diff options
| author | bors <bors@rust-lang.org> | 2013-09-06 23:35:57 -0700 |
|---|---|---|
| committer | bors <bors@rust-lang.org> | 2013-09-06 23:35:57 -0700 |
| commit | 124eb2119c78651cfaaa7a046a101fa2e20f83ca (patch) | |
| tree | 7efa84a80df29080ae0314d0e3788131c72bdceb /src/libsyntax | |
| parent | 3e6de6b7da8ee88bf84b0e217900051334be08da (diff) | |
| parent | b6f3d3f24546a525d1eb80923692c1296eddc4dc (diff) | |
| download | rust-124eb2119c78651cfaaa7a046a101fa2e20f83ca.tar.gz rust-124eb2119c78651cfaaa7a046a101fa2e20f83ca.zip | |
auto merge of #9026 : jbclements/rust/let-var-hygiene, r=jbclements
This is a rebase of my approved pull request from ... the end of June? It introduces hygiene for let-bound variables.
Diffstat (limited to 'src/libsyntax')
| -rw-r--r-- | src/libsyntax/ast.rs | 56 | ||||
| -rw-r--r-- | src/libsyntax/ast_util.rs | 187 | ||||
| -rw-r--r-- | src/libsyntax/ext/base.rs | 115 | ||||
| -rw-r--r-- | src/libsyntax/ext/expand.rs | 847 | ||||
| -rw-r--r-- | src/libsyntax/ext/quote.rs | 26 | ||||
| -rw-r--r-- | src/libsyntax/ext/tt/macro_parser.rs | 19 | ||||
| -rw-r--r-- | src/libsyntax/ext/tt/macro_rules.rs | 16 | ||||
| -rw-r--r-- | src/libsyntax/fold.rs | 96 | ||||
| -rw-r--r-- | src/libsyntax/parse/parser.rs | 15 | ||||
| -rw-r--r-- | src/libsyntax/parse/token.rs | 169 | ||||
| -rw-r--r-- | src/libsyntax/print/pprust.rs | 10 | ||||
| -rw-r--r-- | src/libsyntax/util/interner.rs | 63 | ||||
| -rw-r--r-- | src/libsyntax/util/parser_testing.rs | 21 |
13 files changed, 1275 insertions, 365 deletions
diff --git a/src/libsyntax/ast.rs b/src/libsyntax/ast.rs index c7ebc344a9c..b993f98ec82 100644 --- a/src/libsyntax/ast.rs +++ b/src/libsyntax/ast.rs @@ -20,6 +20,10 @@ use std::option::Option; use std::to_str::ToStr; use extra::serialize::{Encodable, Decodable, Encoder, Decoder}; + +// FIXME #6993: in librustc, uses of "ident" should be replaced +// by just "Name". + // an identifier contains a Name (index into the interner // table) and a SyntaxContext to track renaming and // macro expansion per Flatt et al., "Macros @@ -32,6 +36,36 @@ impl Ident { pub fn new(name: Name) -> Ident { Ident {name: name, ctxt: EMPTY_CTXT}} } +// defining eq in this way is a way of guaranteeing that later stages of the +// compiler don't compare identifiers unhygienically. Unfortunately, some tests +// (specifically debuginfo in no-opt) want to do these comparisons, and that +// seems fine. If only I could find a nice way to statically ensure that +// the compiler "proper" never compares identifiers.... I'm leaving this +// code here (commented out) for potential use in debugging. Specifically, if +// there's a bug where "identifiers aren't matching", it may be because +// they should be compared using mtwt_resolve. In such a case, re-enabling this +// code (and disabling deriving(Eq) for Idents) could help to isolate the +// problem +/* impl Eq for Ident { + fn eq(&self, other: &Ident) -> bool { + if (self.ctxt == other.ctxt) { + self.name == other.name + } else { + // IF YOU SEE ONE OF THESE FAILS: it means that you're comparing + // idents that have different contexts. You can't fix this without + // knowing whether the comparison should be hygienic or non-hygienic. + // if it should be non-hygienic (most things are), just compare the + // 'name' fields of the idents. Or, even better, replace the idents + // with Name's. + fail!(fmt!("not allowed to compare these idents: %?, %?", self, other)); + } + } + fn ne(&self, other: &Ident) -> bool { + ! self.eq(other) + } +} +*/ + /// A SyntaxContext represents a chain of macro-expandings /// and renamings. Each macro expansion corresponds to /// a fresh uint @@ -47,6 +81,15 @@ impl Ident { // storage. pub type SyntaxContext = uint; +// the SCTable contains a table of SyntaxContext_'s. It +// represents a flattened tree structure, to avoid having +// managed pointers everywhere (that caused an ICE). +// the mark_memo and rename_memo fields are side-tables +// that ensure that adding the same mark to the same context +// gives you back the same context as before. This shouldn't +// change the semantics--everything here is immutable--but +// it should cut down on memory use *a lot*; applying a mark +// to a tree containing 50 identifiers would otherwise generate pub struct SCTable { table : ~[SyntaxContext_], mark_memo : HashMap<(SyntaxContext,Mrk),SyntaxContext>, @@ -70,6 +113,7 @@ pub enum SyntaxContext_ { // in the "from" slot. In essence, they're all // pointers to a single "rename" event node. Rename (Ident,Name,SyntaxContext), + // actually, IllegalCtxt may not be necessary. IllegalCtxt } @@ -99,6 +143,7 @@ pub type FnIdent = Option<Ident>; pub struct Lifetime { id: NodeId, span: Span, + // FIXME #7743 : change this to Name! ident: Ident } @@ -443,7 +488,7 @@ pub enum BlockCheckMode { UnsafeBlock, } -#[deriving(Eq, Encodable, Decodable,IterBytes)] +#[deriving(Clone, Eq, Encodable, Decodable,IterBytes)] pub struct Expr { id: NodeId, node: Expr_, @@ -544,10 +589,11 @@ pub enum token_tree { // a delimited sequence (the delimiters appear as the first // and last elements of the vector) tt_delim(@mut ~[token_tree]), + // These only make sense for right-hand-sides of MBE macros: // a kleene-style repetition sequence with a span, a tt_forest, - // an optional separator (?), and a boolean where true indicates + // an optional separator, and a boolean where true indicates // zero or more (*), and false indicates one or more (+). tt_seq(Span, @mut ~[token_tree], Option<::parse::token::Token>, bool), @@ -622,9 +668,13 @@ pub enum matcher_ { pub type mac = Spanned<mac_>; +// represents a macro invocation. The Path indicates which macro +// is being invoked, and the vector of token-trees contains the source +// of the macro invocation. +// There's only one flavor, now, so this could presumably be simplified. #[deriving(Clone, Eq, Encodable, Decodable, IterBytes)] pub enum mac_ { - mac_invoc_tt(Path,~[token_tree]), // new macro-invocation + mac_invoc_tt(Path,~[token_tree],SyntaxContext), // new macro-invocation } pub type lit = Spanned<lit_>; diff --git a/src/libsyntax/ast_util.rs b/src/libsyntax/ast_util.rs index 5ee8537750e..ee898c182e0 100644 --- a/src/libsyntax/ast_util.rs +++ b/src/libsyntax/ast_util.rs @@ -28,6 +28,8 @@ pub fn path_name_i(idents: &[Ident]) -> ~str { idents.map(|i| token::interner_get(i.name)).connect("::") } +// totally scary function: ignores all but the last element, should have +// a different name pub fn path_to_ident(path: &Path) -> Ident { path.segments.last().identifier } @@ -825,9 +827,6 @@ pub fn pat_is_ident(pat: @ast::Pat) -> bool { // HYGIENE FUNCTIONS -/// Construct an identifier with the given name and an empty context: -pub fn new_ident(name: Name) -> Ident { Ident {name: name, ctxt: 0}} - /// Extend a syntax context with a given mark pub fn new_mark(m:Mrk, tail:SyntaxContext) -> SyntaxContext { new_mark_internal(m,tail,get_sctable()) @@ -908,6 +907,15 @@ pub fn get_sctable() -> @mut SCTable { } } +/// print out an SCTable for debugging +pub fn display_sctable(table : &SCTable) { + error!("SC table:"); + for (idx,val) in table.table.iter().enumerate() { + error!("%4u : %?",idx,val); + } +} + + /// Add a value to the end of a vec, return its index fn idx_push<T>(vec: &mut ~[T], val: T) -> uint { vec.push(val); @@ -915,35 +923,76 @@ fn idx_push<T>(vec: &mut ~[T], val: T) -> uint { } /// Resolve a syntax object to a name, per MTWT. -pub fn resolve(id : Ident) -> Name { - resolve_internal(id, get_sctable()) +pub fn mtwt_resolve(id : Ident) -> Name { + resolve_internal(id, get_sctable(), get_resolve_table()) +} + +// FIXME #4536: must be pub for testing +pub type ResolveTable = HashMap<(Name,SyntaxContext),Name>; + +// okay, I admit, putting this in TLS is not so nice: +// fetch the SCTable from TLS, create one if it doesn't yet exist. +pub fn get_resolve_table() -> @mut ResolveTable { + static resolve_table_key: local_data::Key<@@mut ResolveTable> = &local_data::Key; + match local_data::get(resolve_table_key, |k| k.map(|&k| *k)) { + None => { + let new_table = @@mut HashMap::new(); + local_data::set(resolve_table_key,new_table); + *new_table + }, + Some(intr) => *intr + } } // Resolve a syntax object to a name, per MTWT. +// adding memoization to possibly resolve 500+ seconds in resolve for librustc (!) // FIXME #4536 : currently pub to allow testing -pub fn resolve_internal(id : Ident, table : &mut SCTable) -> Name { - match table.table[id.ctxt] { - EmptyCtxt => id.name, - // ignore marks here: - Mark(_,subctxt) => resolve_internal(Ident{name:id.name, ctxt: subctxt},table), - // do the rename if necessary: - Rename(Ident{name,ctxt},toname,subctxt) => { - // this could be cached or computed eagerly: - let resolvedfrom = resolve_internal(Ident{name:name,ctxt:ctxt},table); - let resolvedthis = resolve_internal(Ident{name:id.name,ctxt:subctxt},table); - if ((resolvedthis == resolvedfrom) - && (marksof(ctxt,resolvedthis,table) - == marksof(subctxt,resolvedthis,table))) { - toname - } else { - resolvedthis - } +pub fn resolve_internal(id : Ident, + table : &mut SCTable, + resolve_table : &mut ResolveTable) -> Name { + let key = (id.name,id.ctxt); + match resolve_table.contains_key(&key) { + false => { + let resolved = { + match table.table[id.ctxt] { + EmptyCtxt => id.name, + // ignore marks here: + Mark(_,subctxt) => + resolve_internal(Ident{name:id.name, ctxt: subctxt},table,resolve_table), + // do the rename if necessary: + Rename(Ident{name,ctxt},toname,subctxt) => { + let resolvedfrom = + resolve_internal(Ident{name:name,ctxt:ctxt},table,resolve_table); + let resolvedthis = + resolve_internal(Ident{name:id.name,ctxt:subctxt},table,resolve_table); + if ((resolvedthis == resolvedfrom) + && (marksof(ctxt,resolvedthis,table) + == marksof(subctxt,resolvedthis,table))) { + toname + } else { + resolvedthis + } + } + IllegalCtxt() => fail!(~"expected resolvable context, got IllegalCtxt") + } + }; + resolve_table.insert(key,resolved); + resolved + } + true => { + // it's guaranteed to be there, because we just checked that it was + // there and we never remove anything from the table: + *(resolve_table.find(&key).unwrap()) } - IllegalCtxt() => fail!(~"expected resolvable context, got IllegalCtxt") } } /// Compute the marks associated with a syntax context. +pub fn mtwt_marksof(ctxt: SyntaxContext, stopname: Name) -> ~[Mrk] { + marksof(ctxt, stopname, get_sctable()) +} + +// the internal function for computing marks // it's not clear to me whether it's better to use a [] mutable // vector or a cons-list for this. pub fn marksof(ctxt: SyntaxContext, stopname: Name, table: &SCTable) -> ~[Mrk] { @@ -970,6 +1019,16 @@ pub fn marksof(ctxt: SyntaxContext, stopname: Name, table: &SCTable) -> ~[Mrk] { } } +/// Return the outer mark for a context with a mark at the outside. +/// FAILS when outside is not a mark. +pub fn mtwt_outer_mark(ctxt: SyntaxContext) -> Mrk { + let sctable = get_sctable(); + match sctable.table[ctxt] { + ast::Mark(mrk,_) => mrk, + _ => fail!("can't retrieve outer mark when outside is not a mark") + } +} + /// Push a name... unless it matches the one on top, in which /// case pop and discard (so two of the same marks cancel) pub fn xorPush(marks: &mut ~[uint], mark: uint) { @@ -986,12 +1045,55 @@ pub fn getLast(arr: &~[Mrk]) -> uint { *arr.last() } +// are two paths equal when compared unhygienically? +// since I'm using this to replace ==, it seems appropriate +// to compare the span, global, etc. fields as well. +pub fn path_name_eq(a : &ast::Path, b : &ast::Path) -> bool { + (a.span == b.span) + && (a.global == b.global) + && (segments_name_eq(a.segments, b.segments)) +} + +// are two arrays of segments equal when compared unhygienically? +pub fn segments_name_eq(a : &[ast::PathSegment], b : &[ast::PathSegment]) -> bool { + if (a.len() != b.len()) { + false + } else { + for (idx,seg) in a.iter().enumerate() { + if (seg.identifier.name != b[idx].identifier.name) + // FIXME #7743: ident -> name problems in lifetime comparison? + || (seg.lifetime != b[idx].lifetime) + // can types contain idents? + || (seg.types != b[idx].types) { + return false; + } + } + true + } +} #[cfg(test)] mod test { use ast::*; use super::*; use std::io; + use opt_vec; + use std::hash::HashMap; + + fn ident_to_segment(id : &Ident) -> PathSegment { + PathSegment{identifier:id.clone(), lifetime: None, types: opt_vec::Empty} + } + + #[test] fn idents_name_eq_test() { + assert!(segments_name_eq([Ident{name:3,ctxt:4}, + Ident{name:78,ctxt:82}].map(ident_to_segment), + [Ident{name:3,ctxt:104}, + Ident{name:78,ctxt:182}].map(ident_to_segment))); + assert!(!segments_name_eq([Ident{name:3,ctxt:4}, + Ident{name:78,ctxt:82}].map(ident_to_segment), + [Ident{name:3,ctxt:104}, + Ident{name:77,ctxt:182}].map(ident_to_segment))); + } #[test] fn xorpush_test () { let mut s = ~[]; @@ -1122,29 +1224,30 @@ mod test { #[test] fn resolve_tests () { let a = 40; let mut t = new_sctable_internal(); + let mut rt = HashMap::new(); // - ctxt is MT - assert_eq!(resolve_internal(id(a,EMPTY_CTXT),&mut t),a); + assert_eq!(resolve_internal(id(a,EMPTY_CTXT),&mut t, &mut rt),a); // - simple ignored marks { let sc = unfold_marks(~[1,2,3],EMPTY_CTXT,&mut t); - assert_eq!(resolve_internal(id(a,sc),&mut t),a);} + assert_eq!(resolve_internal(id(a,sc),&mut t, &mut rt),a);} // - orthogonal rename where names don't match { let sc = unfold_test_sc(~[R(id(50,EMPTY_CTXT),51),M(12)],EMPTY_CTXT,&mut t); - assert_eq!(resolve_internal(id(a,sc),&mut t),a);} + assert_eq!(resolve_internal(id(a,sc),&mut t, &mut rt),a);} // - rename where names do match, but marks don't { let sc1 = new_mark_internal(1,EMPTY_CTXT,&mut t); let sc = unfold_test_sc(~[R(id(a,sc1),50), M(1), M(2)], EMPTY_CTXT,&mut t); - assert_eq!(resolve_internal(id(a,sc),&mut t), a);} + assert_eq!(resolve_internal(id(a,sc),&mut t, &mut rt), a);} // - rename where names and marks match { let sc1 = unfold_test_sc(~[M(1),M(2)],EMPTY_CTXT,&mut t); let sc = unfold_test_sc(~[R(id(a,sc1),50),M(1),M(2)],EMPTY_CTXT,&mut t); - assert_eq!(resolve_internal(id(a,sc),&mut t), 50); } + assert_eq!(resolve_internal(id(a,sc),&mut t, &mut rt), 50); } // - rename where names and marks match by literal sharing { let sc1 = unfold_test_sc(~[M(1),M(2)],EMPTY_CTXT,&mut t); let sc = unfold_test_sc(~[R(id(a,sc1),50)],sc1,&mut t); - assert_eq!(resolve_internal(id(a,sc),&mut t), 50); } + assert_eq!(resolve_internal(id(a,sc),&mut t, &mut rt), 50); } // - two renames of the same var.. can only happen if you use // local-expand to prevent the inner binding from being renamed // during the rename-pass caused by the first: @@ -1152,22 +1255,28 @@ mod test { { let sc = unfold_test_sc(~[R(id(a,EMPTY_CTXT),50), R(id(a,EMPTY_CTXT),51)], EMPTY_CTXT,&mut t); - assert_eq!(resolve_internal(id(a,sc),&mut t), 51); } + assert_eq!(resolve_internal(id(a,sc),&mut t, &mut rt), 51); } // the simplest double-rename: { let a_to_a50 = new_rename_internal(id(a,EMPTY_CTXT),50,EMPTY_CTXT,&mut t); let a50_to_a51 = new_rename_internal(id(a,a_to_a50),51,a_to_a50,&mut t); - assert_eq!(resolve_internal(id(a,a50_to_a51),&mut t),51); + assert_eq!(resolve_internal(id(a,a50_to_a51),&mut t, &mut rt),51); // mark on the outside doesn't stop rename: let sc = new_mark_internal(9,a50_to_a51,&mut t); - assert_eq!(resolve_internal(id(a,sc),&mut t),51); + assert_eq!(resolve_internal(id(a,sc),&mut t, &mut rt),51); // but mark on the inside does: let a50_to_a51_b = unfold_test_sc(~[R(id(a,a_to_a50),51), M(9)], a_to_a50, &mut t); - assert_eq!(resolve_internal(id(a,a50_to_a51_b),&mut t),50);} + assert_eq!(resolve_internal(id(a,a50_to_a51_b),&mut t, &mut rt),50);} } + #[test] fn mtwt_resolve_test(){ + let a = 40; + assert_eq!(mtwt_resolve(id(a,EMPTY_CTXT)),a); + } + + #[test] fn hashing_tests () { let mut t = new_sctable_internal(); assert_eq!(new_mark_internal(12,EMPTY_CTXT,&mut t),2); @@ -1177,4 +1286,16 @@ mod test { // I'm assuming that the rename table will behave the same.... } + #[test] fn resolve_table_hashing_tests() { + let mut t = new_sctable_internal(); + let mut rt = HashMap::new(); + assert_eq!(rt.len(),0); + resolve_internal(id(30,EMPTY_CTXT),&mut t, &mut rt); + assert_eq!(rt.len(),1); + resolve_internal(id(39,EMPTY_CTXT),&mut t, &mut rt); + assert_eq!(rt.len(),2); + resolve_internal(id(30,EMPTY_CTXT),&mut t, &mut rt); + assert_eq!(rt.len(),2); + } + } diff --git a/src/libsyntax/ext/base.rs b/src/libsyntax/ext/base.rs index 50683358f87..3d5d62aeadf 100644 --- a/src/libsyntax/ext/base.rs +++ b/src/libsyntax/ext/base.rs @@ -22,8 +22,7 @@ use std::hashmap::HashMap; // new-style macro! tt code: // -// SyntaxExpanderTT, SyntaxExpanderTTItem, MacResult, -// NormalTT, IdentTT +// MacResult, NormalTT, IdentTT // // also note that ast::mac used to have a bunch of extraneous cases and // is now probably a redundant AST node, can be merged with @@ -34,33 +33,46 @@ pub struct MacroDef { ext: SyntaxExtension } -pub type ItemDecorator = @fn(@ExtCtxt, +// No context arg for an Item Decorator macro, simply because +// adding it would require adding a ctxt field to all items. +// we could do this if it turns out to be useful. + +pub type ItemDecoratorFun = @fn(@ExtCtxt, Span, @ast::MetaItem, ~[@ast::item]) -> ~[@ast::item]; -pub struct SyntaxExpanderTT { - expander: SyntaxExpanderTTFun, - span: Option<Span> -} - pub type SyntaxExpanderTTFun = @fn(@ExtCtxt, Span, - &[ast::token_tree]) + &[ast::token_tree], + ast::SyntaxContext) -> MacResult; -pub struct SyntaxExpanderTTItem { - expander: SyntaxExpanderTTItemFun, - span: Option<Span> -} - pub type SyntaxExpanderTTItemFun = @fn(@ExtCtxt, Span, ast::Ident, + ~[ast::token_tree], + ast::SyntaxContext) + -> MacResult; + +// oog... in order to make the presentation of builtin_normal_tt_no_ctxt +// and builtin_ident_tt_no_ctxt palatable, we need one-off types for +// functions that don't consume a ctxt: + +pub type SyntaxExpanderTTFunNoCtxt = @fn(@ExtCtxt, + Span, + &[ast::token_tree]) + -> MacResult; + +pub type SyntaxExpanderTTItemFunNoCtxt = @fn(@ExtCtxt, + Span, + ast::Ident, ~[ast::token_tree]) -> MacResult; + + pub enum MacResult { MRExpr(@ast::Expr), MRItem(@ast::item), @@ -73,10 +85,10 @@ pub enum MacResult { pub enum SyntaxExtension { // #[auto_encode] and such - ItemDecorator(ItemDecorator), + ItemDecorator(ItemDecoratorFun), // Token-tree expanders - NormalTT(SyntaxExpanderTT), + NormalTT(SyntaxExpanderTTFun, Option<Span>), // An IdentTT is a macro that has an // identifier in between the name of the @@ -86,9 +98,10 @@ pub enum SyntaxExtension { // perhaps macro_rules! will lose its odd special identifier argument, // and this can go away also - IdentTT(SyntaxExpanderTTItem), + IdentTT(SyntaxExpanderTTItemFun, Option<Span>), } + // The SyntaxEnv is the environment that's threaded through the expansion // of macros. It contains bindings for macros, and also a special binding // for " block" (not a legal identifier) that maps to a BlockInfo @@ -120,12 +133,16 @@ type RenameList = ~[(ast::Ident,Name)]; // AST nodes into full ASTs pub fn syntax_expander_table() -> SyntaxEnv { // utility function to simplify creating NormalTT syntax extensions - fn builtin_normal_tt(f: SyntaxExpanderTTFun) -> @Transformer { - @SE(NormalTT(SyntaxExpanderTT{expander: f, span: None})) + // that ignore their contexts + fn builtin_normal_tt_no_ctxt(f: SyntaxExpanderTTFunNoCtxt) -> @Transformer { + let wrapped_expander : SyntaxExpanderTTFun = |a,b,c,_d|{f(a,b,c)}; + @SE(NormalTT(wrapped_expander, None)) } // utility function to simplify creating IdentTT syntax extensions - fn builtin_item_tt(f: SyntaxExpanderTTItemFun) -> @Transformer { - @SE(IdentTT(SyntaxExpanderTTItem{expander: f, span: None})) + // that ignore their contexts + fn builtin_item_tt_no_ctxt(f: SyntaxExpanderTTItemFunNoCtxt) -> @Transformer { + let wrapped_expander : SyntaxExpanderTTItemFun = |a,b,c,d,_e|{f(a,b,c,d)}; + @SE(IdentTT(wrapped_expander, None)) } let mut syntax_expanders = HashMap::new(); // NB identifier starts with space, and can't conflict with legal idents @@ -135,16 +152,15 @@ pub fn syntax_expander_table() -> SyntaxEnv { pending_renames : @mut ~[] })); syntax_expanders.insert(intern(&"macro_rules"), - builtin_item_tt( - ext::tt::macro_rules::add_new_extension)); + @SE(IdentTT(ext::tt::macro_rules::add_new_extension, None))); syntax_expanders.insert(intern(&"fmt"), - builtin_normal_tt(ext::fmt::expand_syntax_ext)); + builtin_normal_tt_no_ctxt(ext::fmt::expand_syntax_ext)); syntax_expanders.insert(intern(&"format"), - builtin_normal_tt(ext::ifmt::expand_format)); + builtin_normal_tt_no_ctxt(ext::ifmt::expand_format)); syntax_expanders.insert(intern(&"write"), - builtin_normal_tt(ext::ifmt::expand_write)); + builtin_normal_tt_no_ctxt(ext::ifmt::expand_write)); syntax_expanders.insert(intern(&"writeln"), - builtin_normal_tt(ext::ifmt::expand_writeln)); + builtin_normal_tt_no_ctxt(ext::ifmt::expand_writeln)); syntax_expanders.insert( intern(&"auto_encode"), @SE(ItemDecorator(ext::auto_encode::expand_auto_encode))); @@ -152,16 +168,16 @@ pub fn syntax_expander_table() -> SyntaxEnv { intern(&"auto_decode"), @SE(ItemDecorator(ext::auto_encode::expand_auto_decode))); syntax_expanders.insert(intern(&"env"), - builtin_normal_tt(ext::env::expand_env)); + builtin_normal_tt_no_ctxt(ext::env::expand_env)); syntax_expanders.insert(intern(&"option_env"), - builtin_normal_tt(ext::env::expand_option_env)); + builtin_normal_tt_no_ctxt(ext::env::expand_option_env)); syntax_expanders.insert(intern("bytes"), - builtin_normal_tt(ext::bytes::expand_syntax_ext)); + builtin_normal_tt_no_ctxt(ext::bytes::expand_syntax_ext)); syntax_expanders.insert(intern("concat_idents"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::concat_idents::expand_syntax_ext)); syntax_expanders.insert(intern(&"log_syntax"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::log_syntax::expand_syntax_ext)); syntax_expanders.insert(intern(&"deriving"), @SE(ItemDecorator( @@ -169,49 +185,50 @@ pub fn syntax_expander_table() -> SyntaxEnv { // Quasi-quoting expanders syntax_expanders.insert(intern(&"quote_tokens"), - builtin_normal_tt(ext::quote::expand_quote_tokens)); + builtin_normal_tt_no_ctxt( + ext::quote::expand_quote_tokens)); syntax_expanders.insert(intern(&"quote_expr"), - builtin_normal_tt(ext::quote::expand_quote_expr)); + builtin_normal_tt_no_ctxt(ext::quote::expand_quote_expr)); syntax_expanders.insert(intern(&"quote_ty"), - builtin_normal_tt(ext::quote::expand_quote_ty)); + builtin_normal_tt_no_ctxt(ext::quote::expand_quote_ty)); syntax_expanders.insert(intern(&"quote_item"), - builtin_normal_tt(ext::quote::expand_quote_item)); + builtin_normal_tt_no_ctxt(ext::quote::expand_quote_item)); syntax_expanders.insert(intern(&"quote_pat"), - builtin_normal_tt(ext::quote::expand_quote_pat)); + builtin_normal_tt_no_ctxt(ext::quote::expand_quote_pat)); syntax_expanders.insert(intern(&"quote_stmt"), - builtin_normal_tt(ext::quote::expand_quote_stmt)); + builtin_normal_tt_no_ctxt(ext::quote::expand_quote_stmt)); syntax_expanders.insert(intern(&"line"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::source_util::expand_line)); syntax_expanders.insert(intern(&"col"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::source_util::expand_col)); syntax_expanders.insert(intern(&"file"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::source_util::expand_file)); syntax_expanders.insert(intern(&"stringify"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::source_util::expand_stringify)); syntax_expanders.insert(intern(&"include"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::source_util::expand_include)); syntax_expanders.insert(intern(&"include_str"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::source_util::expand_include_str)); syntax_expanders.insert(intern(&"include_bin"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::source_util::expand_include_bin)); syntax_expanders.insert(intern(&"module_path"), - builtin_normal_tt( + builtin_normal_tt_no_ctxt( ext::source_util::expand_mod)); syntax_expanders.insert(intern(&"asm"), - builtin_normal_tt(ext::asm::expand_asm)); + builtin_normal_tt_no_ctxt(ext::asm::expand_asm)); syntax_expanders.insert(intern(&"cfg"), - builtin_normal_tt(ext::cfg::expand_cfg)); + builtin_normal_tt_no_ctxt(ext::cfg::expand_cfg)); syntax_expanders.insert( intern(&"trace_macros"), - builtin_normal_tt(ext::trace_macros::expand_trace_macros)); + builtin_normal_tt_no_ctxt(ext::trace_macros::expand_trace_macros)); MapChain::new(~syntax_expanders) } diff --git a/src/libsyntax/ext/expand.rs b/src/libsyntax/ext/expand.rs index ea277a8a625..cde418ca991 100644 --- a/src/libsyntax/ext/expand.rs +++ b/src/libsyntax/ext/expand.rs @@ -8,11 +8,12 @@ // option. This file may not be copied, modified, or distributed // except according to those terms. -use ast::{Block, Crate, NodeId, Expr_, ExprMac, Ident, mac_invoc_tt}; -use ast::{item_mac, Stmt_, StmtMac, StmtExpr, StmtSemi}; -use ast::{ILLEGAL_CTXT}; +use ast::{Block, Crate, NodeId, DeclLocal, Expr_, ExprMac, SyntaxContext}; +use ast::{Local, Ident, mac_invoc_tt}; +use ast::{item_mac, Mrk, Stmt_, StmtDecl, StmtMac, StmtExpr, StmtSemi}; +use ast::{token_tree}; use ast; -use ast_util::{new_rename, new_mark, resolve}; +use ast_util::{mtwt_outer_mark, new_rename, new_mark}; use attr; use attr::AttrMetaMethods; use codemap; @@ -23,7 +24,7 @@ use opt_vec; use parse; use parse::{parse_item_from_source_str}; use parse::token; -use parse::token::{ident_to_str, intern}; +use parse::token::{fresh_mark, fresh_name, ident_to_str, intern}; use visit; use visit::Visitor; @@ -41,8 +42,13 @@ pub fn expand_expr(extsbox: @mut SyntaxEnv, // entry-point for all syntax extensions. ExprMac(ref mac) => { match (*mac).node { + // it would almost certainly be cleaner to pass the whole + // macro invocation in, rather than pulling it apart and + // marking the tts and the ctxt separately. This also goes + // for the other three macro invocation chunks of code + // in this file. // Token-tree macros: - mac_invoc_tt(ref pth, ref tts) => { + mac_invoc_tt(ref pth, ref tts, ctxt) => { if (pth.segments.len() > 1u) { cx.span_fatal( pth.span, @@ -58,35 +64,38 @@ pub fn expand_expr(extsbox: @mut SyntaxEnv, pth.span, fmt!("macro undefined: '%s'", extnamestr)) } - Some(@SE(NormalTT(SyntaxExpanderTT{ - expander: exp, - span: exp_sp - }))) => { + Some(@SE(NormalTT(expandfun, exp_span))) => { cx.bt_push(ExpnInfo { call_site: s, callee: NameAndSpan { name: extnamestr, - span: exp_sp, + span: exp_span, }, }); - - let expanded = match exp(cx, mac.span, *tts) { - MRExpr(e) => e, - MRAny(expr_maker,_,_) => expr_maker(), - _ => { - cx.span_fatal( - pth.span, - fmt!( - "non-expr macro in expr pos: %s", - extnamestr + let fm = fresh_mark(); + // mark before: + let marked_before = mark_tts(*tts,fm); + let marked_ctxt = new_mark(fm, ctxt); + let expanded = + match expandfun(cx, mac.span, marked_before, marked_ctxt) { + MRExpr(e) => e, + MRAny(expr_maker,_,_) => expr_maker(), + _ => { + cx.span_fatal( + pth.span, + fmt!( + "non-expr macro in expr pos: %s", + extnamestr + ) ) - ) - } - }; + } + }; + // mark after: + let marked_after = mark_expr(expanded,fm); //keep going, outside-in let fully_expanded = - fld.fold_expr(expanded).node.clone(); + fld.fold_expr(marked_after).node.clone(); cx.bt_pop(); (fully_expanded, s) @@ -309,7 +318,6 @@ pub fn expand_mod_items(extsbox: @mut SyntaxEnv, ast::_mod { items: new_items, ..module_ } } - // eval $e with a new exts frame: macro_rules! with_exts_frame ( ($extsboxexpr:expr,$macros_escape:expr,$e:expr) => @@ -333,25 +341,16 @@ pub fn expand_item(extsbox: @mut SyntaxEnv, fld: @ast_fold, orig: @fn(@ast::item, @ast_fold) -> Option<@ast::item>) -> Option<@ast::item> { - // need to do expansion first... it might turn out to be a module. - let maybe_it = match it.node { - ast::item_mac(*) => expand_item_mac(extsbox, cx, it, fld), - _ => Some(it) - }; - match maybe_it { - Some(it) => { - match it.node { - ast::item_mod(_) | ast::item_foreign_mod(_) => { - cx.mod_push(it.ident); - let macro_escape = contains_macro_escape(it.attrs); - let result = with_exts_frame!(extsbox,macro_escape,orig(it,fld)); - cx.mod_pop(); - result - } - _ => orig(it,fld) - } - } - None => None + match it.node { + ast::item_mac(*) => expand_item_mac(extsbox, cx, it, fld), + ast::item_mod(_) | ast::item_foreign_mod(_) => { + cx.mod_push(it.ident); + let macro_escape = contains_macro_escape(it.attrs); + let result = with_exts_frame!(extsbox,macro_escape,orig(it,fld)); + cx.mod_pop(); + result + }, + _ => orig(it,fld) } } @@ -366,21 +365,22 @@ pub fn expand_item_mac(extsbox: @mut SyntaxEnv, cx: @ExtCtxt, it: @ast::item, fld: @ast_fold) -> Option<@ast::item> { - let (pth, tts) = match it.node { - item_mac(codemap::Spanned { node: mac_invoc_tt(ref pth, ref tts), _}) => { - (pth, (*tts).clone()) + let (pth, tts, ctxt) = match it.node { + item_mac(codemap::Spanned { node: mac_invoc_tt(ref pth, ref tts, ctxt), _}) => { + (pth, (*tts).clone(), ctxt) } _ => cx.span_bug(it.span, "invalid item macro invocation") }; let extname = &pth.segments[0].identifier; let extnamestr = ident_to_str(extname); + let fm = fresh_mark(); let expanded = match (*extsbox).find(&extname.name) { None => cx.span_fatal(pth.span, fmt!("macro undefined: '%s!'", extnamestr)), - Some(@SE(NormalTT(ref expand))) => { - if it.ident != parse::token::special_idents::invalid { + Some(@SE(NormalTT(expander, span))) => { + if it.ident.name != parse::token::special_idents::invalid.name { cx.span_fatal(pth.span, fmt!("macro %s! expects no ident argument, \ given '%s'", extnamestr, @@ -390,13 +390,16 @@ pub fn expand_item_mac(extsbox: @mut SyntaxEnv, call_site: it.span, callee: NameAndSpan { name: extnamestr, - span: expand.span + span: span } }); - ((*expand).expander)(cx, it.span, tts) + // mark before expansion: + let marked_before = mark_tts(tts,fm); + let marked_ctxt = new_mark(fm,ctxt); + expander(cx, it.span, marked_before, marked_ctxt) } - Some(@SE(IdentTT(ref expand))) => { - if it.ident == parse::token::special_idents::invalid { + Some(@SE(IdentTT(expander, span))) => { + if it.ident.name == parse::token::special_idents::invalid.name { cx.span_fatal(pth.span, fmt!("macro %s! expects an ident argument", extnamestr)); @@ -405,21 +408,27 @@ pub fn expand_item_mac(extsbox: @mut SyntaxEnv, call_site: it.span, callee: NameAndSpan { name: extnamestr, - span: expand.span + span: span } }); - ((*expand).expander)(cx, it.span, it.ident, tts) + // mark before expansion: + let marked_tts = mark_tts(tts,fm); + let marked_ctxt = new_mark(fm,ctxt); + expander(cx, it.span, it.ident, marked_tts, marked_ctxt) } _ => cx.span_fatal( it.span, fmt!("%s! is not legal in item position", extnamestr)) }; let maybe_it = match expanded { - MRItem(it) => fld.fold_item(it), + MRItem(it) => mark_item(it,fm).chain(|i| {fld.fold_item(i)}), MRExpr(_) => cx.span_fatal(pth.span, fmt!("expr macro in item position: %s", extnamestr)), - MRAny(_, item_maker, _) => item_maker().chain(|i| {fld.fold_item(i)}), + MRAny(_, item_maker, _) => item_maker().chain(|i| {mark_item(i,fm)}) + .chain(|i| {fld.fold_item(i)}), MRDef(ref mdef) => { + // yikes... no idea how to apply the mark to this. I'm afraid + // we're going to have to wait-and-see on this one. insert_macro(*extsbox,intern(mdef.name), @SE((*mdef).ext)); None } @@ -454,15 +463,17 @@ pub fn expand_stmt(extsbox: @mut SyntaxEnv, orig: @fn(&Stmt_, Span, @ast_fold) -> (Option<Stmt_>, Span)) -> (Option<Stmt_>, Span) { - let (mac, pth, tts, semi) = match *s { + // why the copying here and not in expand_expr? + // looks like classic changed-in-only-one-place + let (mac, pth, tts, semi, ctxt) = match *s { StmtMac(ref mac, semi) => { match mac.node { - mac_invoc_tt(ref pth, ref tts) => { - ((*mac).clone(), pth, (*tts).clone(), semi) + mac_invoc_tt(ref pth, ref tts, ctxt) => { + ((*mac).clone(), pth, (*tts).clone(), semi, ctxt) } } } - _ => return orig(s, sp, fld) + _ => return expand_non_macro_stmt(*extsbox,s,sp,fld,orig) }; if (pth.segments.len() > 1u) { cx.span_fatal( @@ -476,13 +487,16 @@ pub fn expand_stmt(extsbox: @mut SyntaxEnv, None => cx.span_fatal(pth.span, fmt!("macro undefined: '%s'", extnamestr)), - Some(@SE(NormalTT( - SyntaxExpanderTT{expander: exp, span: exp_sp}))) => { + Some(@SE(NormalTT(expandfun, exp_span))) => { cx.bt_push(ExpnInfo { call_site: sp, - callee: NameAndSpan { name: extnamestr, span: exp_sp } + callee: NameAndSpan { name: extnamestr, span: exp_span } }); - let expanded = match exp(cx, mac.span, tts) { + let fm = fresh_mark(); + // mark before expansion: + let marked_tts = mark_tts(tts,fm); + let marked_ctxt = new_mark(fm,ctxt); + let expanded = match expandfun(cx, mac.span, marked_tts, marked_ctxt) { MRExpr(e) => @codemap::Spanned { node: StmtExpr(e, cx.next_id()), span: e.span}, @@ -491,9 +505,10 @@ pub fn expand_stmt(extsbox: @mut SyntaxEnv, pth.span, fmt!("non-stmt macro in stmt pos: %s", extnamestr)) }; + let marked_after = mark_stmt(expanded,fm); //keep going, outside-in - let fully_expanded = match fld.fold_stmt(expanded) { + let fully_expanded = match fld.fold_stmt(marked_after) { Some(stmt) => { let fully_expanded = &stmt.node; cx.bt_pop(); @@ -521,6 +536,73 @@ pub fn expand_stmt(extsbox: @mut SyntaxEnv, } +// expand a non-macro stmt. this is essentially the fallthrough for +// expand_stmt, above. +fn expand_non_macro_stmt (exts: SyntaxEnv, + s: &Stmt_, + sp: Span, + fld: @ast_fold, + orig: @fn(&Stmt_, Span, @ast_fold) -> (Option<Stmt_>, Span)) + -> (Option<Stmt_>,Span) { + // is it a let? + match *s { + StmtDecl(@Spanned{node: DeclLocal(ref local), span: stmt_span}, node_id) => { + let block_info = get_block_info(exts); + let pending_renames = block_info.pending_renames; + + // take it apart: + let @Local{is_mutbl:is_mutbl, + ty:_, + pat:pat, + init:init, + id:id, + span:span + } = *local; + // types can't be copied automatically because of the owned ptr in ty_tup... + let ty = local.ty.clone(); + // expand the pat (it might contain exprs... #:(o)> + let expanded_pat = fld.fold_pat(pat); + // find the pat_idents in the pattern: + // oh dear heaven... this is going to include the enum names, as well.... + // ... but that should be okay, as long as the new names are gensyms + // for the old ones. + let idents = @mut ~[]; + let name_finder = new_name_finder(idents); + name_finder.visit_pat(expanded_pat,()); + // generate fresh names, push them to a new pending list + let new_pending_renames = @mut ~[]; + for ident in idents.iter() { + let new_name = fresh_name(ident); + new_pending_renames.push((*ident,new_name)); + } + let rename_fld = renames_to_fold(new_pending_renames); + // rewrite the pattern using the new names (the old ones + // have already been applied): + let rewritten_pat = rename_fld.fold_pat(expanded_pat); + // add them to the existing pending renames: + for pr in new_pending_renames.iter() {pending_renames.push(*pr)} + // also, don't forget to expand the init: + let new_init_opt = init.map(|e| fld.fold_expr(*e)); + let rewritten_local = + @Local{is_mutbl:is_mutbl, + ty:ty, + pat:rewritten_pat, + init:new_init_opt, + id:id, + span:span}; + (Some(StmtDecl(@Spanned{node:DeclLocal(rewritten_local), + span: stmt_span},node_id)), + sp) + }, + _ => { + orig(s, sp, fld) + } + } +} + +// a visitor that extracts the pat_ident paths +// from a given thingy and puts them in a mutable +// array (passed in to the traversal) #[deriving(Clone)] struct NewNameFinderContext { ident_accumulator: @mut ~[ast::Ident], @@ -664,8 +746,133 @@ impl Visitor<()> for NewNameFinderContext { } } +// a visitor that extracts the paths +// from a given thingy and puts them in a mutable +// array (passed in to the traversal) +#[deriving(Clone)] +struct NewPathExprFinderContext { + path_accumulator: @mut ~[ast::Path], +} + +// XXX : YIKES a lot of boilerplate again.... +impl Visitor<()> for NewPathExprFinderContext { + + fn visit_expr(&mut self, expr: @ast::Expr, _: ()) { + match *expr { + ast::Expr{id:_,span:_,node:ast::ExprPath(ref p)} => { + self.path_accumulator.push(p.clone()); + // not calling visit_path, should be fine. + } + _ => visit::walk_expr(self,expr,()) + } + } + + + // XXX: Methods below can become default methods. + + fn visit_pat(&mut self, pattern: @ast::Pat, _: ()) { + visit::walk_pat(self,pattern,()) + } + + fn visit_mod(&mut self, module: &ast::_mod, _: Span, _: NodeId, _: ()) { + visit::walk_mod(self, module, ()) + } + + fn visit_view_item(&mut self, view_item: &ast::view_item, _: ()) { + visit::walk_view_item(self, view_item, ()) + } + + fn visit_item(&mut self, item: @ast::item, _: ()) { + visit::walk_item(self, item, ()) + } + + fn visit_foreign_item(&mut self, + foreign_item: @ast::foreign_item, + _: ()) { + visit::walk_foreign_item(self, foreign_item, ()) + } + + fn visit_local(&mut self, local: @ast::Local, _: ()) { + visit::walk_local(self, local, ()) + } + + fn visit_block(&mut self, block: &ast::Block, _: ()) { + visit::walk_block(self, block, ()) + } + + fn visit_stmt(&mut self, stmt: @ast::Stmt, _: ()) { + visit::walk_stmt(self, stmt, ()) + } + + fn visit_arm(&mut self, arm: &ast::Arm, _: ()) { + visit::walk_arm(self, arm, ()) + } + + fn visit_decl(&mut self, decl: @ast::Decl, _: ()) { + visit::walk_decl(self, decl, ()) + } + + fn visit_expr_post(&mut self, _: @ast::Expr, _: ()) { + // Empty! + } + + fn visit_ty(&mut self, typ: &ast::Ty, _: ()) { + visit::walk_ty(self, typ, ()) + } + + fn visit_generics(&mut self, generics: &ast::Generics, _: ()) { + visit::walk_generics(self, generics, ()) + } + + fn visit_fn(&mut self, + function_kind: &visit::fn_kind, + function_declaration: &ast::fn_decl, + block: &ast::Block, + span: Span, + node_id: NodeId, + _: ()) { + visit::walk_fn(self, + function_kind, + function_declaration, + block, + span, + node_id, + ()) + } + + fn visit_ty_method(&mut self, ty_method: &ast::TypeMethod, _: ()) { + visit::walk_ty_method(self, ty_method, ()) + } + + fn visit_trait_method(&mut self, + trait_method: &ast::trait_method, + _: ()) { + visit::walk_trait_method(self, trait_method, ()) + } + + fn visit_struct_def(&mut self, + struct_def: @ast::struct_def, + ident: Ident, + generics: &ast::Generics, + node_id: NodeId, + _: ()) { + visit::walk_struct_def(self, + struct_def, + ident, + generics, + node_id, + ()) + } + + fn visit_struct_field(&mut self, + struct_field: @ast::struct_field, + _: ()) { + visit::walk_struct_field(self, struct_field, ()) + } +} + // return a visitor that extracts the pat_ident paths -// from a given pattern and puts them in a mutable +// from a given thingy and puts them in a mutable // array (passed in to the traversal) pub fn new_name_finder(idents: @mut ~[ast::Ident]) -> @mut Visitor<()> { let context = @mut NewNameFinderContext { @@ -674,16 +881,60 @@ pub fn new_name_finder(idents: @mut ~[ast::Ident]) -> @mut Visitor<()> { context as @mut Visitor<()> } +// return a visitor that extracts the paths +// from a given pattern and puts them in a mutable +// array (passed in to the traversal) +pub fn new_path_finder(paths: @mut ~[ast::Path]) -> @mut Visitor<()> { + let context = @mut NewPathExprFinderContext { + path_accumulator: paths, + }; + context as @mut Visitor<()> +} + +// expand a block. pushes a new exts_frame, then calls expand_block_elts pub fn expand_block(extsbox: @mut SyntaxEnv, _cx: @ExtCtxt, blk: &Block, fld: @ast_fold, - orig: @fn(&Block, @ast_fold) -> Block) + _orig: @fn(&Block, @ast_fold) -> Block) -> Block { // see note below about treatment of exts table - with_exts_frame!(extsbox,false,orig(blk,fld)) + with_exts_frame!(extsbox,false, + expand_block_elts(*extsbox, blk, fld)) +} + +// expand the elements of a block. +pub fn expand_block_elts(exts: SyntaxEnv, b: &Block, fld: @ast_fold) -> Block { + let block_info = get_block_info(exts); + let pending_renames = block_info.pending_renames; + let rename_fld = renames_to_fold(pending_renames); + let new_view_items = b.view_items.map(|x| fld.fold_view_item(x)); + let mut new_stmts = ~[]; + for x in b.stmts.iter() { + match fld.fold_stmt(mustbesome(rename_fld.fold_stmt(*x))) { + Some(s) => new_stmts.push(s), + None => () + } + } + let new_expr = b.expr.map(|x| fld.fold_expr(rename_fld.fold_expr(*x))); + Block{ + view_items: new_view_items, + stmts: new_stmts, + expr: new_expr, + id: fld.new_id(b.id), + rules: b.rules, + span: b.span, + } } +// rename_fold should never return "None". +// (basically, just .get() with a better message...) +fn mustbesome<T>(val : Option<T>) -> T { + match val { + Some(v) => v, + None => fail!("rename_fold returned None") + } +} // get the (innermost) BlockInfo from an exts stack fn get_block_info(exts : SyntaxEnv) -> BlockInfo { @@ -694,43 +945,14 @@ fn get_block_info(exts : SyntaxEnv) -> BlockInfo { } } - -// given a mutable list of renames, return a tree-folder that applies those -// renames. -fn renames_to_fold(renames : @mut ~[(ast::Ident,ast::Name)]) -> @ast_fold { - let afp = default_ast_fold(); - let f_pre = @AstFoldFns { - fold_ident: |id,_| { - // the individual elements are memoized... it would - // also be possible to memoize on the whole list at once. - let new_ctxt = renames.iter().fold(id.ctxt,|ctxt,&(from,to)| { - new_rename(from,to,ctxt) - }); - ast::Ident{name:id.name,ctxt:new_ctxt} - }, - .. *afp - }; - make_fold(f_pre) -} - -// perform a bunch of renames -fn apply_pending_renames(folder : @ast_fold, stmt : ast::Stmt) -> @ast::Stmt { - match folder.fold_stmt(&stmt) { - Some(s) => s, - None => fail!(fmt!("renaming of stmt produced None")) - } -} - - - pub fn new_span(cx: @ExtCtxt, sp: Span) -> Span { /* this discards information in the case of macro-defining macros */ return Span {lo: sp.lo, hi: sp.hi, expn_info: cx.backtrace()}; } // FIXME (#2247): this is a moderately bad kludge to inject some macros into -// the default compilation environment. It would be much nicer to use -// a mechanism like syntax_quote to ensure hygiene. +// the default compilation environment in that it injects strings, rather than +// syntax elements. pub fn std_macros() -> @str { return @@ -1179,62 +1401,165 @@ pub fn expand_crate(parse_sess: @mut parse::ParseSess, return ret; } -// given a function from idents to idents, produce -// an ast_fold that applies that function: -pub fn fun_to_ident_folder(f: @fn(ast::Ident)->ast::Ident) -> @ast_fold{ - let afp = default_ast_fold(); - let f_pre = @AstFoldFns{ - fold_ident : |id, _| f(id), - .. *afp - }; - make_fold(f_pre) +// HYGIENIC CONTEXT EXTENSION: +// all of these functions are for walking over +// ASTs and making some change to the context of every +// element that has one. a CtxtFn is a trait-ified +// version of a closure in (SyntaxContext -> SyntaxContext). +// the ones defined here include: +// Renamer - add a rename to a context +// MultiRenamer - add a set of renames to a context +// Marker - add a mark to a context +// Repainter - replace a context (maybe Replacer would be a better name?) + +// a function in SyntaxContext -> SyntaxContext +pub trait CtxtFn{ + fn f(&self, ast::SyntaxContext) -> ast::SyntaxContext; } -// update the ctxts in a path to get a rename node -pub fn new_ident_renamer(from: ast::Ident, - to: ast::Name) -> - @fn(ast::Ident)->ast::Ident { - |id : ast::Ident| - ast::Ident{ - name: id.name, - ctxt: new_rename(from,to,id.ctxt) +// a renamer adds a rename to the syntax context +pub struct Renamer { + from : ast::Ident, + to : ast::Name +} + +impl CtxtFn for Renamer { + fn f(&self, ctxt : ast::SyntaxContext) -> ast::SyntaxContext { + new_rename(self.from,self.to,ctxt) + } +} + +// a renamer that performs a whole bunch of renames +pub struct MultiRenamer { + renames : @mut ~[(ast::Ident,ast::Name)] +} + +impl CtxtFn for MultiRenamer { + fn f(&self, starting_ctxt : ast::SyntaxContext) -> ast::SyntaxContext { + // the individual elements are memoized... it would + // also be possible to memoize on the whole list at once. + self.renames.iter().fold(starting_ctxt,|ctxt,&(from,to)| { + new_rename(from,to,ctxt) + }) + } +} + +// a marker adds the given mark to the syntax context +pub struct Marker { mark : Mrk } + +impl CtxtFn for Marker { + fn f(&self, ctxt : ast::SyntaxContext) -> ast::SyntaxContext { + new_mark(self.mark,ctxt) } } +// a repainter just replaces the given context with the one it's closed over +pub struct Repainter { ctxt : SyntaxContext } -// update the ctxts in a path to get a mark node -pub fn new_ident_marker(mark: uint) -> - @fn(ast::Ident)->ast::Ident { - |id : ast::Ident| - ast::Ident{ - name: id.name, - ctxt: new_mark(mark,id.ctxt) +impl CtxtFn for Repainter { + fn f(&self, _ctxt : ast::SyntaxContext) -> ast::SyntaxContext { + self.ctxt } } -// perform resolution (in the MTWT sense) on all of the -// idents in the tree. This is the final step in expansion. -pub fn new_ident_resolver() -> - @fn(ast::Ident)->ast::Ident { - |id : ast::Ident| - ast::Ident { - name : resolve(id), - ctxt : ILLEGAL_CTXT +// given a function from ctxts to ctxts, produce +// an ast_fold that applies that function to all ctxts: +pub fn fun_to_ctxt_folder<T : 'static + CtxtFn>(cf: @T) -> @AstFoldFns { + let afp = default_ast_fold(); + let fi : @fn(ast::Ident, @ast_fold) -> ast::Ident = + |ast::Ident{name, ctxt}, _| { + ast::Ident{name:name,ctxt:cf.f(ctxt)} + }; + let fm : @fn(&ast::mac_, Span, @ast_fold) -> (ast::mac_,Span) = + |m, sp, fld| { + match *m { + mac_invoc_tt(ref path, ref tts, ctxt) => + (mac_invoc_tt(fld.fold_path(path), + fold_tts(*tts,fld), + cf.f(ctxt)), + sp) + } + + }; + @AstFoldFns{ + fold_ident : fi, + fold_mac : fm, + .. *afp } } + +// given a mutable list of renames, return a tree-folder that applies those +// renames. +// FIXME #4536: currently pub to allow testing +pub fn renames_to_fold(renames : @mut ~[(ast::Ident,ast::Name)]) -> @AstFoldFns { + fun_to_ctxt_folder(@MultiRenamer{renames : renames}) +} + +// just a convenience: +pub fn new_mark_folder(m : Mrk) -> @AstFoldFns { + fun_to_ctxt_folder(@Marker{mark:m}) +} + +pub fn new_rename_folder(from : ast::Ident, to : ast::Name) -> @AstFoldFns { + fun_to_ctxt_folder(@Renamer{from:from,to:to}) +} + +// apply a given mark to the given token trees. Used prior to expansion of a macro. +fn mark_tts(tts : &[token_tree], m : Mrk) -> ~[token_tree] { + fold_tts(tts,new_mark_folder(m) as @ast_fold) +} + +// apply a given mark to the given expr. Used following the expansion of a macro. +fn mark_expr(expr : @ast::Expr, m : Mrk) -> @ast::Expr { + new_mark_folder(m).fold_expr(expr) +} + +// apply a given mark to the given stmt. Used following the expansion of a macro. +fn mark_stmt(expr : &ast::Stmt, m : Mrk) -> @ast::Stmt { + new_mark_folder(m).fold_stmt(expr).unwrap() +} + +// apply a given mark to the given item. Used following the expansion of a macro. +fn mark_item(expr : @ast::item, m : Mrk) -> Option<@ast::item> { + new_mark_folder(m).fold_item(expr) +} + +// replace all contexts in a given expr with the given mark. Used +// for capturing macros +pub fn replace_ctxts(expr : @ast::Expr, ctxt : SyntaxContext) -> @ast::Expr { + fun_to_ctxt_folder(@Repainter{ctxt:ctxt}).fold_expr(expr) +} + +// take the mark from the given ctxt (that has a mark at the outside), +// and apply it to everything in the token trees, thereby cancelling +// that mark. +pub fn mtwt_cancel_outer_mark(tts: &[ast::token_tree], ctxt: ast::SyntaxContext) + -> ~[ast::token_tree] { + let outer_mark = mtwt_outer_mark(ctxt); + mark_tts(tts,outer_mark) +} + + #[cfg(test)] mod test { use super::*; use ast; use ast::{Attribute_, AttrOuter, MetaWord, EMPTY_CTXT}; + use ast_util::{get_sctable, mtwt_marksof, mtwt_resolve, new_rename}; + use ast_util; use codemap; use codemap::Spanned; + use fold; use parse; - use parse::token::{intern, get_ident_interner}; + use parse::token::{fresh_mark, gensym, intern, get_ident_interner, ident_to_str}; + use parse::token; use print::pprust; - use util::parser_testing::{string_to_item, string_to_pat, strs_to_idents}; + use std; + use util::parser_testing::{string_to_crate, string_to_crate_and_sess}; + use util::parser_testing::{string_to_pat, string_to_tts, strs_to_idents}; + use visit; // make sure that fail! is present #[test] fn fail_exists_test () { @@ -1332,28 +1657,251 @@ mod test { } } + #[test] fn cancel_outer_mark_test(){ + let invalid_name = token::special_idents::invalid.name; + let ident_str = @"x"; + let tts = string_to_tts(ident_str); + let fm = fresh_mark(); + let marked_once = fold::fold_tts(tts,new_mark_folder(fm) as @fold::ast_fold); + assert_eq!(marked_once.len(),1); + let marked_once_ctxt = + match marked_once[0] { + ast::tt_tok(_,token::IDENT(id,_)) => id.ctxt, + _ => fail!(fmt!("unexpected shape for marked tts: %?",marked_once[0])) + }; + assert_eq!(mtwt_marksof(marked_once_ctxt,invalid_name),~[fm]); + let remarked = mtwt_cancel_outer_mark(marked_once,marked_once_ctxt); + assert_eq!(remarked.len(),1); + match remarked[0] { + ast::tt_tok(_,token::IDENT(id,_)) => + assert_eq!(mtwt_marksof(id.ctxt,invalid_name),~[]), + _ => fail!(fmt!("unexpected shape for marked tts: %?",remarked[0])) + } + } + #[test] fn renaming () { - let maybe_item_ast = string_to_item(@"fn a() -> int { let b = 13; b }"); - let item_ast = match maybe_item_ast { - Some(x) => x, - None => fail!("test case fail") - }; + let item_ast = string_to_crate(@"fn f() -> int { a }"); let a_name = intern("a"); - let a2_name = intern("a2"); - let renamer = new_ident_renamer(ast::Ident{name:a_name,ctxt:EMPTY_CTXT}, + let a2_name = gensym("a2"); + let renamer = new_rename_folder(ast::Ident{name:a_name,ctxt:EMPTY_CTXT}, a2_name); - let renamed_ast = fun_to_ident_folder(renamer).fold_item(item_ast).unwrap(); - let resolver = new_ident_resolver(); - let resolved_ast = fun_to_ident_folder(resolver).fold_item(renamed_ast).unwrap(); - let resolved_as_str = pprust::item_to_str(resolved_ast, - get_ident_interner()); - assert_eq!(resolved_as_str,~"fn a2() -> int { let b = 13; b }"); + let renamed_ast = renamer.fold_crate(item_ast); + let varrefs = @mut ~[]; + visit::walk_crate(&mut new_path_finder(varrefs), &renamed_ast, ()); + match varrefs { + @[Path{segments:[ref seg],_}] => assert_eq!(mtwt_resolve(seg.identifier),a2_name), + _ => assert_eq!(0,1) + } + + // try a double-rename, with pending_renames. + let a3_name = gensym("a3"); + // a context that renames from ("a",empty) to "a2" : + let ctxt2 = new_rename(ast::Ident::new(a_name),a2_name,EMPTY_CTXT); + let pending_renames = @mut ~[(ast::Ident::new(a_name),a2_name), + (ast::Ident{name:a_name,ctxt:ctxt2},a3_name)]; + let double_renamed = renames_to_fold(pending_renames).fold_crate(item_ast); + let varrefs = @mut ~[]; + visit::walk_crate(&mut new_path_finder(varrefs), &double_renamed, ()); + match varrefs { + @[Path{segments:[ref seg],_}] => assert_eq!(mtwt_resolve(seg.identifier),a3_name), + _ => assert_eq!(0,1) + } + } + fn fake_print_crate(crate: @ast::Crate) { + let s = pprust::rust_printer(std::io::stderr(),get_ident_interner()); + pprust::print_crate_(s, crate); + } + fn expand_crate_str(crate_str: @str) -> @ast::Crate { + let (crate_ast,ps) = string_to_crate_and_sess(crate_str); + // the cfg argument actually does matter, here... + expand_crate(ps,~[],crate_ast) } - // sigh... it looks like I have two different renaming mechanisms, now... + //fn expand_and_resolve(crate_str: @str) -> ast::crate { + //let expanded_ast = expand_crate_str(crate_str); + // std::io::println(fmt!("expanded: %?\n",expanded_ast)); + //mtwt_resolve_crate(expanded_ast) + //} + //fn expand_and_resolve_and_pretty_print (crate_str : @str) -> ~str { + //let resolved_ast = expand_and_resolve(crate_str); + //pprust::to_str(&resolved_ast,fake_print_crate,get_ident_interner()) + //} + + #[test] fn macro_tokens_should_match(){ + expand_crate_str(@"macro_rules! m((a)=>(13)) fn main(){m!(a);}"); + } + + // renaming tests expand a crate and then check that the bindings match + // the right varrefs. The specification of the test case includes the + // text of the crate, and also an array of arrays. Each element in the + // outer array corresponds to a binding in the traversal of the AST + // induced by visit. Each of these arrays contains a list of indexes, + // interpreted as the varrefs in the varref traversal that this binding + // should match. So, for instance, in a program with two bindings and + // three varrefs, the array ~[~[1,2],~[0]] would indicate that the first + // binding should match the second two varrefs, and the second binding + // should match the first varref. + // + // The comparisons are done post-mtwt-resolve, so we're comparing renamed + // names; differences in marks don't matter any more. + // + // oog... I also want tests that check "binding-identifier-=?". That is, + // not just "do these have the same name", but "do they have the same + // name *and* the same marks"? Understanding this is really pretty painful. + // in principle, you might want to control this boolean on a per-varref basis, + // but that would make things even harder to understand, and might not be + // necessary for thorough testing. + type renaming_test = (&'static str, ~[~[uint]], bool); + + #[test] + fn automatic_renaming () { + let tests : ~[renaming_test] = + ~[// b & c should get new names throughout, in the expr too: + ("fn a() -> int { let b = 13; let c = b; b+c }", + ~[~[0,1],~[2]], false), + // both x's should be renamed (how is this causing a bug?) + ("fn main () {let x : int = 13;x;}", + ~[~[0]], false), + // the use of b after the + should be renamed, the other one not: + ("macro_rules! f (($x:ident) => (b + $x)) fn a() -> int { let b = 13; f!(b)}", + ~[~[1]], false), + // the b before the plus should not be renamed (requires marks) + ("macro_rules! f (($x:ident) => ({let b=9; ($x + b)})) fn a() -> int { f!(b)}", + ~[~[1]], false), + // the marks going in and out of letty should cancel, allowing that $x to + // capture the one following the semicolon. + // this was an awesome test case, and caught a *lot* of bugs. + ("macro_rules! letty(($x:ident) => (let $x = 15;)) + macro_rules! user(($x:ident) => ({letty!($x); $x})) + fn main() -> int {user!(z)}", + ~[~[0]], false), + // FIXME #8062: this test exposes a *potential* bug; our system does + // not behave exactly like MTWT, but I haven't thought of a way that + // this could cause a bug in Rust, yet. + // ("fn main() {let hrcoo = 19; macro_rules! getx(()=>(hrcoo)); getx!();}", + // ~[~[0]], true) + // FIXME #6994: the next string exposes the bug referred to in issue 6994, so I'm + // commenting it out. + // the z flows into and out of two macros (g & f) along one path, and one + // (just g) along the other, so the result of the whole thing should + // be "let z_123 = 3; z_123" + //"macro_rules! g (($x:ident) => + // ({macro_rules! f(($y:ident)=>({let $y=3;$x}));f!($x)})) + // fn a(){g!(z)}" + // create a really evil test case where a $x appears inside a binding of $x + // but *shouldnt* bind because it was inserted by a different macro.... + ]; + for (idx,s) in tests.iter().enumerate() { + run_renaming_test(s,idx); + } + } + + // run one of the renaming tests + fn run_renaming_test(t : &renaming_test, test_idx: uint) { + let invalid_name = token::special_idents::invalid.name; + let (teststr, bound_connections, bound_ident_check) = match *t { + (ref str,ref conns, bic) => (str.to_managed(), conns.clone(), bic) + }; + let cr = expand_crate_str(teststr.to_managed()); + // find the bindings: + let bindings = @mut ~[]; + visit::walk_crate(&mut new_name_finder(bindings),cr,()); + // find the varrefs: + let varrefs = @mut ~[]; + visit::walk_crate(&mut new_path_finder(varrefs),cr,()); + // must be one check clause for each binding: + assert_eq!(bindings.len(),bound_connections.len()); + for (binding_idx,shouldmatch) in bound_connections.iter().enumerate() { + let binding_name = mtwt_resolve(bindings[binding_idx]); + let binding_marks = mtwt_marksof(bindings[binding_idx].ctxt,invalid_name); + // shouldmatch can't name varrefs that don't exist: + assert!((shouldmatch.len() == 0) || + (varrefs.len() > *shouldmatch.iter().max().unwrap())); + for (idx,varref) in varrefs.iter().enumerate() { + if shouldmatch.contains(&idx) { + // it should be a path of length 1, and it should + // be free-identifier=? or bound-identifier=? to the given binding + assert_eq!(varref.segments.len(),1); + let varref_name = mtwt_resolve(varref.segments[0].identifier); + let varref_marks = mtwt_marksof(varref.segments[0].identifier.ctxt, + invalid_name); + if (!(varref_name==binding_name)){ + std::io::println("uh oh, should match but doesn't:"); + std::io::println(fmt!("varref: %?",varref)); + std::io::println(fmt!("binding: %?", bindings[binding_idx])); + ast_util::display_sctable(get_sctable()); + } + assert_eq!(varref_name,binding_name); + if (bound_ident_check) { + // we're checking bound-identifier=?, and the marks + // should be the same, too: + assert_eq!(varref_marks,binding_marks.clone()); + } + } else { + let fail = (varref.segments.len() == 1) + && (mtwt_resolve(varref.segments[0].identifier) == binding_name); + // temp debugging: + if (fail) { + println!("failure on test {}",test_idx); + println!("text of test case: \"{}\"", teststr); + println!(""); + println!("uh oh, matches but shouldn't:"); + std::io::println(fmt!("varref: %?",varref)); + // good lord, you can't make a path with 0 segments, can you? + println!("varref's first segment's uint: {}, and string: \"{}\"", + varref.segments[0].identifier.name, + ident_to_str(&varref.segments[0].identifier)); + std::io::println(fmt!("binding: %?", bindings[binding_idx])); + ast_util::display_sctable(get_sctable()); + } + assert!(!fail); + } + } + } + } + + #[test] fn fmt_in_macro_used_inside_module_macro() { + let crate_str = @"macro_rules! fmt_wrap(($b:expr)=>(fmt!(\"left: %?\", $b))) +macro_rules! foo_module (() => (mod generated { fn a() { let xx = 147; fmt_wrap!(xx);}})) +foo_module!() +"; + let cr = expand_crate_str(crate_str); + // find the xx binding + let bindings = @mut ~[]; + visit::walk_crate(&mut new_name_finder(bindings), cr, ()); + let cxbinds : ~[&ast::Ident] = + bindings.iter().filter(|b|{@"xx" == (ident_to_str(*b))}).collect(); + let cxbind = match cxbinds { + [b] => b, + _ => fail!("expected just one binding for ext_cx") + }; + let resolved_binding = mtwt_resolve(*cxbind); + // find all the xx varrefs: + let varrefs = @mut ~[]; + visit::walk_crate(&mut new_path_finder(varrefs), cr, ()); + // the xx binding should bind all of the xx varrefs: + for (idx,v) in varrefs.iter().filter(|p|{ p.segments.len() == 1 + && (@"xx" == (ident_to_str(&p.segments[0].identifier))) + }).enumerate() { + if (mtwt_resolve(v.segments[0].identifier) != resolved_binding) { + std::io::println("uh oh, xx binding didn't match xx varref:"); + std::io::println(fmt!("this is xx varref # %?",idx)); + std::io::println(fmt!("binding: %?",cxbind)); + std::io::println(fmt!("resolves to: %?",resolved_binding)); + std::io::println(fmt!("varref: %?",v.segments[0].identifier)); + std::io::println(fmt!("resolves to: %?",mtwt_resolve(v.segments[0].identifier))); + let table = get_sctable(); + std::io::println("SC table:"); + for (idx,val) in table.table.iter().enumerate() { + std::io::println(fmt!("%4u : %?",idx,val)); + } + } + assert_eq!(mtwt_resolve(v.segments[0].identifier),resolved_binding); + }; + } #[test] fn pat_idents(){ @@ -1363,4 +1911,5 @@ mod test { pat_idents.visit_pat(pat, ()); assert_eq!(idents, @mut strs_to_idents(~["a","c","b","d"])); } + } diff --git a/src/libsyntax/ext/quote.rs b/src/libsyntax/ext/quote.rs index 4e8b7467c5c..6527b083cc1 100644 --- a/src/libsyntax/ext/quote.rs +++ b/src/libsyntax/ext/quote.rs @@ -293,45 +293,51 @@ pub fn expand_quote_tokens(cx: @ExtCtxt, sp: Span, tts: &[ast::token_tree]) -> base::MacResult { let (cx_expr, expr) = expand_tts(cx, sp, tts); - base::MRExpr(expand_wrapper(cx, sp, cx_expr, expr)) + let expanded = expand_wrapper(cx, sp, cx_expr, expr); + base::MRExpr(expanded) } pub fn expand_quote_expr(cx: @ExtCtxt, sp: Span, tts: &[ast::token_tree]) -> base::MacResult { - base::MRExpr(expand_parse_call(cx, sp, "parse_expr", ~[], tts)) + let expanded = expand_parse_call(cx, sp, "parse_expr", ~[], tts); + base::MRExpr(expanded) } pub fn expand_quote_item(cx: @ExtCtxt, sp: Span, tts: &[ast::token_tree]) -> base::MacResult { let e_attrs = cx.expr_vec_uniq(sp, ~[]); - base::MRExpr(expand_parse_call(cx, sp, "parse_item", - ~[e_attrs], tts)) + let expanded = expand_parse_call(cx, sp, "parse_item", + ~[e_attrs], tts); + base::MRExpr(expanded) } pub fn expand_quote_pat(cx: @ExtCtxt, sp: Span, tts: &[ast::token_tree]) -> base::MacResult { let e_refutable = cx.expr_lit(sp, ast::lit_bool(true)); - base::MRExpr(expand_parse_call(cx, sp, "parse_pat", - ~[e_refutable], tts)) + let expanded = expand_parse_call(cx, sp, "parse_pat", + ~[e_refutable], tts); + base::MRExpr(expanded) } pub fn expand_quote_ty(cx: @ExtCtxt, sp: Span, tts: &[ast::token_tree]) -> base::MacResult { let e_param_colons = cx.expr_lit(sp, ast::lit_bool(false)); - base::MRExpr(expand_parse_call(cx, sp, "parse_ty", - ~[e_param_colons], tts)) + let expanded = expand_parse_call(cx, sp, "parse_ty", + ~[e_param_colons], tts); + base::MRExpr(expanded) } pub fn expand_quote_stmt(cx: @ExtCtxt, sp: Span, tts: &[ast::token_tree]) -> base::MacResult { let e_attrs = cx.expr_vec_uniq(sp, ~[]); - base::MRExpr(expand_parse_call(cx, sp, "parse_stmt", - ~[e_attrs], tts)) + let expanded = expand_parse_call(cx, sp, "parse_stmt", + ~[e_attrs], tts); + base::MRExpr(expanded) } fn ids_ext(strs: ~[~str]) -> ~[ast::Ident] { diff --git a/src/libsyntax/ext/tt/macro_parser.rs b/src/libsyntax/ext/tt/macro_parser.rs index 7a9e916421c..aa4183837e3 100644 --- a/src/libsyntax/ext/tt/macro_parser.rs +++ b/src/libsyntax/ext/tt/macro_parser.rs @@ -234,6 +234,15 @@ pub fn parse_or_else( } } +// perform a token equality check, ignoring syntax context (that is, an unhygienic comparison) +pub fn token_name_eq(t1 : &Token, t2 : &Token) -> bool { + match (t1,t2) { + (&token::IDENT(id1,_),&token::IDENT(id2,_)) => + id1.name == id2.name, + _ => *t1 == *t2 + } +} + pub fn parse( sess: @mut ParseSess, cfg: ast::CrateConfig, @@ -297,7 +306,10 @@ pub fn parse( // the *_t vars are workarounds for the lack of unary move match ei.sep { Some(ref t) if idx == len => { // we need a separator - if tok == (*t) { //pass the separator + // i'm conflicted about whether this should be hygienic.... + // though in this case, if the separators are never legal + // idents, it shouldn't matter. + if token_name_eq(&tok, t) { //pass the separator let mut ei_t = ei.clone(); ei_t.idx += 1; next_eis.push(ei_t); @@ -343,7 +355,8 @@ pub fn parse( match_nonterminal(_,_,_) => { bb_eis.push(ei) } match_tok(ref t) => { let mut ei_t = ei.clone(); - if (*t) == tok { + //if (token_name_eq(t,&tok)) { + if (token::mtwt_token_eq(t,&tok)) { ei_t.idx += 1; next_eis.push(ei_t); } @@ -353,7 +366,7 @@ pub fn parse( } /* error messages here could be improved with links to orig. rules */ - if tok == EOF { + if token_name_eq(&tok, &EOF) { if eof_eis.len() == 1u { let mut v = ~[]; for dv in eof_eis[0u].matches.mut_iter() { diff --git a/src/libsyntax/ext/tt/macro_rules.rs b/src/libsyntax/ext/tt/macro_rules.rs index 732ae2ccb96..74de8eaa09e 100644 --- a/src/libsyntax/ext/tt/macro_rules.rs +++ b/src/libsyntax/ext/tt/macro_rules.rs @@ -14,6 +14,7 @@ use ast; use codemap::{Span, Spanned, dummy_sp}; use ext::base::{ExtCtxt, MacResult, MRAny, MRDef, MacroDef, NormalTT}; use ext::base; +use ext::expand; use ext::tt::macro_parser::{error}; use ext::tt::macro_parser::{named_match, matched_seq, matched_nonterminal}; use ext::tt::macro_parser::{parse, parse_or_else, success, failure}; @@ -23,11 +24,17 @@ use parse::token::{get_ident_interner, special_idents, gensym_ident, ident_to_st use parse::token::{FAT_ARROW, SEMI, nt_matchers, nt_tt}; use print; +// this procedure performs the expansion of the +// macro_rules! macro. It parses the RHS and adds +// an extension to the current context. pub fn add_new_extension(cx: @ExtCtxt, sp: Span, name: Ident, - arg: ~[ast::token_tree]) + arg: ~[ast::token_tree], + stx_ctxt: ast::SyntaxContext) -> base::MacResult { + let arg = expand::mtwt_cancel_outer_mark(arg,stx_ctxt); + // Wrap a matcher_ in a spanned to produce a matcher. // these spans won't matter, anyways fn ms(m: matcher_) -> matcher { Spanned { @@ -39,6 +46,7 @@ pub fn add_new_extension(cx: @ExtCtxt, let lhs_nm = gensym_ident("lhs"); let rhs_nm = gensym_ident("rhs"); + // The pattern that macro_rules matches. // The grammar for macro_rules! is: // $( $lhs:mtcs => $rhs:tt );+ // ...quasiquoting this would be nice. @@ -144,11 +152,11 @@ pub fn add_new_extension(cx: @ExtCtxt, cx.span_fatal(best_fail_spot, best_fail_msg); } - let exp: @fn(@ExtCtxt, Span, &[ast::token_tree]) -> MacResult = - |cx, sp, arg| generic_extension(cx, sp, name, arg, *lhses, *rhses); + let exp: @fn(@ExtCtxt, Span, &[ast::token_tree], ctxt: ast::SyntaxContext) -> MacResult = + |cx, sp, arg, _ctxt| generic_extension(cx, sp, name, arg, *lhses, *rhses); return MRDef(MacroDef{ name: ident_to_str(&name), - ext: NormalTT(base::SyntaxExpanderTT{expander: exp, span: Some(sp)}) + ext: NormalTT(exp, Some(sp)) }); } diff --git a/src/libsyntax/fold.rs b/src/libsyntax/fold.rs index 7aa0f3abe87..aad992706fd 100644 --- a/src/libsyntax/fold.rs +++ b/src/libsyntax/fold.rs @@ -14,6 +14,14 @@ use codemap::{Span, Spanned}; use parse::token; use opt_vec::OptVec; +// this file defines an ast_fold trait for objects that can perform +// a "fold" on Rust ASTs. It also contains a structure that implements +// that trait, and a "default_fold" whose fields contain closures +// that perform "default traversals", visiting all of the sub-elements +// and re-assembling the result. The "fun_to_ident_folder" in the +// test module provides a simple example of creating a very simple +// fold that only looks at identifiers. + pub trait ast_fold { fn fold_crate(@self, &Crate) -> Crate; fn fold_view_item(@self, &view_item) -> view_item; @@ -35,6 +43,7 @@ pub trait ast_fold { fn fold_ident(@self, Ident) -> Ident; fn fold_path(@self, &Path) -> Path; fn fold_local(@self, @Local) -> @Local; + fn fold_mac(@self, &mac) -> mac; fn map_exprs(@self, @fn(@Expr) -> @Expr, &[@Expr]) -> ~[@Expr]; fn new_id(@self, NodeId) -> NodeId; fn new_span(@self, Span) -> Span; @@ -64,6 +73,7 @@ pub struct AstFoldFns { fold_ident: @fn(Ident, @ast_fold) -> Ident, fold_path: @fn(&Path, @ast_fold) -> Path, fold_local: @fn(@Local, @ast_fold) -> @Local, + fold_mac: @fn(&mac_, Span, @ast_fold) -> (mac_, Span), map_exprs: @fn(@fn(@Expr) -> @Expr, &[@Expr]) -> ~[@Expr], new_id: @fn(NodeId) -> NodeId, new_span: @fn(Span) -> Span @@ -112,41 +122,31 @@ fn fold_arg_(a: arg, fld: @ast_fold) -> arg { } } -//used in noop_fold_expr, and possibly elsewhere in the future -fn fold_mac_(m: &mac, fld: @ast_fold) -> mac { - Spanned { - node: match m.node { - mac_invoc_tt(ref p,ref tts) => - mac_invoc_tt(fld.fold_path(p), - fold_tts(*tts,fld)) - }, - span: fld.new_span(m.span) - } -} - -fn fold_tts(tts : &[token_tree], fld: @ast_fold) -> ~[token_tree] { +// build a new vector of tts by appling the ast_fold's fold_ident to +// all of the identifiers in the token trees. +pub fn fold_tts(tts : &[token_tree], f : @ast_fold) -> ~[token_tree] { do tts.map |tt| { match *tt { tt_tok(span, ref tok) => - tt_tok(span,maybe_fold_ident(tok,fld)), + tt_tok(span,maybe_fold_ident(tok,f)), tt_delim(ref tts) => - tt_delim(@mut fold_tts(**tts, fld)), + tt_delim(@mut fold_tts(**tts, f)), tt_seq(span, ref pattern, ref sep, is_optional) => tt_seq(span, - @mut fold_tts(**pattern, fld), - sep.map(|tok|maybe_fold_ident(tok,fld)), + @mut fold_tts(**pattern, f), + sep.map(|tok|maybe_fold_ident(tok,f)), is_optional), tt_nonterminal(sp,ref ident) => - tt_nonterminal(sp,fld.fold_ident(*ident)) + tt_nonterminal(sp,f.fold_ident(*ident)) } } } // apply ident folder if it's an ident, otherwise leave it alone -fn maybe_fold_ident(t: &token::Token, fld: @ast_fold) -> token::Token { +fn maybe_fold_ident(t : &token::Token, f: @ast_fold) -> token::Token { match *t { token::IDENT(id,followed_by_colons) => - token::IDENT(fld.fold_ident(id),followed_by_colons), + token::IDENT(f.fold_ident(id),followed_by_colons), _ => (*t).clone() } } @@ -209,6 +209,7 @@ pub fn noop_fold_crate(c: &Crate, fld: @ast_fold) -> Crate { } fn noop_fold_view_item(vi: &view_item_, _fld: @ast_fold) -> view_item_ { + // FIXME #7654: doesn't iterate over idents in a view_item_use return /* FIXME (#2543) */ (*vi).clone(); } @@ -323,11 +324,7 @@ pub fn noop_fold_item_underscore(i: &item_, fld: @ast_fold) -> item_ { ) } item_mac(ref m) => { - // FIXME #2888: we might actually want to do something here. - // ... okay, we're doing something. It would probably be nicer - // to add something to the ast_fold trait, but I'll defer - // that work. - item_mac(fold_mac_(m,fld)) + item_mac(fld.fold_mac(m)) } } } @@ -396,7 +393,6 @@ pub fn noop_fold_block(b: &Block, fld: @ast_fold) -> Block { } fn noop_fold_stmt(s: &Stmt_, fld: @ast_fold) -> Option<Stmt_> { - let fold_mac = |x| fold_mac_(x, fld); match *s { StmtDecl(d, nid) => { match fld.fold_decl(d) { @@ -410,7 +406,7 @@ fn noop_fold_stmt(s: &Stmt_, fld: @ast_fold) -> Option<Stmt_> { StmtSemi(e, nid) => { Some(StmtSemi(fld.fold_expr(e), fld.new_id(nid))) } - StmtMac(ref mac, semi) => Some(StmtMac(fold_mac(mac), semi)) + StmtMac(ref mac, semi) => Some(StmtMac(fld.fold_mac(mac), semi)) } } @@ -478,6 +474,12 @@ fn noop_fold_decl(d: &Decl_, fld: @ast_fold) -> Option<Decl_> { } } +// lift a function in ast-thingy X fold -> ast-thingy to a function +// in (ast-thingy X span X fold) -> (ast-thingy X span). Basically, +// carries the span around. +// It seems strange to me that the call to new_fold doesn't happen +// here but instead in the impl down below.... probably just an +// accident? pub fn wrap<T>(f: @fn(&T, @ast_fold) -> T) -> @fn(&T, Span, @ast_fold) -> (T, Span) { let result: @fn(&T, Span, @ast_fold) -> (T, Span) = |x, s, fld| { @@ -496,8 +498,6 @@ pub fn noop_fold_expr(e: &Expr_, fld: @ast_fold) -> Expr_ { } let fold_field = |x| fold_field_(x, fld); - let fold_mac = |x| fold_mac_(x, fld); - match *e { ExprVstore(e, v) => { ExprVstore(fld.fold_expr(e), v) @@ -544,7 +544,7 @@ pub fn noop_fold_expr(e: &Expr_, fld: @ast_fold) -> Expr_ { ExprDoBody(f) => ExprDoBody(fld.fold_expr(f)), ExprLit(_) => (*e).clone(), ExprCast(expr, ref ty) => { - ExprCast(fld.fold_expr(expr), (*ty).clone()) + ExprCast(fld.fold_expr(expr), fld.fold_ty(ty)) } ExprAddrOf(m, ohs) => ExprAddrOf(m, fld.fold_expr(ohs)), ExprIf(cond, ref tr, fl) => { @@ -629,7 +629,7 @@ pub fn noop_fold_expr(e: &Expr_, fld: @ast_fold) -> Expr_ { .. (*a).clone() }) } - ExprMac(ref mac) => ExprMac(fold_mac(mac)), + ExprMac(ref mac) => ExprMac(fld.fold_mac(mac)), ExprStruct(ref path, ref fields, maybe_expr) => { ExprStruct( fld.fold_path(path), @@ -642,7 +642,6 @@ pub fn noop_fold_expr(e: &Expr_, fld: @ast_fold) -> Expr_ { } pub fn noop_fold_ty(t: &ty_, fld: @ast_fold) -> ty_ { - let fold_mac = |x| fold_mac_(x, fld); fn fold_mt(mt: &mt, fld: @ast_fold) -> mt { mt { ty: ~fld.fold_ty(mt.ty), @@ -698,7 +697,7 @@ pub fn noop_fold_ty(t: &ty_, fld: @ast_fold) -> ty_ { ) } ty_typeof(e) => ty_typeof(fld.fold_expr(e)), - ty_mac(ref mac) => ty_mac(fold_mac(mac)) + ty_mac(ref mac) => ty_mac(fld.fold_mac(mac)) } } @@ -785,6 +784,19 @@ fn noop_fold_local(l: @Local, fld: @ast_fold) -> @Local { } } +// the default macro traversal. visit the path +// using fold_path, and the tts using fold_tts, +// and the span using new_span +fn noop_fold_mac(m: &mac_, fld: @ast_fold) -> mac_ { + match *m { + mac_invoc_tt(ref p,ref tts,ctxt) => + mac_invoc_tt(fld.fold_path(p), + fold_tts(*tts,fld), + ctxt) + } +} + + /* temporarily eta-expand because of a compiler bug with using `fn<T>` as a value */ fn noop_map_exprs(f: @fn(@Expr) -> @Expr, es: &[@Expr]) -> ~[@Expr] { @@ -817,6 +829,7 @@ pub fn default_ast_fold() -> ast_fold_fns { fold_ident: noop_fold_ident, fold_path: noop_fold_path, fold_local: noop_fold_local, + fold_mac: wrap(noop_fold_mac), map_exprs: noop_map_exprs, new_id: noop_id, new_span: noop_span, @@ -922,6 +935,10 @@ impl ast_fold for AstFoldFns { fn fold_local(@self, x: @Local) -> @Local { (self.fold_local)(x, self as @ast_fold) } + fn fold_mac(@self, x: &mac) -> mac { + let (n, s) = (self.fold_mac)(&x.node, x.span, self as @ast_fold); + Spanned { node: n, span: (self.new_span)(s) } + } fn map_exprs(@self, f: @fn(@Expr) -> @Expr, e: &[@Expr]) @@ -946,6 +963,8 @@ impl AstFoldExtensions for @ast_fold { } } +// brson agrees with me that this function's existence is probably +// not a good or useful thing. pub fn make_fold(afp: ast_fold_fns) -> @ast_fold { afp as @ast_fold } @@ -1018,4 +1037,15 @@ mod test { token::get_ident_interner()), ~"zz!zz((zz$zz:zz$(zz $zz:zz)zz+=>(zz$(zz$zz$zz)+)))"); } + + // and in cast expressions... this appears to be an existing bug. + #[test] fn ident_transformation_in_types () { + let zz_fold = fun_to_ident_folder(to_zz()); + let ast = string_to_crate(@"fn a() {let z = 13 as int;}"); + assert_pred!(matches_codepattern, + "matches_codepattern", + pprust::to_str(&zz_fold.fold_crate(ast),fake_print_crate, + token::get_ident_interner()), + ~"fn zz(){let zz=13 as zz;}"); + } } diff --git a/src/libsyntax/parse/parser.rs b/src/libsyntax/parse/parser.rs index 51c5522ae2f..8b11a25f13c 100644 --- a/src/libsyntax/parse/parser.rs +++ b/src/libsyntax/parse/parser.rs @@ -21,7 +21,7 @@ use ast::{_mod, BiAdd, arg, Arm, Attribute, BindByRef, BindInfer}; use ast::{BiBitAnd, BiBitOr, BiBitXor, Block}; use ast::{BlockCheckMode, UnBox}; use ast::{Crate, CrateConfig, Decl, DeclItem}; -use ast::{DeclLocal, DefaultBlock, UnDeref, BiDiv, enum_def, explicit_self}; +use ast::{DeclLocal, DefaultBlock, UnDeref, BiDiv, EMPTY_CTXT, enum_def, explicit_self}; use ast::{Expr, Expr_, ExprAddrOf, ExprMatch, ExprAgain}; use ast::{ExprAssign, ExprAssignOp, ExprBinary, ExprBlock}; use ast::{ExprBreak, ExprCall, ExprCast, ExprDoBody}; @@ -1875,7 +1875,7 @@ impl Parser { |p| p.parse_token_tree()); let hi = self.span.hi; - return self.mk_mac_expr(lo, hi, mac_invoc_tt(pth, tts)); + return self.mk_mac_expr(lo, hi, mac_invoc_tt(pth, tts, EMPTY_CTXT)); } else if *self.token == token::LBRACE { // This might be a struct literal. if self.looking_at_record_literal() { @@ -3197,14 +3197,14 @@ impl Parser { if id == token::special_idents::invalid { return @spanned(lo, hi, StmtMac( - spanned(lo, hi, mac_invoc_tt(pth, tts)), false)); + spanned(lo, hi, mac_invoc_tt(pth, tts, EMPTY_CTXT)), false)); } else { // if it has a special ident, it's definitely an item return @spanned(lo, hi, StmtDecl( @spanned(lo, hi, DeclItem( self.mk_item( lo, hi, id /*id is good here*/, - item_mac(spanned(lo, hi, mac_invoc_tt(pth, tts))), + item_mac(spanned(lo, hi, mac_invoc_tt(pth, tts, EMPTY_CTXT))), inherited, ~[/*no attrs*/]))), self.get_id())); } @@ -3518,7 +3518,10 @@ impl Parser { } fn is_self_ident(&self) -> bool { - *self.token == token::IDENT(special_idents::self_, false) + match *self.token { + token::IDENT(id, false) => id.name == special_idents::self_.name, + _ => false + } } fn expect_self_ident(&self) { @@ -4806,7 +4809,7 @@ impl Parser { _ => self.fatal("expected open delimiter") }; // single-variant-enum... : - let m = ast::mac_invoc_tt(pth, tts); + let m = ast::mac_invoc_tt(pth, tts, EMPTY_CTXT); let m: ast::mac = codemap::Spanned { node: m, span: mk_sp(self.span.lo, self.span.hi) }; diff --git a/src/libsyntax/parse/token.rs b/src/libsyntax/parse/token.rs index 591b4b10bd3..6b3a95a14f8 100644 --- a/src/libsyntax/parse/token.rs +++ b/src/libsyntax/parse/token.rs @@ -9,17 +9,15 @@ // except according to those terms. use ast; -use ast::Name; +use ast::{Name, Mrk}; use ast_util; use parse::token; use util::interner::StrInterner; use util::interner; +use std::cast; use std::char; -use std::cmp::Equiv; use std::local_data; -use std::rand; -use std::rand::RngUtil; #[deriving(Clone, Encodable, Decodable, Eq, IterBytes)] pub enum binop { @@ -309,22 +307,23 @@ pub fn is_bar(t: &Token) -> bool { match *t { BINOP(OR) | OROR => true, _ => false } } - pub mod special_idents { use ast::Ident; - pub static underscore : Ident = Ident { name: 0, ctxt: 0}; + pub static underscore : Ident = Ident { name: 0, ctxt: 0}; // apparently unused? pub static anon : Ident = Ident { name: 1, ctxt: 0}; pub static invalid : Ident = Ident { name: 2, ctxt: 0}; // '' - pub static unary : Ident = Ident { name: 3, ctxt: 0}; - pub static not_fn : Ident = Ident { name: 4, ctxt: 0}; - pub static idx_fn : Ident = Ident { name: 5, ctxt: 0}; - pub static unary_minus_fn : Ident = Ident { name: 6, ctxt: 0}; + pub static unary : Ident = Ident { name: 3, ctxt: 0}; // apparently unused? + pub static not_fn : Ident = Ident { name: 4, ctxt: 0}; // apparently unused? + pub static idx_fn : Ident = Ident { name: 5, ctxt: 0}; // apparently unused? + pub static unary_minus_fn : Ident = Ident { name: 6, ctxt: 0}; // apparently unused? pub static clownshoes_extensions : Ident = Ident { name: 7, ctxt: 0}; pub static self_ : Ident = Ident { name: 8, ctxt: 0}; // 'self' /* for matcher NTs */ + // none of these appear to be used, but perhaps references to + // these are artificially fabricated by the macro system.... pub static item : Ident = Ident { name: 9, ctxt: 0}; pub static block : Ident = Ident { name: 10, ctxt: 0}; pub static stmt : Ident = Ident { name: 11, ctxt: 0}; @@ -336,7 +335,7 @@ pub mod special_idents { pub static tt : Ident = Ident { name: 17, ctxt: 0}; pub static matchers : Ident = Ident { name: 18, ctxt: 0}; - pub static str : Ident = Ident { name: 19, ctxt: 0}; // for the type + pub static str : Ident = Ident { name: 19, ctxt: 0}; // for the type // apparently unused? /* outside of libsyntax */ pub static arg : Ident = Ident { name: 20, ctxt: 0}; @@ -349,10 +348,32 @@ pub mod special_idents { pub static statik : Ident = Ident { name: 27, ctxt: 0}; pub static clownshoes_foreign_mod: Ident = Ident { name: 28, ctxt: 0}; pub static unnamed_field: Ident = Ident { name: 29, ctxt: 0}; - pub static c_abi: Ident = Ident { name: 30, ctxt: 0}; + pub static c_abi: Ident = Ident { name: 30, ctxt: 0}; // apparently unused? pub static type_self: Ident = Ident { name: 31, ctxt: 0}; // `Self` } +// here are the ones that actually occur in the source. Maybe the rest +// should be removed? +/* +special_idents::anon +special_idents::arg +special_idents::blk +special_idents::clownshoe_abi +special_idents::clownshoe_stack_shim +special_idents::clownshoes_extensions +special_idents::clownshoes_foreign_mod +special_idents::descrim +special_idents::invalid +special_idents::main +special_idents::matchers +special_idents::opaque +special_idents::self_ +special_idents::statik +special_idents::tt +special_idents::type_self +special_idents::unnamed_field +*/ + /** * Maps a token to a record specifying the corresponding binary * operator @@ -381,30 +402,8 @@ pub fn token_to_binop(tok: &Token) -> Option<ast::BinOp> { } } -pub struct ident_interner { - priv interner: StrInterner, -} - -impl ident_interner { - pub fn intern(&self, val: &str) -> Name { - self.interner.intern(val) - } - pub fn gensym(&self, val: &str) -> Name { - self.interner.gensym(val) - } - pub fn get(&self, idx: Name) -> @str { - self.interner.get(idx) - } - // is this really something that should be exposed? - pub fn len(&self) -> uint { - self.interner.len() - } - pub fn find_equiv<Q:Hash + IterBytes + Equiv<@str>>(&self, val: &Q) - -> Option<Name> { - self.interner.find_equiv(val) - } -} - +// looks like we can get rid of this completely... +pub type ident_interner = StrInterner; // return a fresh interner, preloaded with special identifiers. fn mk_fresh_ident_interner() -> @ident_interner { @@ -485,9 +484,7 @@ fn mk_fresh_ident_interner() -> @ident_interner { "typeof", // 67 ]; - @ident_interner { - interner: interner::StrInterner::prefill(init_vec) - } + @interner::StrInterner::prefill(init_vec) } // if an interner exists in TLS, return it. Otherwise, prepare a @@ -508,7 +505,7 @@ pub fn get_ident_interner() -> @ident_interner { /* for when we don't care about the contents; doesn't interact with TLD or serialization */ pub fn mk_fake_ident_interner() -> @ident_interner { - @ident_interner { interner: interner::StrInterner::new() } + @interner::StrInterner::new() } // maps a string to its interned representation @@ -543,18 +540,44 @@ pub fn gensym_ident(str : &str) -> ast::Ident { ast::Ident::new(gensym(str)) } +// create a fresh name that maps to the same string as the old one. +// note that this guarantees that str_ptr_eq(ident_to_str(src),interner_get(fresh_name(src))); +// that is, that the new name and the old one are connected to ptr_eq strings. +pub fn fresh_name(src : &ast::Ident) -> Name { + let interner = get_ident_interner(); + interner.gensym_copy(src.name) + // following: debug version. Could work in final except that it's incompatible with + // good error messages and uses of struct names in ambiguous could-be-binding + // locations. Also definitely destroys the guarantee given above about ptr_eq. + /*let num = rand::rng().gen_uint_range(0,0xffff); + gensym(fmt!("%s_%u",ident_to_str(src),num))*/ +} + +// it looks like there oughta be a str_ptr_eq fn, but no one bothered to implement it? + +// determine whether two @str values are pointer-equal +pub fn str_ptr_eq(a : @str, b : @str) -> bool { + unsafe { + let p : uint = cast::transmute(a); + let q : uint = cast::transmute(b); + let result = p == q; + // got to transmute them back, to make sure the ref count is correct: + let _junk1 : @str = cast::transmute(p); + let _junk2 : @str = cast::transmute(q); + result + } +} + +// return true when two identifiers refer (through the intern table) to the same ptr_eq +// string. This is used to compare identifiers in places where hygienic comparison is +// not wanted (i.e. not lexical vars). +pub fn ident_spelling_eq(a : &ast::Ident, b : &ast::Ident) -> bool { + str_ptr_eq(interner_get(a.name),interner_get(b.name)) +} -// create a fresh name. In principle, this is just a -// gensym, but for debugging purposes, you'd like the -// resulting name to have a suggestive stringify, without -// paying the cost of guaranteeing that the name is -// truly unique. I'm going to try to strike a balance -// by using a gensym with a name that has a random number -// at the end. So, the gensym guarantees the uniqueness, -// and the int helps to avoid confusion. -pub fn fresh_name(src_name : &str) -> Name { - let num = rand::rng().gen_uint_range(0,0xffff); - gensym(fmt!("%s_%u",src_name,num)) +// create a fresh mark. +pub fn fresh_mark() -> Mrk { + gensym("mark") } /** @@ -694,12 +717,48 @@ pub fn is_reserved_keyword(tok: &Token) -> bool { } } +pub fn mtwt_token_eq(t1 : &Token, t2 : &Token) -> bool { + match (t1,t2) { + (&IDENT(id1,_),&IDENT(id2,_)) => + ast_util::mtwt_resolve(id1) == ast_util::mtwt_resolve(id2), + _ => *t1 == *t2 + } +} + + #[cfg(test)] mod test { use super::*; - #[test] fn t1() { - let a = fresh_name("ghi"); - printfln!("interned name: %u,\ntextual name: %s\n", - a, interner_get(a)); + use ast; + use ast_util; + + fn mark_ident(id : ast::Ident, m : ast::Mrk) -> ast::Ident { + ast::Ident{name:id.name,ctxt:ast_util::new_mark(m,id.ctxt)} } + + #[test] fn mtwt_token_eq_test() { + assert!(mtwt_token_eq(>,>)); + let a = str_to_ident("bac"); + let a1 = mark_ident(a,92); + assert!(mtwt_token_eq(&IDENT(a,true),&IDENT(a1,false))); + } + + + #[test] fn str_ptr_eq_tests(){ + let a = @"abc"; + let b = @"abc"; + let c = a; + assert!(str_ptr_eq(a,c)); + assert!(!str_ptr_eq(a,b)); + } + + #[test] fn fresh_name_pointer_sharing() { + let ghi = str_to_ident("ghi"); + assert_eq!(ident_to_str(&ghi),@"ghi"); + assert!(str_ptr_eq(ident_to_str(&ghi),ident_to_str(&ghi))) + let fresh = ast::Ident::new(fresh_name(&ghi)); + assert_eq!(ident_to_str(&fresh),@"ghi"); + assert!(str_ptr_eq(ident_to_str(&ghi),ident_to_str(&fresh))); + } + } diff --git a/src/libsyntax/print/pprust.rs b/src/libsyntax/print/pprust.rs index 32cf30fd3a0..f440e0a1771 100644 --- a/src/libsyntax/print/pprust.rs +++ b/src/libsyntax/print/pprust.rs @@ -619,7 +619,8 @@ pub fn print_item(s: @ps, item: &ast::item) { } bclose(s, item.span); } - ast::item_mac(codemap::Spanned { node: ast::mac_invoc_tt(ref pth, ref tts), + // I think it's reasonable to hide the context here: + ast::item_mac(codemap::Spanned { node: ast::mac_invoc_tt(ref pth, ref tts, _), _}) => { print_visibility(s, item.vis); print_path(s, pth, false); @@ -1021,7 +1022,8 @@ pub fn print_if(s: @ps, test: &ast::Expr, blk: &ast::Block, pub fn print_mac(s: @ps, m: &ast::mac) { match m.node { - ast::mac_invoc_tt(ref pth, ref tts) => { + // I think it's reasonable to hide the ctxt here: + ast::mac_invoc_tt(ref pth, ref tts, _) => { print_path(s, pth, false); word(s.s, "!"); popen(s); @@ -1924,8 +1926,8 @@ pub fn print_arg(s: @ps, input: &ast::arg) { match input.pat.node { ast::PatIdent(_, ref path, _) if path.segments.len() == 1 && - path.segments[0].identifier == - parse::token::special_idents::invalid => { + path.segments[0].identifier.name == + parse::token::special_idents::invalid.name => { // Do nothing. } _ => { diff --git a/src/libsyntax/util/interner.rs b/src/libsyntax/util/interner.rs index 46676ce1093..2b1e7eaa9b2 100644 --- a/src/libsyntax/util/interner.rs +++ b/src/libsyntax/util/interner.rs @@ -117,6 +117,23 @@ impl StrInterner { new_idx } + // I want these gensyms to share name pointers + // with existing entries. This would be automatic, + // except that the existing gensym creates its + // own managed ptr using to_managed. I think that + // adding this utility function is the most + // lightweight way to get what I want, though not + // necessarily the cleanest. + + // create a gensym with the same name as an existing + // entry. + pub fn gensym_copy(&self, idx : uint) -> uint { + let new_idx = self.len(); + // leave out of map to avoid colliding + self.vect.push(self.vect[idx]); + new_idx + } + // this isn't "pure" in the traditional sense, because it can go from // failing to returning a value as items are interned. But for typestate, // where we first check a pred and then rely on it, ceasing to fail is ok. @@ -144,23 +161,23 @@ mod tests { } #[test] - fn i2 () { + fn interner_tests () { let i : Interner<@str> = Interner::new(); // first one is zero: - assert_eq!(i.intern (@"dog"), 0); + assert_eq!(i.intern(@"dog"), 0); // re-use gets the same entry: - assert_eq!(i.intern (@"dog"), 0); + assert_eq!(i.intern(@"dog"), 0); // different string gets a different #: - assert_eq!(i.intern (@"cat"), 1); - assert_eq!(i.intern (@"cat"), 1); + assert_eq!(i.intern(@"cat"), 1); + assert_eq!(i.intern(@"cat"), 1); // dog is still at zero - assert_eq!(i.intern (@"dog"), 0); + assert_eq!(i.intern(@"dog"), 0); // gensym gets 3 - assert_eq!(i.gensym (@"zebra" ), 2); + assert_eq!(i.gensym(@"zebra" ), 2); // gensym of same string gets new number : assert_eq!(i.gensym (@"zebra" ), 3); // gensym of *existing* string gets new number: - assert_eq!(i.gensym (@"dog"), 4); + assert_eq!(i.gensym(@"dog"), 4); assert_eq!(i.get(0), @"dog"); assert_eq!(i.get(1), @"cat"); assert_eq!(i.get(2), @"zebra"); @@ -176,4 +193,34 @@ mod tests { assert_eq!(i.get(2), @"Carol"); assert_eq!(i.intern(@"Bob"), 1); } + + #[test] + fn string_interner_tests() { + let i : StrInterner = StrInterner::new(); + // first one is zero: + assert_eq!(i.intern("dog"), 0); + // re-use gets the same entry: + assert_eq!(i.intern ("dog"), 0); + // different string gets a different #: + assert_eq!(i.intern("cat"), 1); + assert_eq!(i.intern("cat"), 1); + // dog is still at zero + assert_eq!(i.intern("dog"), 0); + // gensym gets 3 + assert_eq!(i.gensym("zebra"), 2); + // gensym of same string gets new number : + assert_eq!(i.gensym("zebra"), 3); + // gensym of *existing* string gets new number: + assert_eq!(i.gensym("dog"), 4); + // gensym tests again with gensym_copy: + assert_eq!(i.gensym_copy(2), 5); + assert_eq!(i.get(5), @"zebra"); + assert_eq!(i.gensym_copy(2), 6); + assert_eq!(i.get(6), @"zebra"); + assert_eq!(i.get(0), @"dog"); + assert_eq!(i.get(1), @"cat"); + assert_eq!(i.get(2), @"zebra"); + assert_eq!(i.get(3), @"zebra"); + assert_eq!(i.get(4), @"dog"); + } } diff --git a/src/libsyntax/util/parser_testing.rs b/src/libsyntax/util/parser_testing.rs index ca1e53f7fcd..23396c06a9a 100644 --- a/src/libsyntax/util/parser_testing.rs +++ b/src/libsyntax/util/parser_testing.rs @@ -22,6 +22,12 @@ pub fn string_to_tts_and_sess (source_str : @str) -> (~[ast::token_tree],@mut Pa (filemap_to_tts(ps,string_to_filemap(ps,source_str,@"bogofile")),ps) } +// map a string to tts, using a made-up filename: +pub fn string_to_tts(source_str : @str) -> ~[ast::token_tree] { + let (tts,_) = string_to_tts_and_sess(source_str); + tts +} + pub fn string_to_parser_and_sess(source_str: @str) -> (Parser,@mut ParseSess) { let ps = new_parse_sess(None); (new_parser_from_source_str(ps,~[],@"bogofile",source_str),ps) @@ -40,12 +46,19 @@ fn with_error_checking_parse<T>(s: @str, f: &fn(&mut Parser) -> T) -> T { x } +// parse a string, return a crate. pub fn string_to_crate (source_str : @str) -> @ast::Crate { do with_error_checking_parse(source_str) |p| { p.parse_crate_mod() } } +// parse a string, return a crate and the ParseSess +pub fn string_to_crate_and_sess (source_str : @str) -> (@ast::Crate,@mut ParseSess) { + let (p,ps) = string_to_parser_and_sess(source_str); + (p.parse_crate_mod(),ps) +} + // parse a string, return an expr pub fn string_to_expr (source_str : @str) -> @ast::Expr { do with_error_checking_parse(source_str) |p| { @@ -60,14 +73,6 @@ pub fn string_to_item (source_str : @str) -> Option<@ast::item> { } } -// parse a string, return an item and the ParseSess -pub fn string_to_item_and_sess (source_str : @str) -> (Option<@ast::item>,@mut ParseSess) { - let (p,ps) = string_to_parser_and_sess(source_str); - let io = p.parse_item(~[]); - p.abort_if_errors(); - (io,ps) -} - // parse a string, return a stmt pub fn string_to_stmt(source_str : @str) -> @ast::Stmt { do with_error_checking_parse(source_str) |p| { |
