Commit Graph

21519 Commits

Author SHA1 Message Date
Chris Lattner
93eea9c8bc Simplify ASTContext's transforming of NamedTypeDecls into types. It now doesn't unique them, the decls do for themselves. ASTContext still has the type symbol table.
Swift SVN r228
2011-02-22 06:07:32 +00:00
Chris Lattner
cb9b1838a6 Introduce and create a new TypeAliasDecl for typealiases.
Swift SVN r227
2011-02-22 05:47:08 +00:00
Chris Lattner
ad4389edc8 Introduce a new TypeAliasDecl class and NamedTypeDecl parent class. We're not creating it yet, but hey it's there.
Swift SVN r226
2011-02-22 05:37:01 +00:00
Chris Lattner
818cd27748 Implement tuple to tuple conversions that involve default element values.
Swift SVN r221
2010-11-27 17:17:57 +00:00
Chris Lattner
e488b69182 Implement sema support for TupleExprs with default elements (represented with null), implement support for conversions from scalar to tuples with multiple default elements.
Swift SVN r220
2010-11-25 23:02:23 +00:00
Chris Lattner
906ba908fd Rename 'data' to 'oneof', resolving a fixme in the spec. Yay for CLU!
Swift SVN r212
2010-11-11 01:20:36 +00:00
Chris Lattner
6f7564235d Wire up sema support for creating array types.
Swift SVN r207
2010-10-17 13:12:47 +00:00
Chris Lattner
ff38a2db1b AST Support for arrays.
Swift SVN r206
2010-10-17 12:56:04 +00:00
Chris Lattner
f92a3d56d7 Implement parser support for structs, improve datadecl dumping.
Swift SVN r192
2010-10-10 06:38:24 +00:00
Chris Lattner
bffee378d2 Generalize the expression walker to allow rewrites to be implemented.
Swift SVN r191
2010-10-10 06:19:16 +00:00
Chris Lattner
a885cd5dc5 Diagnose completely unresolved expressions with an error like this:
data.swift:114:10: error: ambiguous expression could not resolve a concrete type
var xx = :Zero;
         ^

Do this with a very general pre/post-order walking function.

Swift SVN r190
2010-10-10 06:11:47 +00:00
Chris Lattner
272cbcaba3 Implement parser, ast and minimal sema support for :foo expressions, type resolution isn't done yet.
Swift SVN r185
2010-10-10 00:15:46 +00:00
Chris Lattner
47e034fb0e Fix element values to have a proper function type when they take an argument. Now this:
data MaybeInt {
  None,
  Some int
}
var c = MaybeInt::Some 4;
var d = MaybeInt::Some;

-ast-dump's into:

(vardecl 'c' type='MaybeInt'
 (apply_expr type='MaybeInt'
   (declref_expr type='int -> MaybeInt' decl=Some)
   (integer_literal type='__builtin_int32_type' value=4)))
(vardecl 'd' type='int -> MaybeInt'
 (declref_expr type='int -> MaybeInt' decl=Some))


Swift SVN r183
2010-10-09 23:51:38 +00:00
Chris Lattner
ad438c55e3 Implement the rest of sema for scoped identifier references (X::Y) we now get a proper DeclRefExpr, and this:
var def = DataSearchFlags::None;

-ast-dump's into:

(vardecl 'def' type='DataSearchFlags'
 (declref_expr type='DataSearchFlags' decl=None))
 

Swift SVN r182
2010-10-09 23:46:16 +00:00
Chris Lattner
10e3eacc83 Create a DataElementDecl to represent the elements of a data, and wire up sema support for creating them when the data is defined.
Swift SVN r181
2010-10-09 23:24:56 +00:00
Chris Lattner
09a023c3cd Now that we have simple user defined data types, remove the __builtin_else_hack type. woot.
Swift SVN r179
2010-10-09 22:06:13 +00:00
Chris Lattner
ac855b690b Move ActOnTypeAlias to SemaDecl since a type alias *is* a decl that *installs* a type. Add some fixme's and rename some junk in ASTContext relating to the typemap. Diagnose redefinitions of named types. Install a typemap entry when a data declaration is defined.
We can now declare variables to be of data type.

Swift SVN r178
2010-10-09 22:01:25 +00:00
Chris Lattner
9ba168ace5 Fix Alias uniquing to reuse the already-uniqued identifier object instead of being a string map. Make getAliasType fail if a redefinition happens and make getNamedType return datatype's as well.
Swift SVN r177
2010-10-09 21:30:56 +00:00
Chris Lattner
aefb083fed Make the DataType class, wire it into ASTContext.
Swift SVN r176
2010-10-09 21:11:27 +00:00
Chris Lattner
a43a8f27f5 Wire data decls up to -ast-dump. Fix some minor bugs exposed along the way.
Swift SVN r175
2010-10-09 20:46:10 +00:00
Chris Lattner
0ad8498114 declare DeclData and create it in sema.
Swift SVN r174
2010-10-09 20:40:19 +00:00
Chris Lattner
57c697172a Split NamedDecl into NamedDecl with a ValueDecl subclass that has the type and value of the decl.
Swift SVN r173
2010-10-09 20:28:11 +00:00
Chris Lattner
91f5abd1e9 Implement parser support for assignment, what a hack :-)
Swift SVN r166
2010-09-18 17:26:01 +00:00
Chris Lattner
d2abfde434 introduce a new builtin type to represent the result of the else binary operator (which is a temporary hack until we get user defined types).
Use it to define ife/else.  Fix some precedence/parsing bugs that implementing else exposed. 

Swift SVN r162
2010-08-08 22:37:56 +00:00
Chris Lattner
d5f834512e Next step to allow use of named arguments: Create a new ArgDecl AST node, and have SemaDecl form them. We now correctly diagnose argument name redefinitions (though the diagnostic sucks because we don't have loc info)
Swift SVN r159
2010-08-07 20:25:05 +00:00
Chris Lattner
5afc6bdbd0 Add Type::getAs<> and support for minimally desugaring types. Start preserving more type sugar in the type conversion code.
Swift SVN r154
2010-08-05 23:05:54 +00:00
Chris Lattner
cbecebb18d -ast-dump shouldn't crash on empty identifiers
Swift SVN r153
2010-08-05 22:40:48 +00:00
Chris Lattner
429ecb81da Finish ast and sema support for name binding elements of tuples in var decls.
Swift SVN r152
2010-08-05 22:22:22 +00:00
Chris Lattner
c1294622bf Cleanup type stuff.
Swift SVN r147
2010-08-04 06:35:54 +00:00
Chris Lattner
e1e335610f int is no longer a keyword, it is a type alias for the new __builtin_int32_type type.
Swift SVN r146
2010-08-04 05:27:45 +00:00
Chris Lattner
a8916a6106 implement semantic analysis and AST representation for AliasType (aka typedefs in C).
Swift SVN r144
2010-08-04 05:14:57 +00:00
Chris Lattner
f704c58c42 Two significant changes:
1. Change the representation of TupleType to contain TupleTypeElt structs instead of PointerUnion<Type*, NameDecl*>
 2. Change type canonicalization to not strip field names off tuples, and change type conversion logic to use canonical types again.

Also adds some stuff to the todo list, destined to never be short.

Swift SVN r141
2010-08-04 03:35:53 +00:00
Chris Lattner
04f7a08621 add support for tuple field access with ".", e.g.:
var a : (int, var f : int, int);
  var b = a.field0+a.field1+a.f;

This also eliminates TupleConvertExpr.

Swift SVN r137
2010-08-03 06:55:08 +00:00
Chris Lattner
6858e37cf2 implement a lot more semantic analysis for tuple conversions.
Swift SVN r136
2010-08-03 05:25:10 +00:00
Chris Lattner
1bb344c282 strength reduce our handling of dependent types. Now, you can't compose dependent types, there isn't a (dep, int) type, the whole tuple gets dependent type. This simplifies things a bit.
Swift SVN r134
2010-08-03 04:36:05 +00:00
Chris Lattner
7a5b428772 start reworking tuple compatibility: two different tuples aren't compatible because their canonical type is the same, they are compatible because one can be converted to the other. I'm just going to rip out canonical type support for now.
Swift SVN r133
2010-08-03 04:17:26 +00:00
Chris Lattner
422da5de1f Give _0 a dependent type and use type inference to determine what it is based on context. Still some "bottom up" type inference logic missing, and we don't infer all the types in the expression tree of things like:
func6((_0 4) + _0);

yet.


Swift SVN r121
2010-07-30 22:04:44 +00:00
Chris Lattner
1a9f102090 switch expression printing over to use a visitor instead of hand rolled dispatch.
Swift SVN r120
2010-07-30 21:08:58 +00:00
Chris Lattner
9107a90d55 add support for representing dependent types.
Swift SVN r117
2010-07-30 18:18:28 +00:00
Chris Lattner
667078650a Implement support for anonymous closure arguments. For example, we can now compile this:
var func6 : ((int,int) -> int) -> ();  // Takes a function, returns nothing.
func funcdecl5() {
  func6(_0 + _1);         // Closure with two named anonymous arguments
}
into:

  (apply_expr type='()'
    (declref_expr type='(int, int) -> int -> ()' decl=func6)
    (closure_expr type='(int, int) -> int'
     (anondecl '_0' type='int')
     (anondecl '_1' type='int')
     (tuple_expr type='int'
      (binary_expr '+' type='int'
       (declref_expr type='int' decl=_0)
       (declref_expr type='int' decl=_1))))))))

However, there are still some problems with this (and we're definitely not doing type inference yet, all anon args are assumed 'int').

Swift SVN r111
2010-07-28 06:25:16 +00:00
Chris Lattner
9d33ef3616 add infrasturctuer for binding to "named anonymous" arguments.
Swift SVN r108
2010-07-27 21:37:29 +00:00
Chris Lattner
874421d4ce Implement AST and Sema support for auto-closures. We now parse:
var closure1 : () -> int = 4;  // Function producing 4 whenever it is called.
var closure2 : (int,int) -> int = 4; // Has some (dead) arguments.

into:

(vardecl 'closure1' type='() -> int'
 (closure_expr type='() -> int'
  (integer_literal type='int' value=4)))
(vardecl 'closure2' type='(int, int) -> int'
 (closure_expr type='(int, int) -> int'
  (integer_literal type='int' value=4)))


Swift SVN r101
2010-07-27 06:27:02 +00:00
Chris Lattner
8562512be0 test commit
Swift SVN r97
2010-07-27 05:39:40 +00:00
Chris Lattner
da8a6af90a add a new expression to model sequences instead of hacking up braceexpr to do it.
Swift SVN r94
2010-07-25 23:22:29 +00:00
Chris Lattner
a6b810b3e6 Enforce type rules in SemaExpr::ActOnBinaryExpr, validating that proper types are passed into the LHS/RHS of the operator.
Swift SVN r92
2010-07-25 23:04:03 +00:00
Chris Lattner
fa78da101b implement lazy construction of canonical types, ala clang. This only affect tuples at the moment, but will affect type aliases someday.
Swift SVN r90
2010-07-25 22:41:05 +00:00
Chris Lattner
7e0123c768 Kill off add/sub/mul/div as special AST nodes. Instead, + is just yet-another identifier token and BinaryExpr is a generalized infix call expression.
Swift SVN r88
2010-07-25 19:16:45 +00:00
Chris Lattner
12fa92e146 Add a new ApplyExpr expression AST node for function application.
Swift SVN r85
2010-07-25 05:37:45 +00:00
Chris Lattner
9b42d1b273 enhance parenexpr to be tupleexpr. The grouping paren case is just a special case of a tuple literal with one child.
Swift SVN r81
2010-07-24 21:24:44 +00:00
Chris Lattner
2bafb0e8fb Introduce new NamedDecl and FuncDecl classes. The difference between the two is syntactic sugar, so most clients want NamedDecl. This doesn't implement FuncDecl yet, which isn't a high prio for me in the short term.
Swift SVN r77
2010-07-24 19:09:09 +00:00