lix/src/libexpr/parser.y

658 lines
19 KiB
Plaintext
Raw Normal View History

%glr-parser
%pure-parser
%locations
%error-verbose
2006-10-02 14:43:15 +00:00
%defines
/* %no-lines */
2013-03-14 17:31:08 +00:00
%parse-param { void * scanner }
%parse-param { nix::ParseData * data }
%lex-param { void * scanner }
%lex-param { nix::ParseData * data }
2012-04-13 12:28:26 +00:00
%expect 1
%expect-rr 1
%code requires {
2013-09-02 14:29:15 +00:00
#ifndef BISON_HEADER
#define BISON_HEADER
2013-09-02 14:29:15 +00:00
#include "util.hh"
2013-09-02 14:29:15 +00:00
#include "nixexpr.hh"
#include "eval.hh"
namespace nix {
2013-09-02 14:29:15 +00:00
struct ParseData
{
EvalState & state;
SymbolTable & symbols;
Expr * result;
Path basePath;
Symbol path;
string error;
Symbol sLetBody;
ParseData(EvalState & state)
: state(state)
, symbols(state.symbols)
, sLetBody(symbols.create("<let-body>"))
{ };
};
2013-09-02 14:29:15 +00:00
}
#define YY_DECL int yylex \
(YYSTYPE * yylval_param, YYLTYPE * yylloc_param, yyscan_t yyscanner, nix::ParseData * data)
#endif
}
%{
#include "parser-tab.hh"
#include "lexer-tab.hh"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
YY_DECL;
using namespace nix;
namespace nix {
2013-09-02 14:29:15 +00:00
static void dupAttr(const AttrPath & attrPath, const Pos & pos, const Pos & prevPos)
{
throw ParseError(format("attribute `%1%' at %2% already defined at %3%")
% showAttrPath(attrPath) % pos % prevPos);
}
2013-09-02 14:29:15 +00:00
static void dupAttr(Symbol attr, const Pos & pos, const Pos & prevPos)
{
throw ParseError(format("attribute `%1%' at %2% already defined at %3%")
% attr % pos % prevPos);
}
2013-09-02 14:29:15 +00:00
static void addAttr(ExprAttrs * attrs, AttrPath & attrPath,
Expr * e, const Pos & pos)
{
AttrPath::iterator i;
// All attrpaths have at least one attr
assert(!attrPath.empty());
for (i = attrPath.begin(); i + 1 < attrPath.end(); i++) {
if (i->symbol.set()) {
ExprAttrs::AttrDefs::iterator j = attrs->attrs.find(i->symbol);
if (j != attrs->attrs.end()) {
if (!j->second.inherited) {
ExprAttrs * attrs2 = dynamic_cast<ExprAttrs *>(j->second.e);
if (!attrs2) dupAttr(attrPath, pos, j->second.pos);
attrs = attrs2;
} else
dupAttr(attrPath, pos, j->second.pos);
} else {
2010-04-12 23:33:23 +00:00
ExprAttrs * nested = new ExprAttrs;
attrs->attrs[i->symbol] = ExprAttrs::AttrDef(nested, pos);
2010-04-12 23:33:23 +00:00
attrs = nested;
}
} else {
ExprAttrs *nested = new ExprAttrs;
attrs->dynamicAttrs.push_back(ExprAttrs::DynamicAttrDef(i->expr, nested, pos));
attrs = nested;
}
}
if (i->symbol.set()) {
ExprAttrs::AttrDefs::iterator j = attrs->attrs.find(i->symbol);
if (j != attrs->attrs.end()) {
dupAttr(attrPath, pos, j->second.pos);
} else {
attrs->attrs[i->symbol] = ExprAttrs::AttrDef(e, pos);
e->setName(i->symbol);
}
} else {
attrs->dynamicAttrs.push_back(ExprAttrs::DynamicAttrDef(i->expr, e, pos));
}
}
static void addFormal(const Pos & pos, Formals * formals, const Formal & formal)
{
if (formals->argNames.find(formal.name) != formals->argNames.end())
throw ParseError(format("duplicate formal function argument `%1%' at %2%")
% formal.name % pos);
formals->formals.push_front(formal);
formals->argNames.insert(formal.name);
}
static Expr * stripIndentation(const Pos & pos, SymbolTable & symbols, vector<Expr *> & es)
{
if (es.empty()) return new ExprString(symbols.create(""));
2013-09-02 14:29:15 +00:00
/* Figure out the minimum indentation. Note that by design
whitespace-only final lines are not taken into account. (So
the " " in "\n ''" is ignored, but the " " in "\n foo''" is.) */
bool atStartOfLine = true; /* = seen only whitespace in the current line */
unsigned int minIndent = 1000000;
unsigned int curIndent = 0;
2010-04-12 22:03:27 +00:00
foreach (vector<Expr *>::iterator, i, es) {
ExprIndStr * e = dynamic_cast<ExprIndStr *>(*i);
if (!e) {
/* Anti-quotations end the current start-of-line whitespace. */
if (atStartOfLine) {
atStartOfLine = false;
if (curIndent < minIndent) minIndent = curIndent;
}
continue;
}
2010-04-12 22:03:27 +00:00
for (unsigned int j = 0; j < e->s.size(); ++j) {
if (atStartOfLine) {
2010-04-12 22:03:27 +00:00
if (e->s[j] == ' ')
curIndent++;
2010-04-12 22:03:27 +00:00
else if (e->s[j] == '\n') {
/* Empty line, doesn't influence minimum
indentation. */
curIndent = 0;
} else {
atStartOfLine = false;
if (curIndent < minIndent) minIndent = curIndent;
}
2010-04-12 22:03:27 +00:00
} else if (e->s[j] == '\n') {
atStartOfLine = true;
curIndent = 0;
}
}
}
/* Strip spaces from each line. */
2010-04-12 22:03:27 +00:00
vector<Expr *> * es2 = new vector<Expr *>;
atStartOfLine = true;
unsigned int curDropped = 0;
2010-04-12 22:03:27 +00:00
unsigned int n = es.size();
for (vector<Expr *>::iterator i = es.begin(); i != es.end(); ++i, --n) {
ExprIndStr * e = dynamic_cast<ExprIndStr *>(*i);
if (!e) {
atStartOfLine = false;
curDropped = 0;
2010-04-12 22:03:27 +00:00
es2->push_back(*i);
continue;
}
2013-09-02 14:29:15 +00:00
string s2;
2010-04-12 22:03:27 +00:00
for (unsigned int j = 0; j < e->s.size(); ++j) {
if (atStartOfLine) {
2010-04-12 22:03:27 +00:00
if (e->s[j] == ' ') {
if (curDropped++ >= minIndent)
2010-04-12 22:03:27 +00:00
s2 += e->s[j];
}
2010-04-12 22:03:27 +00:00
else if (e->s[j] == '\n') {
curDropped = 0;
2010-04-12 22:03:27 +00:00
s2 += e->s[j];
} else {
atStartOfLine = false;
curDropped = 0;
2010-04-12 22:03:27 +00:00
s2 += e->s[j];
}
} else {
2010-04-12 22:03:27 +00:00
s2 += e->s[j];
if (e->s[j] == '\n') atStartOfLine = true;
}
}
/* Remove the last line if it is empty and consists only of
spaces. */
if (n == 1) {
string::size_type p = s2.find_last_of('\n');
if (p != string::npos && s2.find_first_not_of(' ', p + 1) == string::npos)
s2 = string(s2, 0, p + 1);
}
2010-04-12 22:03:27 +00:00
es2->push_back(new ExprString(symbols.create(s2)));
}
/* If this is a single string, then don't do a concatenation. */
return es2->size() == 1 && dynamic_cast<ExprString *>((*es2)[0]) ? (*es2)[0] : new ExprConcatStrings(pos, true, es2);
}
void backToString(yyscan_t scanner);
void backToIndString(yyscan_t scanner);
static inline Pos makeCurPos(const YYLTYPE & loc, ParseData * data)
{
return Pos(data->path, loc.first_line, loc.first_column);
}
#define CUR_POS makeCurPos(*yylocp, data)
}
2006-10-02 14:43:15 +00:00
void yyerror(YYLTYPE * loc, yyscan_t scanner, ParseData * data, const char * error)
{
2010-04-12 21:21:24 +00:00
data->error = (format("%1%, at %2%")
% error % makeCurPos(*loc, data)).str();
}
%}
%union {
2013-09-02 14:29:15 +00:00
// !!! We're probably leaking stuff here.
nix::Expr * e;
nix::ExprList * list;
nix::ExprAttrs * attrs;
nix::Formals * formals;
nix::Formal * formal;
nix::NixInt n;
2014-01-21 17:29:55 +00:00
const char * id; // !!! -> Symbol
char * path;
char * uri;
std::vector<nix::AttrName> * attrNames;
2010-04-12 21:21:24 +00:00
std::vector<nix::Expr *> * string_parts;
}
%type <e> start expr expr_function expr_if expr_op
%type <e> expr_app expr_select expr_simple
%type <list> expr_list
%type <attrs> binds
%type <formals> formals
%type <formal> formal
%type <attrNames> attrs attrpath
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
%type <string_parts> string_parts_interpolated ind_string_parts
%type <e> string_parts string_attr
%type <id> attr
%token <id> ID ATTRPATH
2010-04-12 21:21:24 +00:00
%token <e> STR IND_STR
%token <n> INT
%token <path> PATH SPATH
%token <uri> URI
%token IF THEN ELSE ASSERT WITH LET IN REC INHERIT EQ NEQ AND OR IMPL OR_KW
%token DOLLAR_CURLY /* == ${ */
%token IND_STRING_OPEN IND_STRING_CLOSE
%token ELLIPSIS
%nonassoc IMPL
%left OR
%left AND
%nonassoc EQ NEQ
%left '<' '>' LEQ GEQ
%right UPDATE
%left NOT
%left '+' '-'
%left '*' '/'
%right CONCAT
%nonassoc '?'
%nonassoc '~'
%nonassoc NEGATE
%%
start: expr { data->result = $1; };
expr: expr_function;
expr_function
: ID ':' expr_function
{ $$ = new ExprLambda(CUR_POS, data->symbols.create($1), false, 0, $3); }
| '{' formals '}' ':' expr_function
{ $$ = new ExprLambda(CUR_POS, data->symbols.create(""), true, $2, $5); }
| '{' formals '}' '@' ID ':' expr_function
{ $$ = new ExprLambda(CUR_POS, data->symbols.create($5), true, $2, $7); }
| ID '@' '{' formals '}' ':' expr_function
{ $$ = new ExprLambda(CUR_POS, data->symbols.create($1), true, $4, $7); }
2010-04-12 21:21:24 +00:00
| ASSERT expr ';' expr_function
{ $$ = new ExprAssert(CUR_POS, $2, $4); }
| WITH expr ';' expr_function
{ $$ = new ExprWith(CUR_POS, $2, $4); }
| LET binds IN expr_function
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
{ if (!$2->dynamicAttrs.empty())
throw ParseError(format("dynamic attributes not allowed in let at %1%")
% CUR_POS);
$$ = new ExprLet($2, $4);
}
| expr_if
;
expr_if
: IF expr THEN expr ELSE expr { $$ = new ExprIf($2, $4, $6); }
| expr_op
;
expr_op
: '!' expr_op %prec NOT { $$ = new ExprOpNot($2); }
| '-' expr_op %prec NEGATE { $$ = new ExprApp(CUR_POS, new ExprApp(new ExprBuiltin(data->symbols.create("sub")), new ExprInt(0)), $2); }
2010-04-12 21:21:24 +00:00
| expr_op EQ expr_op { $$ = new ExprOpEq($1, $3); }
| expr_op NEQ expr_op { $$ = new ExprOpNEq($1, $3); }
| expr_op '<' expr_op { $$ = new ExprApp(CUR_POS, new ExprApp(new ExprBuiltin(data->symbols.create("lessThan")), $1), $3); }
| expr_op LEQ expr_op { $$ = new ExprOpNot(new ExprApp(CUR_POS, new ExprApp(new ExprBuiltin(data->symbols.create("lessThan")), $3), $1)); }
| expr_op '>' expr_op { $$ = new ExprApp(CUR_POS, new ExprApp(new ExprBuiltin(data->symbols.create("lessThan")), $3), $1); }
| expr_op GEQ expr_op { $$ = new ExprOpNot(new ExprApp(CUR_POS, new ExprApp(new ExprBuiltin(data->symbols.create("lessThan")), $1), $3)); }
| expr_op AND expr_op { $$ = new ExprOpAnd(CUR_POS, $1, $3); }
| expr_op OR expr_op { $$ = new ExprOpOr(CUR_POS, $1, $3); }
| expr_op IMPL expr_op { $$ = new ExprOpImpl(CUR_POS, $1, $3); }
| expr_op UPDATE expr_op { $$ = new ExprOpUpdate(CUR_POS, $1, $3); }
| expr_op '?' attrpath { $$ = new ExprOpHasAttr($1, *$3); }
2010-04-12 21:21:24 +00:00
| expr_op '+' expr_op
{ vector<Expr *> * l = new vector<Expr *>;
l->push_back($1);
l->push_back($3);
$$ = new ExprConcatStrings(CUR_POS, false, l);
2010-04-12 21:21:24 +00:00
}
| expr_op '-' expr_op { $$ = new ExprApp(CUR_POS, new ExprApp(new ExprBuiltin(data->symbols.create("sub")), $1), $3); }
| expr_op '*' expr_op { $$ = new ExprApp(CUR_POS, new ExprApp(new ExprBuiltin(data->symbols.create("mul")), $1), $3); }
| expr_op '/' expr_op { $$ = new ExprApp(CUR_POS, new ExprApp(new ExprBuiltin(data->symbols.create("div")), $1), $3); }
| expr_op CONCAT expr_op { $$ = new ExprOpConcatLists(CUR_POS, $1, $3); }
| expr_app
;
expr_app
: expr_app expr_select
{ $$ = new ExprApp(CUR_POS, $1, $2); }
| expr_select { $$ = $1; }
;
expr_select
: expr_simple '.' attrpath
{ $$ = new ExprSelect(CUR_POS, $1, *$3, 0); }
| expr_simple '.' attrpath OR_KW expr_select
{ $$ = new ExprSelect(CUR_POS, $1, *$3, $5); }
| /* Backwards compatibility: because Nixpkgs has a rarely used
function named or, allow stuff like map or [...]. */
expr_simple OR_KW
{ $$ = new ExprApp(CUR_POS, $1, new ExprVar(CUR_POS, data->symbols.create("or"))); }
| expr_simple { $$ = $1; }
;
expr_simple
: ID {
if (strcmp($1, "__curPos") == 0)
$$ = new ExprPos(CUR_POS);
else
$$ = new ExprVar(CUR_POS, data->symbols.create($1));
}
2010-04-12 21:21:24 +00:00
| INT { $$ = new ExprInt($1); }
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
| '"' string_parts '"' { $$ = $2; }
| IND_STRING_OPEN ind_string_parts IND_STRING_CLOSE {
$$ = stripIndentation(CUR_POS, data->symbols, *$2);
}
| PATH { $$ = new ExprPath(absPath($1, data->basePath)); }
| SPATH {
string path($1 + 1, strlen($1) - 2);
Path path2 = data->state.findFile(path);
/* The file wasn't found in the search path. However, we can't
throw an error here, because the expression might never be
evaluated. So return an expression that lazily calls
throw. */
$$ = path2 == ""
? (Expr * ) new ExprApp(
new ExprBuiltin(data->symbols.create("throw")),
new ExprString(data->symbols.create(
(format("file `%1%' was not found in the Nix search path (add it using $NIX_PATH or -I)") % path).str())))
: (Expr * ) new ExprPath(path2);
}
| URI { $$ = new ExprString(data->symbols.create($1)); }
| '(' expr ')' { $$ = $2; }
/* Let expressions `let {..., body = ...}' are just desugared
2010-04-12 21:21:24 +00:00
into `(rec {..., body = ...}).body'. */
| LET '{' binds '}'
{ $3->recursive = true; $$ = new ExprSelect(noPos, $3, data->symbols.create("body")); }
| REC '{' binds '}'
2010-04-12 23:33:23 +00:00
{ $3->recursive = true; $$ = $3; }
| '{' binds '}'
2010-04-12 23:33:23 +00:00
{ $$ = $2; }
| '[' expr_list ']' { $$ = $2; }
;
string_parts
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
: STR
| string_parts_interpolated { $$ = new ExprConcatStrings(CUR_POS, true, $1); }
2014-01-21 14:34:04 +00:00
| { $$ = new ExprString(data->symbols.create("")); }
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
;
string_parts_interpolated
: string_parts_interpolated STR { $$ = $1; $1->push_back($2); }
| string_parts_interpolated DOLLAR_CURLY expr '}' { backToString(scanner); $$ = $1; $1->push_back($3); }
| STR DOLLAR_CURLY expr '}'
{
backToString(scanner);
$$ = new vector<Expr *>;
$$->push_back($1);
$$->push_back($3);
}
| DOLLAR_CURLY expr '}'
{
backToString(scanner);
$$ = new vector<Expr *>;
$$->push_back($2);
}
;
ind_string_parts
2010-04-12 22:03:27 +00:00
: ind_string_parts IND_STR { $$ = $1; $1->push_back($2); }
| ind_string_parts DOLLAR_CURLY expr '}' { backToIndString(scanner); $$ = $1; $1->push_back($3); }
| { $$ = new vector<Expr *>; }
;
binds
: binds attrpath '=' expr ';' { $$ = $1; addAttr($$, *$2, $4, makeCurPos(@2, data)); }
| binds INHERIT attrs ';'
{ $$ = $1;
foreach (AttrPath::iterator, i, *$3) {
if ($$->attrs.find(i->symbol) != $$->attrs.end())
dupAttr(i->symbol, makeCurPos(@3, data), $$->attrs[i->symbol].pos);
Pos pos = makeCurPos(@3, data);
$$->attrs[i->symbol] = ExprAttrs::AttrDef(new ExprVar(CUR_POS, i->symbol), pos, true);
}
}
| binds INHERIT '(' expr ')' attrs ';'
{ $$ = $1;
/* !!! Should ensure sharing of the expression in $4. */
foreach (AttrPath::iterator, i, *$6) {
if ($$->attrs.find(i->symbol) != $$->attrs.end())
dupAttr(i->symbol, makeCurPos(@6, data), $$->attrs[i->symbol].pos);
$$->attrs[i->symbol] = ExprAttrs::AttrDef(new ExprSelect(CUR_POS, $4, i->symbol), makeCurPos(@6, data));
}
}
| { $$ = new ExprAttrs; }
;
attrs
: attrs attr { $$ = $1; $1->push_back(AttrName(data->symbols.create($2))); }
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
| attrs string_attr
{ $$ = $1;
ExprString *str = dynamic_cast<ExprString *>($2);
if (str) {
$$->push_back(AttrName(str->s));
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
delete str;
} else
throw ParseError(format("dynamic attributes not allowed in inherit at %1%")
% makeCurPos(@2, data));
}
| { $$ = new AttrPath; }
;
attrpath
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
: attrpath '.' attr { $$ = $1; $1->push_back(AttrName(data->symbols.create($3))); }
| attrpath '.' string_attr
{ $$ = $1;
ExprString *str = dynamic_cast<ExprString *>($3);
if (str) {
$$->push_back(AttrName(str->s));
delete str;
} else
$$->push_back(AttrName(static_cast<ExprConcatStrings *>($3)));
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
}
| attr { $$ = new vector<AttrName>; $$->push_back(AttrName(data->symbols.create($1))); }
| string_attr
{ $$ = new vector<AttrName>;
ExprString *str = dynamic_cast<ExprString *>($1);
if (str) {
$$->push_back(AttrName(str->s));
delete str;
} else
$$->push_back(AttrName(static_cast<ExprConcatStrings *>($1)));
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
}
;
attr
: ID { $$ = $1; }
| OR_KW { $$ = "or"; }
Dynamic attrs This adds new syntax for attribute names: * attrs."${name}" => getAttr name attrs * attrs ? "${name}" => isAttrs attrs && hasAttr attrs name * attrs."${name}" or def => if attrs ? "${name}" then attrs."${name}" else def * { "${name}" = value; } => listToAttrs [{ inherit name value; }] Of course, it's a bit more complicated than that. The attribute chains can be arbitrarily long and contain combinations of static and dynamic parts (e.g. attrs."${foo}".bar."${baz}" or qux), which is relatively straightforward for the getAttrs/hasAttrs cases but is more complex for the listToAttrs case due to rules about duplicate attribute definitions. For attribute sets with dynamic attribute names, duplicate static attributes are detected at parse time while duplicate dynamic attributes are detected when the attribute set is forced. So, for example, { a = null; a.b = null; "${"c"}" = true; } will be a parse-time error, while { a = {}; "${"a"}".b = null; c = true; } will be an eval-time error (technically that case could theoretically be detected at parse time, but the general case would require full evaluation). Moreover, duplicate dynamic attributes are not allowed even in cases where they would be with static attributes ({ a.b.d = true; a.b.c = false; } is legal, but { a."${"b"}".d = true; a."${"b"}".c = false; } is not). This restriction might be relaxed in the future in cases where the static variant would not be an error, but it is not obvious that that is desirable. Finally, recursive attribute sets with dynamic attributes have the static attributes in scope but not the dynamic ones. So rec { a = true; "${"b"}" = a; } is equivalent to { a = true; b = true; } but rec { "${"a"}" = true; b = a; } would be an error or use a from the surrounding scope if it exists. Note that the getAttr, getAttr or default, and hasAttr are all implemented purely in the parser as syntactic sugar, while attribute sets with dynamic attribute names required changes to the AST to be implemented cleanly. This is an alternative solution to and closes #167 Signed-off-by: Shea Levy <shea@shealevy.com>
2013-09-21 03:25:30 +00:00
;
string_attr
: '"' string_parts '"' { $$ = $2; }
| DOLLAR_CURLY expr '}' { $$ = new ExprConcatStrings(CUR_POS, true, new vector<Expr*>(1, $2)); }
;
expr_list
: expr_list expr_select { $$ = $1; $1->elems.push_back($2); /* !!! dangerous */ }
| { $$ = new ExprList; }
;
formals
: formal ',' formals
{ $$ = $3; addFormal(CUR_POS, $$, *$1); }
| formal
{ $$ = new Formals; addFormal(CUR_POS, $$, *$1); $$->ellipsis = false; }
|
{ $$ = new Formals; $$->ellipsis = false; }
| ELLIPSIS
{ $$ = new Formals; $$->ellipsis = true; }
;
formal
: ID { $$ = new Formal(data->symbols.create($1), 0); }
| ID '?' expr { $$ = new Formal(data->symbols.create($1), $3); }
;
2013-09-02 14:29:15 +00:00
%%
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <unistd.h>
#include <eval.hh>
namespace nix {
2013-09-02 14:29:15 +00:00
Expr * EvalState::parse(const char * text,
2013-09-02 16:34:04 +00:00
const Path & path, const Path & basePath, StaticEnv & staticEnv)
{
yyscan_t scanner;
ParseData data(*this);
data.basePath = basePath;
data.path = data.symbols.create(path);
yylex_init(&scanner);
yy_scan_string(text, scanner);
int res = yyparse(scanner, &data);
yylex_destroy(scanner);
2013-09-02 14:29:15 +00:00
if (res) throw ParseError(data.error);
data.result->bindVars(staticEnv);
2013-09-02 14:29:15 +00:00
return data.result;
}
Path resolveExprPath(Path path)
{
assert(path[0] == '/');
/* If `path' is a symlink, follow it. This is so that relative
path references work. */
struct stat st;
while (true) {
if (lstat(path.c_str(), &st))
throw SysError(format("getting status of `%1%'") % path);
if (!S_ISLNK(st.st_mode)) break;
path = absPath(readLink(path), dirOf(path));
}
/* If `path' refers to a directory, append `/default.nix'. */
if (S_ISDIR(st.st_mode))
path = canonPath(path + "/default.nix");
return path;
}
Expr * EvalState::parseExprFromFile(const Path & path)
{
Add primop ‘scopedImport’ ‘scopedImport’ works like ‘import’, except that it takes a set of attributes to be added to the lexical scope of the expression, essentially extending or overriding the builtin variables. For instance, the expression scopedImport { x = 1; } ./foo.nix where foo.nix contains ‘x’, will evaluate to 1. This has a few applications: * It allows getting rid of function argument specifications in package expressions. For instance, a package expression like: { stdenv, fetchurl, libfoo }: stdenv.mkDerivation { ... buildInputs = [ libfoo ]; } can now we written as just stdenv.mkDerivation { ... buildInputs = [ libfoo ]; } and imported in all-packages.nix as: bar = scopedImport pkgs ./bar.nix; So whereas we once had dependencies listed in three places (buildInputs, the function, and the call site), they now only need to appear in one place. * It allows overriding builtin functions. For instance, to trace all calls to ‘map’: let overrides = { map = f: xs: builtins.trace "map called!" (map f xs); # Ensure that our override gets propagated by calls to # import/scopedImport. import = fn: scopedImport overrides fn; scopedImport = attrs: fn: scopedImport (overrides // attrs) fn; # Also update ‘builtins’. builtins = builtins // overrides; }; in scopedImport overrides ./bla.nix * Similarly, it allows extending the set of builtin functions. For instance, during Nixpkgs/NixOS evaluation, the Nixpkgs library functions could be added to the default scope. There is a downside: calls to scopedImport are not memoized, unlike import. So importing a file multiple times leads to multiple parsings / evaluations. It would be possible to construct the AST only once, but that would require careful handling of variables/environments.
2014-05-26 11:46:11 +00:00
return parseExprFromFile(path, staticBaseEnv);
}
Expr * EvalState::parseExprFromFile(const Path & path, StaticEnv & staticEnv)
{
return parse(readFile(path).c_str(), path, dirOf(path), staticEnv);
}
2013-09-02 16:34:04 +00:00
Expr * EvalState::parseExprFromString(const string & s, const Path & basePath, StaticEnv & staticEnv)
{
return parse(s.c_str(), "(string)", basePath, staticEnv);
}
Expr * EvalState::parseExprFromString(const string & s, const Path & basePath)
{
2013-09-02 16:34:04 +00:00
return parseExprFromString(s, basePath, staticBaseEnv);
}
Add primop ‘scopedImport’ ‘scopedImport’ works like ‘import’, except that it takes a set of attributes to be added to the lexical scope of the expression, essentially extending or overriding the builtin variables. For instance, the expression scopedImport { x = 1; } ./foo.nix where foo.nix contains ‘x’, will evaluate to 1. This has a few applications: * It allows getting rid of function argument specifications in package expressions. For instance, a package expression like: { stdenv, fetchurl, libfoo }: stdenv.mkDerivation { ... buildInputs = [ libfoo ]; } can now we written as just stdenv.mkDerivation { ... buildInputs = [ libfoo ]; } and imported in all-packages.nix as: bar = scopedImport pkgs ./bar.nix; So whereas we once had dependencies listed in three places (buildInputs, the function, and the call site), they now only need to appear in one place. * It allows overriding builtin functions. For instance, to trace all calls to ‘map’: let overrides = { map = f: xs: builtins.trace "map called!" (map f xs); # Ensure that our override gets propagated by calls to # import/scopedImport. import = fn: scopedImport overrides fn; scopedImport = attrs: fn: scopedImport (overrides // attrs) fn; # Also update ‘builtins’. builtins = builtins // overrides; }; in scopedImport overrides ./bla.nix * Similarly, it allows extending the set of builtin functions. For instance, during Nixpkgs/NixOS evaluation, the Nixpkgs library functions could be added to the default scope. There is a downside: calls to scopedImport are not memoized, unlike import. So importing a file multiple times leads to multiple parsings / evaluations. It would be possible to construct the AST only once, but that would require careful handling of variables/environments.
2014-05-26 11:46:11 +00:00
void EvalState::addToSearchPath(const string & s, bool warn)
{
size_t pos = s.find('=');
string prefix;
Path path;
if (pos == string::npos) {
path = s;
} else {
prefix = string(s, 0, pos);
path = string(s, pos + 1);
}
2013-09-02 14:29:15 +00:00
path = absPath(path);
if (pathExists(path)) {
debug(format("adding path `%1%' to the search path") % path);
searchPath.push_back(std::pair<string, Path>(prefix, path));
} else if (warn)
printMsg(lvlError, format("warning: Nix search path entry `%1%' does not exist, ignoring") % path);
}
Path EvalState::findFile(const string & path)
{
foreach (SearchPath::iterator, i, searchPath) {
Path res;
2013-09-02 14:29:15 +00:00
if (i->first.empty())
res = i->second + "/" + path;
else {
if (path.compare(0, i->first.size(), i->first) != 0 ||
(path.size() > i->first.size() && path[i->first.size()] != '/'))
continue;
res = i->second +
(path.size() == i->first.size() ? "" : "/" + string(path, i->first.size()));
}
if (pathExists(res)) return canonPath(res);
}
return "";
}
}