lix/src/libexpr/value.hh

352 lines
9.1 KiB
C++
Raw Normal View History

#pragma once
#include "symbol-table.hh"
2016-08-30 11:12:12 +00:00
#if HAVE_BOEHMGC
#include <gc/gc_allocator.h>
#endif
namespace nix {
typedef enum {
tInt = 1,
tBool,
tString,
tPath,
tNull,
tAttrs,
tList1,
tList2,
tListN,
tThunk,
tApp,
tLambda,
tBlackhole,
tPrimOp,
tPrimOpApp,
tExternal,
tFloat
} InternalType;
// This type abstracts over all actual value types in the language,
// grouping together implementation details like tList*, different function
// types, and types in non-normal form (so thunks and co.)
typedef enum {
nThunk,
nInt,
nFloat,
nBool,
nString,
nPath,
nNull,
nAttrs,
nList,
nFunction,
nExternal
} ValueType;
2014-01-21 17:29:55 +00:00
class Bindings;
struct Env;
struct Expr;
struct ExprLambda;
struct PrimOp;
2014-01-21 17:29:55 +00:00
class Symbol;
struct Pos;
class EvalState;
class XMLWriter;
class JSONPlaceholder;
libexpr: Use int64_t for NixInt Using a 64bit integer on 32bit systems will come with a bit of a performance overhead, but given that Nix doesn't use a lot of integers compared to other types, I think the overhead is negligible also considering that 32bit systems are in decline. The biggest advantage however is that when we use a consistent integer size across all platforms it's less likely that we miss things that we break due to that. One example would be: https://github.com/NixOS/nixpkgs/pull/44233 On Hydra it will evaluate, because the evaluator runs on a 64bit machine, but when evaluating the same on a 32bit machine it will fail, so using 64bit integers should make that consistent. While the change of the type in value.hh is rather easy to do, we have a few more options available for doing the conversion in the lexer: * Via an #ifdef on the architecture and using strtol() or strtoll() accordingly depending on which architecture we are. For the #ifdef we would need another AX_COMPILE_CHECK_SIZEOF in configure.ac. * Using istringstream, which would involve copying the value. * As we're already using boost, lexical_cast might be a good idea. Spoiler: I went for the latter, first of all because lexical_cast does have an overload for const char* and second of all, because it doesn't involve copying around the input string. Also, because istringstream seems to come with a bigger overhead than boost::lexical_cast: https://www.boost.org/doc/libs/release/doc/html/boost_lexical_cast/performance.html The first method (still using strtol/strtoll) also wasn't something I pursued further, because it is also locale-aware which I doubt is what we want, given that the regex for int is [0-9]+. Signed-off-by: aszlig <aszlig@nix.build> Fixes: #2339
2018-08-28 22:23:51 +00:00
typedef int64_t NixInt;
typedef double NixFloat;
/* External values must descend from ExternalValueBase, so that
* type-agnostic nix functions (e.g. showType) can be implemented
*/
class ExternalValueBase
{
friend std::ostream & operator << (std::ostream & str, const ExternalValueBase & v);
protected:
/* Print out the value */
virtual std::ostream & print(std::ostream & str) const = 0;
public:
/* Return a simple string describing the type */
virtual string showType() const = 0;
/* Return a string to be used in builtins.typeOf */
virtual string typeOf() const = 0;
/* Coerce the value to a string. Defaults to uncoercable, i.e. throws an
* error
*/
virtual string coerceToString(const Pos & pos, PathSet & context, bool copyMore, bool copyToStore) const;
/* Compare to another value of the same type. Defaults to uncomparable,
* i.e. always false.
*/
virtual bool operator==(const ExternalValueBase & b) const;
/* Print the value as JSON. Defaults to unconvertable, i.e. throws an error */
virtual void printValueAsJSON(EvalState & state, bool strict,
JSONPlaceholder & out, PathSet & context) const;
/* Print the value as XML. Defaults to unevaluated */
virtual void printValueAsXML(EvalState & state, bool strict, bool location,
XMLWriter & doc, PathSet & context, PathSet & drvsSeen) const;
virtual ~ExternalValueBase()
{
};
};
std::ostream & operator << (std::ostream & str, const ExternalValueBase & v);
struct Value
{
private:
InternalType internalType;
friend std::string showType(const Value & v);
friend void printValue(std::ostream & str, std::set<const Value *> & active, const Value & v);
public:
inline void setInt() { internalType = tInt; };
inline void setBool() { internalType = tBool; };
inline void setString() { internalType = tString; };
inline void setPath() { internalType = tPath; };
inline void setNull() { internalType = tNull; };
inline void setAttrs() { internalType = tAttrs; };
inline void setList1() { internalType = tList1; };
inline void setList2() { internalType = tList2; };
inline void setListN() { internalType = tListN; };
inline void setThunk() { internalType = tThunk; };
inline void setApp() { internalType = tApp; };
inline void setLambda() { internalType = tLambda; };
inline void setBlackhole() { internalType = tBlackhole; };
inline void setPrimOp() { internalType = tPrimOp; };
inline void setPrimOpApp() { internalType = tPrimOpApp; };
inline void setExternal() { internalType = tExternal; };
inline void setFloat() { internalType = tFloat; };
// Functions needed to distinguish the type
// These should be removed eventually, by putting the functionality that's
// needed by callers into methods of this type
// type() == nThunk
inline bool isThunk() const { return internalType == tThunk; };
inline bool isApp() const { return internalType == tApp; };
inline bool isBlackhole() const { return internalType == tBlackhole; };
// type() == nFunction
inline bool isLambda() const { return internalType == tLambda; };
inline bool isPrimOp() const { return internalType == tPrimOp; };
inline bool isPrimOpApp() const { return internalType == tPrimOpApp; };
union
{
NixInt integer;
bool boolean;
2013-08-06 12:15:11 +00:00
/* Strings in the evaluator carry a so-called `context' which
is a list of strings representing store paths. This is to
allow users to write things like
"--with-freetype2-library=" + freetype + "/lib"
where `freetype' is a derivation (or a source to be copied
to the store). If we just concatenated the strings without
keeping track of the referenced store paths, then if the
string is used as a derivation attribute, the derivation
will not have the correct dependencies in its inputDrvs and
inputSrcs.
The semantics of the context is as follows: when a string
with context C is used as a derivation attribute, then the
derivations in C will be added to the inputDrvs of the
derivation, and the other store paths in C will be added to
the inputSrcs of the derivations.
For canonicity, the store paths should be in sorted order. */
struct {
const char * s;
const char * * context; // must be in sorted order
} string;
const char * path;
Bindings * attrs;
struct {
2018-05-02 11:56:34 +00:00
size_t size;
Value * * elems;
} bigList;
Value * smallList[2];
struct {
Env * env;
Expr * expr;
} thunk;
struct {
Value * left, * right;
} app;
struct {
Env * env;
ExprLambda * fun;
} lambda;
PrimOp * primOp;
struct {
Value * left, * right;
} primOpApp;
ExternalValueBase * external;
NixFloat fpoint;
};
// Returns the normal type of a Value. This only returns nThunk if the
// Value hasn't been forceValue'd
inline ValueType type() const
{
switch (internalType) {
case tInt: return nInt;
case tBool: return nBool;
case tString: return nString;
case tPath: return nPath;
case tNull: return nNull;
case tAttrs: return nAttrs;
case tList1: case tList2: case tListN: return nList;
case tLambda: case tPrimOp: case tPrimOpApp: return nFunction;
case tExternal: return nExternal;
case tFloat: return nFloat;
case tThunk: case tApp: case tBlackhole: return nThunk;
}
abort();
}
bool isList() const
{
return internalType == tList1 || internalType == tList2 || internalType == tListN;
}
Value * * listElems()
{
return internalType == tList1 || internalType == tList2 ? smallList : bigList.elems;
}
const Value * const * listElems() const
{
return internalType == tList1 || internalType == tList2 ? smallList : bigList.elems;
}
2018-05-02 11:56:34 +00:00
size_t listSize() const
{
return internalType == tList1 ? 1 : internalType == tList2 ? 2 : bigList.size;
}
/* Check whether forcing this value requires a trivial amount of
computation. In particular, function applications are
non-trivial. */
bool isTrivial() const;
2020-06-29 17:08:37 +00:00
std::vector<std::pair<Path, std::string>> getContext();
};
/* After overwriting an app node, be sure to clear pointers in the
Value to ensure that the target isn't kept alive unnecessarily. */
static inline void clearValue(Value & v)
{
v.app.left = v.app.right = 0;
}
static inline void mkInt(Value & v, NixInt n)
{
clearValue(v);
v.setInt();
v.integer = n;
}
static inline void mkFloat(Value & v, NixFloat n)
{
clearValue(v);
v.setFloat();
v.fpoint = n;
}
static inline void mkBool(Value & v, bool b)
{
clearValue(v);
v.setBool();
v.boolean = b;
}
static inline void mkNull(Value & v)
{
clearValue(v);
v.setNull();
}
static inline void mkApp(Value & v, Value & left, Value & right)
{
v.setApp();
v.app.left = &left;
v.app.right = &right;
}
static inline void mkPrimOpApp(Value & v, Value & left, Value & right)
{
v.setPrimOpApp();
v.app.left = &left;
v.app.right = &right;
}
static inline void mkStringNoCopy(Value & v, const char * s)
{
v.setString();
v.string.s = s;
v.string.context = 0;
}
static inline void mkString(Value & v, const Symbol & s)
{
mkStringNoCopy(v, ((const string &) s).c_str());
}
void mkString(Value & v, const char * s);
static inline void mkPathNoCopy(Value & v, const char * s)
{
clearValue(v);
v.setPath();
v.path = s;
}
void mkPath(Value & v, const char * s);
#if HAVE_BOEHMGC
typedef std::vector<Value *, traceable_allocator<Value *> > ValueVector;
typedef std::map<Symbol, Value *, std::less<Symbol>, traceable_allocator<std::pair<const Symbol, Value *> > > ValueMap;
#else
typedef std::vector<Value *> ValueVector;
typedef std::map<Symbol, Value *> ValueMap;
#endif
/* A value allocated in traceable memory. */
typedef std::shared_ptr<Value *> RootValue;
RootValue allocRootValue(Value * v);
}