path.hh"                           // for CanonPath
+#include "content-address.hh"                      // for FileIngestionMethod
+#include "derivations.hh"                          // for hashPlaceholder
+#include "derived-path.hh"                         // for SingleDerivedPath
+#include "error.hh"                                // for Error, SysError
+#include "eval-error.hh"                           // for EvalError, EvalErr...
+#include "eval-settings.hh"                        // for EvalSettings, eval...
+#include "eval.hh"                                 // for EvalState, PrimOp
+#include "experimental-features.hh"                // for ExperimentalFeature
+#include "fetch-to-store.hh"                       // for fetchToStore
+#include "globals.hh"                              // for Settings, settings
+#include "hash.hh"                                 // for Hash, newHashAllow...
+#include "input-accessor.hh"                       // for InputAccessor
+#include "outputs-spec.hh"                         // for OutputNameView
+#include "path-info.hh"                            // for ValidPathInfo
+#include "path-references.hh"                      // for PathRefScanSink
+#include "path.hh"                                 // for StorePath, StorePa...
+#include "pos-idx.hh"                              // for PosIdx, noPos
+#include "pos-table.hh"                            // for PosTable
+#include "primops.hh"                              // for RealisePathFlags
+#include "ref.hh"                                  // for ref, make_ref
+#include "search-path.hh"                          // for SearchPath
+#include "serialise.hh"                            // for operator<<, Sink
+#include "source-path.hh"                          // for SourcePath, Symlin...
+#include "store-api.hh"                            // for Store
+#include "symbol-table.hh"                         // for Symbol, SymbolStr
+#include "types.hh"                                // for BackedStringView
+#include "util.hh"                                 // for rewriteStrings
+#include "value.hh"                                // for Value, nPath, nString
+#include "value/context.hh"                        // for NixStringContext
+
+namespace nix {
+
+// TODO: Use a const map instead ?
+static std::string_view fileTypeToString(InputAccessor::Type type)
+{
+    return type == InputAccessor::Type::tRegular  ? "regular"
+        : type == InputAccessor::Type::tDirectory ? "directory"
+        : type == InputAccessor::Type::tSymlink   ? "symlink"
+                                                  : "unknown";
+}
+
+static void addPath(
+    EvalState & state,
+    const PosIdx pos,
+    std::string_view name,
+    Path path,
+    Value * filterFun,
+    FileIngestionMethod method,
+    const std::optional<Hash> expectedHash,
+    Value & v,
+    const NixStringContext & context
+)
+{
+    try {
+        // FIXME: handle CA derivation outputs (where path needs to
+        // be rewritten to the actual output).
+        auto rewrites = state.realiseContext(context);
+        path = state.toRealPath(rewriteStrings(path, rewrites), context);
+
+        StorePathSet refs;
+
+        if (state.store->isInStore(path)) {
+            try {
+                auto [storePath, subPath] = state.store->toStorePath(path);
+                // FIXME: we should scanForReferences on the path before adding it
+                refs = state.store->queryPathInfo(storePath)->references;
+                path = state.store->toRealPath(storePath) + subPath;
+            } catch (Error &) { // FIXME: should be InvalidPathError
+            }
+        }
+
+        path = evalSettings.pureEval && expectedHash
+            ? path
+            : state.checkSourcePath(CanonPath(path)).path.abs();
+
+        PathFilter filter = filterFun ? ([&](const Path & path) {
+            auto st = lstat(path);
+
+            /* Call the filter function.  The first argument is the path,
+               the second is a string indicating the type of the file. */
+            Value arg1;
+            arg1.mkString(path);
+
+            Value arg2;
+            arg2.mkString(
+                S_ISREG(st.st_mode)       ? "regular"
+                    : S_ISDIR(st.st_mode) ? "directory"
+                    : S_ISLNK(st.st_mode) ? "symlink"
+                                          : "unknown" /* not supported, will fail! */
+            );
+
+            Value * args[]{&arg1, &arg2};
+            Value res;
+            state.callFunction(*filterFun, 2, args, res, pos);
+
+            return state.forceBool(
+                res, pos, "while evaluating the return value of the path filter function"
+            );
+        })
+                                      : defaultPathFilter;
+
+        std::optional<StorePath> expectedStorePath;
+        if (expectedHash) {
+            expectedStorePath = state.store->makeFixedOutputPath(
+                name,
+                FixedOutputInfo{
+                    .method = method,
+                    .hash = *expectedHash,
+                    .references = {},
+                }
+            );
+        }
+
+        if (!expectedHash || !state.store->isValidPath(*expectedStorePath)) {
+            auto dstPath = fetchToStore(
+                *state.store, state.rootPath(CanonPath(path)), name, method, &filter, state.repair
+            );
+            if (expectedHash && expectedStorePath != dstPath) {
+                state
+                    .error<EvalError>(
+                        "store path mismatch in (possibly filtered) path added from '%s'", path
+                    )
+                    .atPos(pos)
+                    .debugThrow();
+            }
+            state.allowAndSetStorePathString(dstPath, v);
+        } else {
+            state.allowAndSetStorePathString(*expectedStorePath, v);
+        }
+    } catch (Error & e) {
+        e.addTrace(state.positions[pos], "while adding path '%s'", path);
+        throw;
+    }
+}
+
+SourcePath realisePath(EvalState & state, const PosIdx pos, Value & v, const RealisePathFlags flags)
+{
+    NixStringContext context;
+
+    auto path = state.coerceToPath(noPos, v, context, "while realising the context of a path");
+
+    try {
+        StringMap rewrites = state.realiseContext(context);
+
+        auto realPath = state.rootPath(CanonPath(state.toRealPath(rewriteStrings(path.path.abs(), rewrites), context)));
+
+        return flags.checkForPureEval
+            ? state.checkSourcePath(realPath)
+            : realPath;
+    } catch (Error & e) {
+        e.addTrace(state.positions[pos], "while realising the context of path '%s'", path);
+        throw;
+    }
+}
+
+/**
+ * builtins.dirOf
+ */
+
+static void prim_dirOf(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    if (args[0]->type() == nPath) {
+        auto path = args[0]->path();
+        v.mkPath(path.path.isRoot() ? path : path.parent());
+    } else {
+        NixStringContext context;
+        auto path = state.coerceToString(
+            pos,
+            *args[0],
+            context,
+            "while evaluating the first argument passed to 'builtins.dirOf'",
+            false,
+            false
+        );
+        auto dir = dirOf(*path);
+        v.mkString(dir, context);
+    }
+}
+
+PrimOp primop_dirOf({
+    .name = "dirOf",
+    .args = {"s"},
+    .doc = R"(
+      Return the directory part of the string *s*, that is, everything
+      before the final slash in the string. This is similar to the GNU
+      `dirname` command.
+    )",
+    .fun = prim_dirOf,
+});
+
+/**
+ * builtins.filterSource
+ */
+
+static void prim_filterSource(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto path = state.coerceToPath(
+        pos,
+        *args[1],
+        context,
+        "while evaluating the second argument (the path to filter) passed to builtins.filterSource"
+    );
+    state.forceFunction(
+        *args[0], pos, "while evaluating the first argument passed to builtins.filterSource"
+    );
+    addPath(
+        state,
+        pos,
+        path.baseName(),
+        path.path.abs(),
+        args[0],
+        FileIngestionMethod::Recursive,
+        std::nullopt,
+        v,
+        context
+    );
+}
+
+PrimOp primop_filterSource({
+    .name = "__filterSource",
+    .args = {"e1", "e2"},
+    .doc = R"(
+      > **Warning**
+      >
+      > `filterSource` should not be used to filter store paths. Since
+      > `filterSource` uses the name of the input directory while naming
+      > the output directory, doing so will produce a directory name in
+      > the form of `<hash2>-<hash>-<name>`, where `<hash>-<name>` is
+      > the name of the input directory. Since `<hash>` depends on the
+      > unfiltered directory, the name of the output directory will
+      > indirectly depend on files that are filtered out by the
+      > function. This will trigger a rebuild even when a filtered out
+      > file is changed. Use `builtins.path` instead, which allows
+      > specifying the name of the output directory.
+
+      This function allows you to copy sources into the Nix store while
+      filtering certain files. For instance, suppose that you want to use
+      the directory `source-dir` as an input to a Nix expression, e.g.
+
+      ```nix
+      stdenv.mkDerivation {
+        ...
+        src = ./source-dir;
+      }
+      ```
+
+      However, if `source-dir` is a Subversion working copy, then all
+      those annoying `.svn` subdirectories will also be copied to the
+      store. Worse, the contents of those directories may change a lot,
+      causing lots of spurious rebuilds. With `filterSource` you can
+      filter out the `.svn` directories:
+
+      ```nix
+      src = builtins.filterSource
+        (path: type: type != "directory" || baseNameOf path != ".svn")
+        ./source-dir;
+      ```
+
+      Thus, the first argument *e1* must be a predicate function that is
+      called for each regular file, directory or symlink in the source
+      tree *e2*. If the function returns `true`, the file is copied to the
+      Nix store, otherwise it is omitted. The function is called with two
+      arguments. The first is the full path of the file. The second is a
+      string that identifies the type of the file, which is either
+      `"regular"`, `"directory"`, `"symlink"` or `"unknown"` (for other
+      kinds of files such as device nodes or fifos — but note that those
+      cannot be copied to the Nix store, so if the predicate returns
+      `true` for them, the copy will fail). If you exclude a directory,
+      the entire corresponding subtree of *e2* will be excluded.
+    )",
+    .fun = prim_filterSource,
+});
+
+/**
+ * builtins.findFile
+ */
+
+static void prim_findFile(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(
+        *args[0], pos, "while evaluating the first argument passed to builtins.findFile"
+    );
+
+    SearchPath searchPath;
+
+    for (auto v2 : args[0]->listItems()) {
+        state.forceAttrs(
+            *v2, pos, "while evaluating an element of the list passed to builtins.findFile"
+        );
+
+        std::string prefix;
+        Bindings::iterator i = v2->attrs->find(state.sPrefix);
+        if (i != v2->attrs->end()) {
+            prefix = state.forceStringNoCtx(
+                *i->value,
+                pos,
+                "while evaluating the `prefix` attribute of an element of the list passed to "
+                "builtins.findFile"
+            );
+        }
+
+        i = getAttr(state, state.sPath, v2->attrs, "in an element of the __nixPath");
+
+        NixStringContext context;
+        auto path = state
+                        .coerceToString(
+                            pos,
+                            *i->value,
+                            context,
+                            "while evaluating the `path` attribute of an element of the list "
+                            "passed to builtins.findFile",
+                            false,
+                            false
+                        )
+                        .toOwned();
+
+        try {
+            auto rewrites = state.realiseContext(context);
+            path = rewriteStrings(path, rewrites);
+        } catch (InvalidPathError & e) {
+            state.error<EvalError>("cannot find '%1%', since path '%2%' is not valid", path, e.path)
+                .atPos(pos)
+                .debugThrow();
+        }
+
+        searchPath.elements.emplace_back(SearchPath::Elem{
+            .prefix = SearchPath::Prefix{.s = prefix},
+            .path = SearchPath::Path{.s = path},
+        });
+    }
+
+    auto path = state.forceStringNoCtx(
+        *args[1], pos, "while evaluating the second argument passed to builtins.findFile"
+    );
+
+    v.mkPath(state.checkSourcePath(state.findFile(searchPath, path, pos)));
+}
+
+PrimOp primop_findFile(PrimOp{
+    .name = "__findFile",
+    .args = {"search path", "lookup path"},
+    .doc = R"(
+      Look up the given path with the given search path.
+
+      A search path is represented list of [attribute sets](./values.md#attribute-set) with two attributes, `prefix`, and `path`.
+      `prefix` is a relative path.
+      `path` denotes a file system location; the exact syntax depends on the command line interface.
+
+      Examples of search path attribute sets:
+
+      - ```
+        {
+          prefix = "nixos-config";
+          path = "/etc/nixos/configuration.nix";
+        }
+        ```
+
+      - ```
+        {
+          prefix = "";
+          path = "/nix/var/nix/profiles/per-user/root/channels";
+        }
+        ```
+
+      The lookup algorithm checks each entry until a match is found, returning a [path value](@docroot@/language/values.html#type-path) of the match.
+
+      This is the process for each entry:
+      If the lookup path matches `prefix`, then the remainder of the lookup path (the "suffix") is searched for within the directory denoted by `patch`.
+      Note that the `path` may need to be downloaded at this point to look inside.
+      If the suffix is found inside that directory, then the entry is a match;
+      the combined absolute path of the directory (now downloaded if need be) and the suffix is returned.
+
+      The syntax
+
+      ```nix
+      <nixpkgs>
+      ```
+
+      is equivalent to:
+
+      ```nix
+      builtins.findFile builtins.nixPath "nixpkgs"
+      ```
+    )",
+    .fun = prim_findFile,
+});
+
+/**
+ * builtins.outputOf
+ */
+
+static void prim_outputOf(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    SingleDerivedPath drvPath = state.coerceToSingleDerivedPath(
+        pos, *args[0], "while evaluating the first argument to builtins.outputOf"
+    );
+
+    OutputNameView outputName = state.forceStringNoCtx(
+        *args[1], pos, "while evaluating the second argument to builtins.outputOf"
+    );
+
+    state.mkSingleDerivedPathString(
+        SingleDerivedPath::Built{
+            .drvPath = make_ref<SingleDerivedPath>(drvPath),
+            .output = std::string{outputName},
+        },
+        v
+    );
+}
+
+PrimOp primop_outputOf({
+    .name = "__outputOf",
+    .args = {"derivation-reference", "output-name"},
+    .doc = R"(
+      Return the output path of a derivation, literally or using a placeholder if needed.
+
+      If the derivation has a statically-known output path (i.e. the derivation output is input-addressed, or fixed content-addresed), the output path will just be returned.
+      But if the derivation is content-addressed or if the derivation is itself not-statically produced (i.e. is the output of another derivation), a placeholder will be returned instead.
+
+      *`derivation reference`* must be a string that may contain a regular store path to a derivation, or may be a placeholder reference. If the derivation is produced by a derivation, you must explicitly select `drv.outPath`.
+      This primop can be chained arbitrarily deeply.
+      For instance,
+
+      ```nix
+      builtins.outputOf
+        (builtins.outputOf myDrv "out)
+        "out"
+      ```
+
+      will return a placeholder for the output of the output of `myDrv`.
+
+      This primop corresponds to the `^` sigil for derivable paths, e.g. as part of installable syntax on the command line.
+    )",
+    .fun = prim_outputOf,
+    .experimentalFeature = Xp::DynamicDerivations,
+});
+
+/**
+ * builtins.path
+ */
+
+static void prim_path(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    std::optional<SourcePath> path;
+    std::string name;
+    Value * filterFun = nullptr;
+    auto method = FileIngestionMethod::Recursive;
+    std::optional<Hash> expectedHash;
+    NixStringContext context;
+
+    state.forceAttrs(*args[0], pos, "while evaluating the argument passed to 'builtins.path'");
+
+    for (auto & attr : *args[0]->attrs) {
+        auto n = state.symbols[attr.name];
+        if (n == "path") {
+            path.emplace(state.coerceToPath(
+                attr.pos,
+                *attr.value,
+                context,
+                "while evaluating the 'path' attribute passed to 'builtins.path'"
+            ));
+        } else if (attr.name == state.sName) {
+            name = state.forceStringNoCtx(
+                *attr.value,
+                attr.pos,
+                "while evaluating the `name` attribute passed to builtins.path"
+            );
+        } else if (n == "filter") {
+            state.forceFunction(
+                *(filterFun = attr.value),
+                attr.pos,
+                "while evaluating the `filter` parameter passed to builtins.path"
+            );
+        } else if (n == "recursive") {
+            method = FileIngestionMethod{state.forceBool(
+                *attr.value,
+                attr.pos,
+                "while evaluating the `recursive` attribute passed to builtins.path"
+            )};
+        } else if (n == "sha256") {
+            expectedHash = newHashAllowEmpty(
+                state.forceStringNoCtx(
+                    *attr.value,
+                    attr.pos,
+                    "while evaluating the `sha256` attribute passed to builtins.path"
+                ),
+                htSHA256
+            );
+        } else {
+            state
+                .error<EvalError>(
+                    "unsupported argument '%1%' to 'addPath'", state.symbols[attr.name]
+                )
+                .atPos(attr.pos)
+                .debugThrow();
+        }
+    }
+    if (!path) {
+        state
+            .error<EvalError>(
+                "missing required 'path' attribute in the first argument to builtins.path"
+            )
+            .atPos(pos)
+            .debugThrow();
+    }
+    if (name.empty()) {
+        name = path->baseName();
+    }
+
+    addPath(state, pos, name, path->path.abs(), filterFun, method, expectedHash, v, context);
+}
+
+PrimOp primop_path({
+    .name = "__path",
+    .args = {"args"},
+    .doc = R"(
+      An enrichment of the built-in path type, based on the attributes
+      present in *args*. All are optional except `path`:
+
+        - path\
+          The underlying path.
+
+        - name\
+          The name of the path when added to the store. This can used to
+          reference paths that have nix-illegal characters in their names,
+          like `@`.
+
+        - filter\
+          A function of the type expected by `builtins.filterSource`,
+          with the same semantics.
+
+        - recursive\
+          When `false`, when `path` is added to the store it is with a
+          flat hash, rather than a hash of the NAR serialization of the
+          file. Thus, `path` must refer to a regular file, not a
+          directory. This allows similar behavior to `fetchurl`. Defaults
+          to `true`.
+
+        - sha256\
+          When provided, this is the expected hash of the file at the
+          path. Evaluation will fail if the hash is incorrect, and
+          providing a hash allows `builtins.path` to be used even when the
+          `pure-eval` nix config option is on.
+    )",
+    .fun = prim_path,
+});
+
+/**
+ * builtins.pathExists
+ */
+
+static void prim_pathExists(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto & arg = *args[0];
+
+    /* We don’t check the path right now, because we don’t want to
+       throw if the path isn’t allowed, but just return false (and we
+       can’t just catch the exception here because we still want to
+       throw if something in the evaluation of `arg` tries to
+       access an unauthorized path). */
+    auto path = realisePath(state, pos, arg, {.checkForPureEval = false});
+
+    /* SourcePath doesn't know about trailing slash. */
+    auto mustBeDir =
+        arg.type() == nString && (arg.str().ends_with("/") || arg.str().ends_with("/."));
+
+    try {
+        auto checked = state.checkSourcePath(path).resolveSymlinks(
+            mustBeDir ? SymlinkResolution::Full : SymlinkResolution::Ancestors
+        );
+
+        auto st = checked.maybeLstat();
+        auto exists = st && (!mustBeDir || st->type == InputAccessor::tDirectory);
+        v.mkBool(exists);
+    } catch (SysError & e) {
+        /* Don't give away info from errors while canonicalising
+           ‘path’ in restricted mode. */
+        v.mkBool(false);
+    } catch (RestrictedPathError & e) {
+        v.mkBool(false);
+    }
+}
+
+PrimOp primop_pathExists({
+    .name = "__pathExists",
+    .args = {"path"},
+    .doc = R"(
+      Return `true` if the path *path* exists at evaluation time, and
+      `false` otherwise.
+    )",
+    .fun = prim_pathExists,
+});
+
+/**
+ * builtins.palceholder
+ */
+
+/* Return a placeholder string for the specified output that will be
+   substituted by the corresponding output path at build time. For
+   example, 'placeholder "out"' returns the string
+   /1rz4g4znpzjwh1xymhjpm42vipw92pr73vdgl6xs1hycac8kf2n9. At build
+   time, any occurrence of this string in an derivation attribute will
+   be replaced with the concrete path in the Nix store of the output
+   ‘out’. */
+static void prim_placeholder(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    v.mkString(hashPlaceholder(state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.placeholder"
+    )));
+}
+
+PrimOp primop_placeholder({
+    .name = "placeholder",
+    .args = {"output"},
+    .doc = R"(
+      Return a placeholder string for the specified *output* that will be
+      substituted by the corresponding output path at build time. Typical
+      outputs would be `"out"`, `"bin"` or `"dev"`.
+    )",
+    .fun = prim_placeholder,
+});
+
+/**
+ * builtins.toPath
+ * WARNING: deprecated
+ */
+
+static void prim_toPath(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto path = state.coerceToPath(
+        pos, *args[0], context, "while evaluating the first argument passed to builtins.toPath"
+    );
+    v.mkString(path.path.abs(), context);
+}
+
+PrimOp primop_toPath({
+    .name = "__toPath",
+    .args = {"s"},
+    .doc = R"(
+      **DEPRECATED.** Use `/. + "/path"` to convert a string into an absolute
+      path. For relative paths, use `./. + "/path"`.
+    )",
+    .fun = prim_toPath,
+});
+
+/**
+ * builtins.readDir
+ */
+
+static void prim_readDir(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto path = realisePath(state, pos, *args[0]);
+
+    // Retrieve directory entries for all nodes in a directory.
+    // This is similar to `getFileType` but is optimized to reduce system calls
+    // on many systems.
+    auto entries = path.readDirectory();
+    auto attrs = state.buildBindings(entries.size());
+
+    // If we hit unknown directory entry types we may need to fallback to
+    // using `getFileType` on some systems.
+    // In order to reduce system calls we make each lookup lazy by using
+    // `builtins.readFileType` application.
+    Value * readFileType = nullptr;
+
+    for (auto & [name, type] : entries) {
+        auto & attr = attrs.alloc(name);
+        if (!type) {
+            // Some filesystems or operating systems may not be able to return
+            // detailed node info quickly in this case we produce a thunk to
+            // query the file type lazily.
+            auto epath = state.allocValue();
+            epath->mkPath(path + name);
+            if (!readFileType) {
+                readFileType = &state.getBuiltin("readFileType");
+            }
+            attr.mkApp(readFileType, epath);
+        } else {
+            // This branch of the conditional is much more likely.
+            // Here we just stringize the directory entry type.
+            attr.mkString(fileTypeToString(*type));
+        }
+    }
+
+    v.mkAttrs(attrs);
+}
+
+PrimOp primop_readDir({
+    .name = "__readDir",
+    .args = {"path"},
+    .doc = R"(
+      Return the contents of the directory *path* as a set mapping
+      directory entries to the corresponding file type. For instance, if
+      directory `A` contains a regular file `B` and another directory
+      `C`, then `builtins.readDir ./A` will return the set
+
+      ```nix
+      { B = "regular"; C = "directory"; }
+      ```
+
+      The possible values for the file type are `"regular"`,
+      `"directory"`, `"symlink"` and `"unknown"`.
+    )",
+    .fun = prim_readDir,
+});
+
+/**
+ * builtins.readFile
+ */
+
+static void prim_readFile(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto path = realisePath(state, pos, *args[0]);
+    auto s = path.readFile();
+    if (s.find((char) 0) != std::string::npos) {
+        state
+            .error<EvalError>(
+                "the contents of the file '%1%' cannot be represented as a Nix string", path
+            )
+            .atPos(pos)
+            .debugThrow();
+    }
+    StorePathSet refs;
+    if (state.store->isInStore(path.path.abs())) {
+        try {
+            refs = state.store->queryPathInfo(state.store->toStorePath(path.path.abs()).first)
+                       ->references;
+        } catch (Error &) { // FIXME: should be InvalidPathError
+        }
+        // Re-scan references to filter down to just the ones that actually occur in the file.
+        auto refsSink = PathRefScanSink::fromPaths(refs);
+        refsSink << s;
+        refs = refsSink.getResultPaths();
+    }
+    NixStringContext context;
+    for (auto && p : std::move(refs)) {
+        context.insert(NixStringContextElem::Opaque{
+            .path = std::move((StorePath &&) p),
+        });
+    }
+    v.mkString(s, context);
+}
+
+PrimOp primop_readFile({
+    .name = "__readFile",
+    .args = {"path"},
+    .doc = R"(
+      Return the contents of the file *path* as a string.
+    )",
+    .fun = prim_readFile,
+});
+
+/**
+ * builtins.readFileType
+ */
+
+static void prim_readFileType(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto path = realisePath(state, pos, *args[0]);
+    /* Retrieve the directory entry type and stringize it. */
+    v.mkString(fileTypeToString(path.lstat().type));
+}
+
+PrimOp primop_readFileType({
+    .name = "__readFileType",
+    .args = {"p"},
+    .doc = R"(
+      Determine the directory entry type of a filesystem node, being
+      one of "directory", "regular", "symlink", or "unknown".
+    )",
+    .fun = prim_readFileType,
+});
+
+/**
+ * builtins.storePath
+ */
+
+static void prim_storePath(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    if (evalSettings.pureEval) {
+        state.error<EvalError>("'%s' is not allowed in pure evaluation mode", "builtins.storePath")
+            .atPos(pos)
+            .debugThrow();
+    }
+
+    NixStringContext context;
+    auto path = state
+                    .checkSourcePath(state.coerceToPath(
+                        pos,
+                        *args[0],
+                        context,
+                        "while evaluating the first argument passed to builtins.storePath"
+                    ))
+                    .path;
+    /* Resolve symlinks in ‘path’, unless ‘path’ itself is a symlink
+       directly in the store.  The latter condition is necessary so
+       e.g. nix-push does the right thing. */
+    if (!state.store->isStorePath(path.abs())) {
+        path = CanonPath(canonPath(path.abs(), true));
+    }
+    if (!state.store->isInStore(path.abs())) {
+        state.error<EvalError>("path '%1%' is not in the Nix store", path).atPos(pos).debugThrow();
+    }
+    auto path2 = state.store->toStorePath(path.abs()).first;
+    if (!settings.readOnlyMode) {
+        state.store->ensurePath(path2);
+    }
+    context.insert(NixStringContextElem::Opaque{.path = path2});
+    v.mkString(path.abs(), context);
+}
+
+PrimOp primop_storePath({
+    .name = "__storePath",
+    .args = {"path"},
+    .doc = R"(
+      This function allows you to define a dependency on an already
+      existing store path. For example, the derivation attribute `src
+      = builtins.storePath /nix/store/f1d18v1y…-source` causes the
+      derivation to depend on the specified path, which must exist or
+      be substitutable. Note that this differs from a plain path
+      (e.g. `src = /nix/store/f1d18v1y…-source`) in that the latter
+      causes the path to be *copied* again to the Nix store, resulting
+      in a new path (e.g. `/nix/store/ld01dnzc…-source-source`).
+
+      Not available in [pure evaluation mode](@docroot@/command-ref/conf-file.md#conf-pure-eval).
+
+      See also [`builtins.fetchClosure`](#builtins-fetchClosure).
+    )",
+    .fun = prim_storePath,
+});
+
+/**
+ * builtins.toFile
+ */
+
+static void prim_toFile(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    std::string name(state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.toFile"
+    ));
+    std::string contents(state.forceString(
+        *args[1], context, pos, "while evaluating the second argument passed to builtins.toFile"
+    ));
+
+    StorePathSet refs;
+
+    for (auto c : context) {
+        if (auto p = std::get_if<NixStringContextElem::Opaque>(&c.raw)) {
+            refs.insert(p->path);
+        } else {
+            state
+                .error<EvalError>(
+                    "files created by %1% may not reference derivations, but %2% references %3%",
+                    "builtins.toFile",
+                    name,
+                    c.to_string()
+                )
+                .atPos(pos)
+                .debugThrow();
+        }
+    }
+
+    auto storePath = settings.readOnlyMode
+        ? state.store->computeStorePathForText(name, contents, refs)
+        : state.store->addTextToStore(name, contents, refs, state.repair);
+
+    /* Note: we don't need to add `context' to the context of the
+       result, since `storePath' itself has references to the paths
+       used in args[1]. */
+
+    /* Add the output of this to the allowed paths. */
+    state.allowAndSetStorePathString(storePath, v);
+}
+
+PrimOp primop_toFile({
+    .name = "__toFile",
+    .args = {"name", "s"},
+    .doc = R"(
+      Store the string *s* in a file in the Nix store and return its
+      path.  The file has suffix *name*. This file can be used as an
+      input to derivations. One application is to write builders
+      “inline”. For instance, the following Nix expression combines the
+      Nix expression for GNU Hello and its build script into one file:
+
+      ```nix
+      { stdenv, fetchurl, perl }:
+
+      stdenv.mkDerivation {
+        name = "hello-2.1.1";
+
+        builder = builtins.toFile "builder.sh" "
+          source $stdenv/setup
+
+          PATH=$perl/bin:$PATH
+
+          tar xvfz $src
+          cd hello-*
+          ./configure --prefix=$out
+          make
+          make install
+        ";
+
+        src = fetchurl {
+          url = "http://ftp.nluug.nl/pub/gnu/hello/hello-2.1.1.tar.gz";
+          sha256 = "1md7jsfd8pa45z73bz1kszpp01yw6x5ljkjk2hx7wl800any6465";
+        };
+        inherit perl;
+      }
+      ```
+
+      It is even possible for one file to refer to another, e.g.,
+
+      ```nix
+      builder = let
+        configFile = builtins.toFile "foo.conf" "
+          # This is some dummy configuration file.
+          ...
+        ";
+      in builtins.toFile "builder.sh" "
+        source $stdenv/setup
+        ...
+        cp ${configFile} $out/etc/foo.conf
+      ";
+      ```
+
+      Note that `${configFile}` is a
+      [string interpolation](@docroot@/language/values.md#type-string), so the result of the
+      expression `configFile`
+      (i.e., a path like `/nix/store/m7p7jfny445k...-foo.conf`) will be
+      spliced into the resulting string.
+
+      It is however *not* allowed to have files mutually referring to each
+      other, like so:
+
+      ```nix
+      let
+        foo = builtins.toFile "foo" "...${bar}...";
+        bar = builtins.toFile "bar" "...${foo}...";
+      in foo
+      ```
+
+      This is not allowed because it would cause a cyclic dependency in
+      the computation of the cryptographic hashes for `foo` and `bar`.
+
+      It is also not possible to reference the result of a derivation. If
+      you are using Nixpkgs, the `writeTextFile` function is able to do
+      that.
+    )",
+    .fun = prim_toFile,
+});
+
+}

src/libexpr/primops/string.cc

@@ -0,0 +1,641 @@
+#include <assert.h>          // for assert
+#include <stddef.h>          // for size_t
+#include <iterator>          // for distance
+#include <memory>            // for shared_ptr, __shared_ptr_access, make_sh...
+#include <regex>             // for regex_error, match_results, cregex_iterator
+#include <string>            // for basic_string, char_traits, string
+#include <string_view>       // for basic_string_view
+#include <unordered_map>     // for unordered_map, operator==, _Node_iterator
+#include <utility>           // for move, pair
+#include <vector>            // for vector
+#include "attr-set.hh"       // for BindingsBuilder
+#include "eval-error.hh"     // for EvalError, EvalErrorBuilder
+#include "eval.hh"           // for EvalState, PrimOp, makeRegexCache, prim_...
+#include "names.hh"          // for DrvName, compareVersions, nextComponent
+#include "pos-idx.hh"        // for PosIdx
+#include "primops.hh"        // for RegexCache
+#include "symbol-table.hh"   // for Symbol
+#include "types.hh"          // for BackedStringView, Strings
+#include "util.hh"           // for baseNameOf, enumerate
+#include "value.hh"          // for Value, nString
+#include "value/context.hh"  // for NixStringContext
+
+namespace nix {
+
+std::shared_ptr<RegexCache> makeRegexCache()
+{
+    return std::make_shared<RegexCache>();
+}
+
+/**
+ * builtins.baseNameOf
+ */
+
+static void prim_baseNameOf(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    v.mkString(
+        baseNameOf(*state.coerceToString(
+            pos,
+            *args[0],
+            context,
+            "while evaluating the first argument passed to builtins.baseNameOf",
+            false,
+            false
+        )),
+        context
+    );
+}
+
+PrimOp primop_baseNameOf({
+    .name = "baseNameOf",
+    .args = {"s"},
+    .doc = R"(
+      Return the *base name* of the string *s*, that is, everything
+      following the final slash in the string. This is similar to the GNU
+      `basename` command.
+    )",
+    .fun = prim_baseNameOf,
+});
+
+/**
+ * builtins.compareVersions
+ */
+
+static void prim_compareVersions(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto version1 = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.compareVersions"
+    );
+    auto version2 = state.forceStringNoCtx(
+        *args[1], pos, "while evaluating the second argument passed to builtins.compareVersions"
+    );
+    v.mkInt(compareVersions(version1, version2));
+}
+
+PrimOp primop_compareVersions({
+    .name = "__compareVersions",
+    .args = {"s1", "s2"},
+    .doc = R"(
+      Compare two strings representing versions and return `-1` if
+      version *s1* is older than version *s2*, `0` if they are the same,
+      and `1` if *s1* is newer than *s2*. The version comparison
+      algorithm is the same as the one used by [`nix-env
+      -u`](../command-ref/nix-env.md#operation---upgrade).
+    )",
+    .fun = prim_compareVersions,
+});
+
+/**
+ * builtins.concatStringsSep
+ */
+
+static void prim_concatStringsSep(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+
+    auto sep = state.forceString(
+        *args[0],
+        context,
+        pos,
+        "while evaluating the first argument (the separator string) passed to "
+        "builtins.concatStringsSep"
+    );
+    state.forceList(
+        *args[1],
+        pos,
+        "while evaluating the second argument (the list of strings to concat) passed to "
+        "builtins.concatStringsSep"
+    );
+
+    std::string res;
+    res.reserve((args[1]->listSize() + 32) * sep.size());
+    bool first = true;
+
+    for (auto elem : args[1]->listItems()) {
+        if (first) {
+            first = false;
+        } else {
+            res += sep;
+        }
+        res += *state.coerceToString(
+            pos,
+            *elem,
+            context,
+            "while evaluating one element of the list of strings to concat passed to "
+            "builtins.concatStringsSep"
+        );
+    }
+
+    v.mkString(res, context);
+}
+
+PrimOp primop_concatStringsSep(
+    {.name = "__concatStringsSep",
+     .args = {"separator", "list"},
+     .doc = R"(
+      Concatenate a list of strings with a separator between each
+      element, e.g. `concatStringsSep "/" ["usr" "local" "bin"] ==
+      "usr/local/bin"`.
+    )",
+     .fun = prim_concatStringsSep}
+);
+
+/**
+ * builtins.match
+ */
+
+void prim_match(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto re = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.match"
+    );
+
+    try {
+
+        auto regex = state.regexCache->get(re);
+
+        NixStringContext context;
+        const auto str = state.forceString(
+            *args[1], context, pos, "while evaluating the second argument passed to builtins.match"
+        );
+
+        std::cmatch match;
+        if (!std::regex_match(str.begin(), str.end(), match, regex)) {
+            v.mkNull();
+            return;
+        }
+
+        // the first match is the whole string
+        const size_t len = match.size() - 1;
+        state.mkList(v, len);
+        for (size_t i = 0; i < len; ++i) {
+            if (!match[i + 1].matched) {
+                (v.listElems()[i] = state.allocValue())->mkNull();
+            } else {
+                (v.listElems()[i] = state.allocValue())->mkString(match[i + 1].str());
+            }
+        }
+
+    } catch (std::regex_error & e) {
+        if (e.code() == std::regex_constants::error_space) {
+            // limit is _GLIBCXX_REGEX_STATE_LIMIT for libstdc++
+            state.error<EvalError>("memory limit exceeded by regular expression '%s'", re)
+                .atPos(pos)
+                .debugThrow();
+        } else {
+            state.error<EvalError>("invalid regular expression '%s'", re).atPos(pos).debugThrow();
+        }
+    }
+}
+
+PrimOp primop_match({
+    .name = "__match",
+    .args = {"regex", "str"},
+    .doc = R"s(
+      Returns a list if the [extended POSIX regular
+      expression](http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap09.html#tag_09_04)
+      *regex* matches *str* precisely, otherwise returns `null`. Each item
+      in the list is a regex group.
+
+      ```nix
+      builtins.match "ab" "abc"
+      ```
+
+      Evaluates to `null`.
+
+      ```nix
+      builtins.match "abc" "abc"
+      ```
+
+      Evaluates to `[ ]`.
+
+      ```nix
+      builtins.match "a(b)(c)" "abc"
+      ```
+
+      Evaluates to `[ "b" "c" ]`.
+
+      ```nix
+      builtins.match "[[:space:]]+([[:upper:]]+)[[:space:]]+" "  FOO   "
+      ```
+
+      Evaluates to `[ "FOO" ]`.
+    )s",
+    .fun = prim_match,
+});
+
+/**
+ * builtins.parseDrvName
+ */
+
+static void prim_parseDrvName(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto name = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.parseDrvName"
+    );
+    DrvName parsed(name);
+    auto attrs = state.buildBindings(2);
+    attrs.alloc(state.sName).mkString(parsed.name);
+    attrs.alloc("version").mkString(parsed.version);
+    v.mkAttrs(attrs);
+}
+
+PrimOp primop_parseDrvName({
+    .name = "__parseDrvName",
+    .args = {"s"},
+    .doc = R"(
+      Split the string *s* into a package name and version. The package
+      name is everything up to but not including the first dash not followed
+      by a letter, and the version is everything following that dash. The
+      result is returned in a set `{ name, version }`. Thus,
+      `builtins.parseDrvName "nix-0.12pre12876"` returns `{ name =
+      "nix"; version = "0.12pre12876"; }`.
+    )",
+    .fun = prim_parseDrvName,
+});
+
+/**
+ * builtins.replaceStrings
+ */
+
+static void prim_replaceStrings(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(
+        *args[0], pos, "while evaluating the first argument passed to builtins.replaceStrings"
+    );
+    state.forceList(
+        *args[1], pos, "while evaluating the second argument passed to builtins.replaceStrings"
+    );
+    if (args[0]->listSize() != args[1]->listSize()) {
+        state
+            .error<EvalError>(
+                "'from' and 'to' arguments passed to builtins.replaceStrings have different lengths"
+            )
+            .atPos(pos)
+            .debugThrow();
+    }
+
+    std::vector<std::string> from;
+    from.reserve(args[0]->listSize());
+    for (auto elem : args[0]->listItems()) {
+        from.emplace_back(state.forceString(
+            *elem,
+            pos,
+            "while evaluating one of the strings to replace passed to builtins.replaceStrings"
+        ));
+    }
+
+    std::unordered_map<size_t, std::string> cache;
+    auto to = args[1]->listItems();
+
+    NixStringContext context;
+    auto s = state.forceString(
+        *args[2],
+        context,
+        pos,
+        "while evaluating the third argument passed to builtins.replaceStrings"
+    );
+
+    std::string res;
+    // Loops one past last character to handle the case where 'from' contains an empty string.
+    for (size_t p = 0; p <= s.size();) {
+        bool found = false;
+        auto i = from.begin();
+        auto j = to.begin();
+        size_t j_index = 0;
+        for (; i != from.end(); ++i, ++j, ++j_index) {
+            if (s.compare(p, i->size(), *i) == 0) {
+                found = true;
+                auto v = cache.find(j_index);
+                if (v == cache.end()) {
+                    NixStringContext ctx;
+                    auto ts = state.forceString(
+                        **j,
+                        ctx,
+                        pos,
+                        "while evaluating one of the replacement strings passed to "
+                        "builtins.replaceStrings"
+                    );
+                    v = (cache.emplace(j_index, ts)).first;
+                    for (auto & path : ctx) {
+                        context.insert(path);
+                    }
+                }
+                res += v->second;
+                if (i->empty()) {
+                    if (p < s.size()) {
+                        res += s[p];
+                    }
+                    p++;
+                } else {
+                    p += i->size();
+                }
+                break;
+            }
+        }
+        if (!found) {
+            if (p < s.size()) {
+                res += s[p];
+            }
+            p++;
+        }
+    }
+
+    v.mkString(res, context);
+}
+
+PrimOp primop_replaceStrings({
+    .name = "__replaceStrings",
+    .args = {"from", "to", "s"},
+    .doc = R"(
+      Given string *s*, replace every occurrence of the strings in *from*
+      with the corresponding string in *to*.
+
+      The argument *to* is lazy, that is, it is only evaluated when its corresponding pattern in *from* is matched in the string *s*
+
+      Example:
+
+      ```nix
+      builtins.replaceStrings ["oo" "a"] ["a" "i"] "foobar"
+      ```
+
+      evaluates to `"fabir"`.
+    )",
+    .fun = prim_replaceStrings,
+});
+
+/**
+ * builtins.split
+ */
+
+void prim_split(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto re = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.split"
+    );
+
+    try {
+
+        auto regex = state.regexCache->get(re);
+
+        NixStringContext context;
+        const auto str = state.forceString(
+            *args[1], context, pos, "while evaluating the second argument passed to builtins.split"
+        );
+
+        auto begin = std::cregex_iterator(str.begin(), str.end(), regex);
+        auto end = std::cregex_iterator();
+
+        // Any matches results are surrounded by non-matching results.
+        const size_t len = std::distance(begin, end);
+        state.mkList(v, 2 * len + 1);
+        size_t idx = 0;
+
+        if (len == 0) {
+            v.listElems()[idx++] = args[1];
+            return;
+        }
+
+        for (auto i = begin; i != end; ++i) {
+            assert(idx <= 2 * len + 1 - 3);
+            auto match = *i;
+
+            // Add a string for non-matched characters.
+            (v.listElems()[idx++] = state.allocValue())->mkString(match.prefix().str());
+
+            // Add a list for matched substrings.
+            const size_t slen = match.size() - 1;
+            auto elem = v.listElems()[idx++] = state.allocValue();
+
+            // Start at 1, beacause the first match is the whole string.
+            state.mkList(*elem, slen);
+            for (size_t si = 0; si < slen; ++si) {
+                if (!match[si + 1].matched) {
+                    (elem->listElems()[si] = state.allocValue())->mkNull();
+                } else {
+                    (elem->listElems()[si] = state.allocValue())->mkString(match[si + 1].str());
+                }
+            }
+
+            // Add a string for non-matched suffix characters.
+            if (idx == 2 * len) {
+                (v.listElems()[idx++] = state.allocValue())->mkString(match.suffix().str());
+            }
+        }
+
+        assert(idx == 2 * len + 1);
+
+    } catch (std::regex_error & e) {
+        if (e.code() == std::regex_constants::error_space) {
+            // limit is _GLIBCXX_REGEX_STATE_LIMIT for libstdc++
+            state.error<EvalError>("memory limit exceeded by regular expression '%s'", re)
+                .atPos(pos)
+                .debugThrow();
+        } else {
+            state.error<EvalError>("invalid regular expression '%s'", re).atPos(pos).debugThrow();
+        }
+    }
+}
+
+PrimOp primop_split({
+    .name = "__split",
+    .args = {"regex", "str"},
+    .doc = R"s(
+      Returns a list composed of non matched strings interleaved with the
+      lists of the [extended POSIX regular
+      expression](http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap09.html#tag_09_04)
+      *regex* matches of *str*. Each item in the lists of matched
+      sequences is a regex group.
+
+      ```nix
+      builtins.split "(a)b" "abc"
+      ```
+
+      Evaluates to `[ "" [ "a" ] "c" ]`.
+
+      ```nix
+      builtins.split "([ac])" "abc"
+      ```
+
+      Evaluates to `[ "" [ "a" ] "b" [ "c" ] "" ]`.
+
+      ```nix
+      builtins.split "(a)|(c)" "abc"
+      ```
+
+      Evaluates to `[ "" [ "a" null ] "b" [ null "c" ] "" ]`.
+
+      ```nix
+      builtins.split "([[:upper:]]+)" " FOO "
+      ```
+
+      Evaluates to `[ " " [ "FOO" ] " " ]`.
+    )s",
+    .fun = prim_split,
+});
+
+/**
+ * builtins.splitVersion
+ */
+
+static void prim_splitVersion(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    auto version = state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.splitVersion"
+    );
+    auto iter = version.cbegin();
+    Strings components;
+    while (iter != version.cend()) {
+        auto component = nextComponent(iter, version.cend());
+        if (component.empty()) {
+            break;
+        }
+        components.emplace_back(component);
+    }
+    state.mkList(v, components.size());
+    for (const auto & [n, component] : enumerate(components)) {
+        (v.listElems()[n] = state.allocValue())->mkString(std::move(component));
+    }
+}
+
+PrimOp primop_splitVersion({
+    .name = "__splitVersion",
+    .args = {"s"},
+    .doc = R"(
+      Split a string representing a version into its components, by the
+      same version splitting logic underlying the version comparison in
+      [`nix-env -u`](../command-ref/nix-env.md#operation---upgrade).
+    )",
+    .fun = prim_splitVersion,
+});
+
+/**
+ * builtins.stringLength
+ */
+
+static void prim_stringLength(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto s = state.coerceToString(
+        pos, *args[0], context, "while evaluating the argument passed to builtins.stringLength"
+    );
+    v.mkInt(s->size());
+}
+
+PrimOp primop_stringLength({
+    .name = "__stringLength",
+    .args = {"e"},
+    .doc = R"(
+      Return the length of the string *e*. If *e* is not a string,
+      evaluation is aborted.
+    )",
+    .fun = prim_stringLength,
+});
+
+/**
+ * builtins.substring
+ */
+
+static void prim_substring(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    int start = state.forceInt(
+        *args[0],
+        pos,
+        "while evaluating the first argument (the start offset) passed to builtins.substring"
+    );
+
+    if (start < 0) {
+        state.error<EvalError>("negative start position in 'substring'").atPos(pos).debugThrow();
+    }
+
+    int len = state.forceInt(
+        *args[1],
+        pos,
+        "while evaluating the second argument (the substring length) passed to builtins.substring"
+    );
+
+    // Special-case on empty substring to avoid O(n) strlen
+    // This allows for the use of empty substrings to efficently capture string context
+    if (len == 0) {
+        state.forceValue(*args[2], pos);
+        if (args[2]->type() == nString) {
+            v.mkString("", args[2]->string.context);
+            return;
+        }
+    }
+
+    NixStringContext context;
+    auto s = state.coerceToString(
+        pos,
+        *args[2],
+        context,
+        "while evaluating the third argument (the string) passed to builtins.substring"
+    );
+
+    v.mkString((unsigned int) start >= s->size() ? "" : s->substr(start, len), context);
+}
+
+PrimOp primop_substring({
+    .name = "__substring",
+    .args = {"start", "len", "s"},
+    .doc = R"(
+      Return the substring of *s* from character position *start*
+      (zero-based) up to but not including *start + len*. If *start* is
+      greater than the length of the string, an empty string is returned,
+      and if *start + len* lies beyond the end of the string, only the
+      substring up to the end of the string is returned. *start* must be
+      non-negative. For example,
+
+      ```nix
+      builtins.substring 0 3 "nixos"
+      ```
+
+      evaluates to `"nix"`.
+    )",
+    .fun = prim_substring,
+});
+
+/**
+ * builtins.toString
+ */
+
+static void prim_toString(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    NixStringContext context;
+    auto s = state.coerceToString(
+        pos,
+        *args[0],
+        context,
+        "while evaluating the first argument passed to builtins.toString",
+        true,
+        false
+    );
+    v.mkString(*s, context);
+}
+
+PrimOp primop_toString({
+    .name = "toString",
+    .args = {"e"},
+    .doc = R"(
+      Convert the expression *e* to a string. *e* can be:
+
+        - A string (in which case the string is returned unmodified).
+
+        - A path (e.g., `toString /foo/bar` yields `"/foo/bar"`.
+
+        - A set containing `{ __toString = self: ...; }` or `{ outPath = ...; }`.
+
+        - An integer.
+
+        - A list, in which case the string representations of its elements
+          are joined with spaces.
+
+        - A Boolean (`false` yields `""`, `true` yields `"1"`).
+
+        - `null`, which yields the empty string.
+    )",
+    .fun = prim_toString,
+});
+
+}

src/libexpr/primops/system.cc

@@ -0,0 +1,116 @@
+#include <optional>                                // for optional
+#include <string>                                  // for basic_string, char...
+#include <utility>                                 // for move
+#include "canon-path.hh"                           // for CanonPath
+#include "error.hh"                                // for Error
+#include "eval-error.hh"                           // for EvalError, EvalErr...
+#include "eval-settings.hh"                        // for EvalSettings, eval...
+#include "eval.hh"                                 // for EvalState, PrimOp
+#include "pos-idx.hh"                              // for PosIdx
+#include "pos-table.hh"                            // for PosTable
+#include "primops.hh"                              // for prim_exec
+#include "types.hh"                                // for BackedStringView
+#include "util.hh"                                 // for getEnv, runProgram
+#include "value.hh"                                // for Value
+#include "value/context.hh"                        // for NixStringContext
+
+namespace nix {
+
+struct Expr;
+
+/**
+ * builtins.exec
+ */
+
+void prim_exec(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceList(*args[0], pos, "while evaluating the first argument passed to builtins.exec");
+    auto elems = args[0]->listElems();
+    auto count = args[0]->listSize();
+    if (count == 0) {
+        state.error<EvalError>("at least one argument to 'exec' required").atPos(pos).debugThrow();
+    }
+    NixStringContext context;
+    auto program =
+        state
+            .coerceToString(
+                pos,
+                *elems[0],
+                context,
+                "while evaluating the first element of the argument passed to builtins.exec",
+                false,
+                false
+            )
+            .toOwned();
+    Strings commandArgs;
+    for (unsigned int i = 1; i < args[0]->listSize(); ++i) {
+        commandArgs.push_back(
+            state
+                .coerceToString(
+                    pos,
+                    *elems[i],
+                    context,
+                    "while evaluating an element of the argument passed to builtins.exec",
+                    false,
+                    false
+                )
+                .toOwned()
+        );
+    }
+    try {
+        auto _ = state.realiseContext(context); // FIXME: Handle CA derivations
+    } catch (InvalidPathError & e) {
+        state
+            .error<EvalError>(
+                "cannot execute '%1%', since path '%2%' is not valid", program, e.path
+            )
+            .atPos(pos)
+            .debugThrow();
+    }
+
+    auto output = runProgram(program, true, commandArgs);
+    Expr * parsed;
+    try {
+        parsed = state.parseExprFromString(std::move(output), state.rootPath(CanonPath::root));
+    } catch (Error & e) {
+        e.addTrace(state.positions[pos], "while parsing the output from '%1%'", program);
+        throw;
+    }
+    try {
+        state.eval(parsed, v);
+    } catch (Error & e) {
+        e.addTrace(state.positions[pos], "while evaluating the output from '%1%'", program);
+        throw;
+    }
+}
+
+/**
+ * builtins.getEnv
+ */
+
+static void prim_getEnv(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    std::string name(state.forceStringNoCtx(
+        *args[0], pos, "while evaluating the first argument passed to builtins.getEnv"
+    ));
+    v.mkString(evalSettings.restrictEval || evalSettings.pureEval ? "" : getEnv(name).value_or(""));
+}
+
+PrimOp primop_getEnv({
+    .name = "__getEnv",
+    .args = {"s"},
+    .doc = R"(
+      `getEnv` returns the value of the environment variable *s*, or an
+      empty string if the variable doesn’t exist. This function should be
+      used with care, as it can introduce all sorts of nasty environment
+      dependencies in your Nix expression.
+
+      `getEnv` is used in Nix Packages to locate the file
+      `~/.nixpkgs/config.nix`, which contains user-local settings for Nix
+      Packages. (That is, it does a `getEnv "HOME"` to locate the user’s
+      home directory.)
+    )",
+    .fun = prim_getEnv,
+});
+
+}

src/libexpr/primops/toXML.cc

@@ -0,0 +1,119 @@
+#include <sstream>           // for basic_ostringstream, basic_ios, basic_os...
+#include "eval.hh"           // for PrimOp, EvalState (ptr only)
+#include "pos-idx.hh"        // for PosIdx
+#include "value.hh"          // for Value
+#include "value-to-xml.hh"   // for printValueAsXML
+#include "value/context.hh"  // for NixStringContext
+
+namespace nix {
+
+/**
+ * builtins.toXML
+ */
+
+static void prim_toXML(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    std::ostringstream out;
+    NixStringContext context;
+    printValueAsXML(state, true, false, *args[0], out, context, pos);
+    v.mkString(out.str(), context);
+}
+
+PrimOp primop_toXML({
+    .name = "__toXML",
+    .args = {"e"},
+    .doc = R"(
+      Return a string containing an XML representation of *e*. The main
+      application for `toXML` is to communicate information with the
+      builder in a more structured format than plain environment
+      variables.
+
+      Here is an example where this is the case:
+
+      ```nix
+      { stdenv, fetchurl, libxslt, jira, uberwiki }:
+
+      stdenv.mkDerivation (rec {
+        name = "web-server";
+
+        buildInputs = [ libxslt ];
+
+        builder = builtins.toFile "builder.sh" "
+          source $stdenv/setup
+          mkdir $out
+          echo "$servlets" | xsltproc ${stylesheet} - > $out/server-conf.xml ①
+        ";
+
+        stylesheet = builtins.toFile "stylesheet.xsl" ②
+         "<?xml version='1.0' encoding='UTF-8'?>
+          <xsl:stylesheet xmlns:xsl='http://www.w3.org/1999/XSL/Transform' version='1.0'>
+            <xsl:template match='/'>
+              <Configure>
+                <xsl:for-each select='/expr/list/attrs'>
+                  <Call name='addWebApplication'>
+                    <Arg><xsl:value-of select=\"attr[@name = 'path']/string/@value\" /></Arg>
+                    <Arg><xsl:value-of select=\"attr[@name = 'war']/path/@value\" /></Arg>
+                  </Call>
+                </xsl:for-each>
+              </Configure>
+            </xsl:template>
+          </xsl:stylesheet>
+        ";
+
+        servlets = builtins.toXML [ ③
+          { path = "/bugtracker"; war = jira + "/lib/atlassian-jira.war"; }
+          { path = "/wiki"; war = uberwiki + "/uberwiki.war"; }
+        ];
+      })
+      ```
+
+      The builder is supposed to generate the configuration file for a
+      [Jetty servlet container](http://jetty.mortbay.org/). A servlet
+      container contains a number of servlets (`*.war` files) each
+      exported under a specific URI prefix. So the servlet configuration
+      is a list of sets containing the `path` and `war` of the servlet
+      (①). This kind of information is difficult to communicate with the
+      normal method of passing information through an environment
+      variable, which just concatenates everything together into a
+      string (which might just work in this case, but wouldn’t work if
+      fields are optional or contain lists themselves). Instead the Nix
+      expression is converted to an XML representation with `toXML`,
+      which is unambiguous and can easily be processed with the
+      appropriate tools. For instance, in the example an XSLT stylesheet
+      (at point ②) is applied to it (at point ①) to generate the XML
+      configuration file for the Jetty server. The XML representation
+      produced at point ③ by `toXML` is as follows:
+
+      ```xml
+      <?xml version='1.0' encoding='utf-8'?>
+      <expr>
+        <list>
+          <attrs>
+            <attr name="path">
+              <string value="/bugtracker" />
+            </attr>
+            <attr name="war">
+              <path value="/nix/store/d1jh9pasa7k2...-jira/lib/atlassian-jira.war" />
+            </attr>
+          </attrs>
+          <attrs>
+            <attr name="path">
+              <string value="/wiki" />
+            </attr>
+            <attr name="war">
+              <path value="/nix/store/y6423b1yi4sx...-uberwiki/uberwiki.war" />
+            </attr>
+          </attrs>
+        </list>
+      </expr>
+      ```
+
+      Note that we used the `toFile` built-in to write the builder and
+      the stylesheet “inline” in the Nix expression. The path of the
+      stylesheet is spliced into the builder using the syntax `xsltproc
+      ${stylesheet}`.
+    )",
+    .fun = prim_toXML,
+});
+
+}

src/libexpr/primops/types.cc

@@ -0,0 +1,245 @@
+#include <stdlib.h>         // for abort
+#include <string>           // for basic_string, char_traits, string
+#include "attr-set.hh"      // for BindingsBuilder
+#include "eval-error.hh"    // for TypeError, EvalErrorBuilder
+#include "eval.hh"          // for PrimOp, EvalState
+#include "nixexpr.hh"       // for ExprLambda, Formal, Formals
+#include "pos-idx.hh"       // for PosIdx
+#include "symbol-table.hh"  // for Symbol
+#include "value.hh"         // for Value, nAttrs, nBool, nFloat, nFunction
+
+namespace nix {
+
+/**
+ * Generic isType function
+ */
+
+static inline auto prim_isType(auto n)
+{
+    return [n](EvalState & state, const PosIdx pos, Value ** args, Value & v) {
+        state.forceValue(*args[0], pos);
+        v.mkBool(args[0]->type() == n);
+    };
+}
+
+/**
+ * builtins.functionArgs
+ */
+
+static void prim_functionArgs(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    if (args[0]->isPrimOpApp() || args[0]->isPrimOp()) {
+        v.mkAttrs(&state.emptyBindings);
+        return;
+    }
+    if (!args[0]->isLambda()) {
+        state.error<TypeError>("'functionArgs' requires a function").atPos(pos).debugThrow();
+    }
+
+    if (!args[0]->lambda.fun->hasFormals()) {
+        v.mkAttrs(&state.emptyBindings);
+        return;
+    }
+
+    auto attrs = state.buildBindings(args[0]->lambda.fun->formals->formals.size());
+    for (auto & i : args[0]->lambda.fun->formals->formals) {
+        // !!! should optimise booleans (allocate only once)
+        attrs.alloc(i.name, i.pos).mkBool(i.def);
+    }
+    v.mkAttrs(attrs);
+}
+
+PrimOp primop_functionArgs({
+    .name = "__functionArgs",
+    .args = {"f"},
+    .doc = R"(
+      Return a set containing the names of the formal arguments expected
+      by the function *f*. The value of each attribute is a Boolean
+      denoting whether the corresponding argument has a default value. For
+      instance, `functionArgs ({ x, y ? 123}: ...) = { x = false; y =
+      true; }`.
+
+      "Formal argument" here refers to the attributes pattern-matched by
+      the function. Plain lambdas are not included, e.g. `functionArgs (x:
+      ...) = { }`.
+    )",
+    .fun = prim_functionArgs,
+});
+
+/**
+ * builtins.isAttrs
+ */
+
+PrimOp primop_isAttrs({
+    .name = "__isAttrs",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to a set, and `false` otherwise.
+    )",
+    .fun = prim_isType(nAttrs),
+});
+
+/**
+ * builtins.isBool
+ */
+
+PrimOp primop_isBool({
+    .name = "__isBool",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to a bool, and `false` otherwise.
+    )",
+    .fun = prim_isType(nBool),
+});
+
+/**
+ * builtins.Float
+ */
+
+PrimOp primop_isFloat({
+    .name = "__isFloat",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to a float, and `false` otherwise.
+    )",
+    .fun = prim_isType(nFloat),
+});
+
+/**
+ * builtins.isFunction
+ */
+
+PrimOp primop_isFunction({
+    .name = "__isFunction",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to a function, and `false` otherwise.
+    )",
+    .fun = prim_isType(nFunction),
+});
+
+/**
+ * builtins.isInt
+ */
+
+PrimOp primop_isInt({
+    .name = "__isInt",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to an integer, and `false` otherwise.
+    )",
+    .fun = prim_isType(nInt),
+});
+
+/**
+ * builtins.isList
+ */
+
+PrimOp primop_isList({
+    .name = "__isList",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to a list, and `false` otherwise.
+    )",
+    .fun = prim_isType(nList),
+});
+
+/**
+ * builtins.isNull
+ */
+
+PrimOp primop_isNull({
+    .name = "isNull",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to `null`, and `false` otherwise.
+
+      This is equivalent to `e == null`.
+    )",
+    .fun = prim_isType(nNull),
+});
+
+/**
+ * builtins.isPath
+ */
+
+PrimOp primop_isPath({
+    .name = "__isPath",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to a path, and `false` otherwise.
+    )",
+    .fun = prim_isType(nPath),
+});
+
+/**
+ * builtins.isString
+ */
+
+PrimOp primop_isString({
+    .name = "__isString",
+    .args = {"e"},
+    .doc = R"(
+      Return `true` if *e* evaluates to a string, and `false` otherwise.
+    )",
+    .fun = prim_isType(nString),
+});
+
+/**
+ * builtins.typeOf
+ */
+
+static void prim_typeOf(EvalState & state, const PosIdx pos, Value ** args, Value & v)
+{
+    state.forceValue(*args[0], pos);
+    std::string t;
+    switch (args[0]->type()) {
+    case nInt:
+        t = "int";
+        break;
+    case nBool:
+        t = "bool";
+        break;
+    case nString:
+        t = "string";
+        break;
+    case nPath:
+        t = "path";
+        break;
+    case nNull:
+        t = "null";
+        break;
+    case nAttrs:
+        t = "set";
+        break;
+    case nList:
+        t = "list";
+        break;
+    case nFunction:
+        t = "lambda";
+        break;
+    case nExternal:
+        t = args[0]->external->typeOf();
+        break;
+    case nFloat:
+        t = "float";
+        break;
+    case nThunk:
+        abort();
+    }
+    v.mkString(t);
+}
+
+PrimOp primop_typeOf({
+    .name = "__typeOf",
+    .args = {"e"},
+    .doc = R"(
+      Return a string representing the type of the value *e*, namely
+      `"int"`, `"bool"`, `"string"`, `"path"`, `"null"`, `"set"`,
+      `"list"`, `"lambda"` or `"float"`.
+    )",
+    .fun = prim_typeOf,
+});
+
+}