rust/src/libsyntax/ext/expand.rs

1601 lines
62 KiB
Rust
Raw Normal View History

// Copyright 2012-2014 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
2014-05-19 20:59:35 +00:00
use ast::{P, Block, Crate, DeclLocal, ExprMac, PatMac};
use ast::{Local, Ident, MacInvocTT};
use ast::{ItemMac, Mrk, Stmt, StmtDecl, StmtMac, StmtExpr, StmtSemi};
use ast::TokenTree;
use ast;
use ext::mtwt;
use ext::build::AstBuilder;
2013-02-11 16:36:42 +00:00
use attr;
use attr::AttrMetaMethods;
use codemap;
use codemap::{Span, Spanned, ExpnInfo, NameAndSpan, MacroBang, MacroAttribute};
log: Introduce liblog, the old std::logging This commit moves all logging out of the standard library into an external crate. This crate is the new crate which is responsible for all logging macros and logging implementation. A few reasons for this change are: * The crate map has always been a bit of a code smell among rust programs. It has difficulty being loaded on almost all platforms, and it's used almost exclusively for logging and only logging. Removing the crate map is one of the end goals of this movement. * The compiler has a fair bit of special support for logging. It has the __log_level() expression as well as generating a global word per module specifying the log level. This is unfairly favoring the built-in logging system, and is much better done purely in libraries instead of the compiler itself. * Initialization of logging is much easier to do if there is no reliance on a magical crate map being available to set module log levels. * If the logging library can be written outside of the standard library, there's no reason that it shouldn't be. It's likely that we're not going to build the highest quality logging library of all time, so third-party libraries should be able to provide just as high-quality logging systems as the default one provided in the rust distribution. With a migration such as this, the change does not come for free. There are some subtle changes in the behavior of liblog vs the previous logging macros: * The core change of this migration is that there is no longer a physical log-level per module. This concept is still emulated (it is quite useful), but there is now only a global log level, not a local one. This global log level is a reflection of the maximum of all log levels specified. The previously generated logging code looked like: if specified_level <= __module_log_level() { println!(...) } The newly generated code looks like: if specified_level <= ::log::LOG_LEVEL { if ::log::module_enabled(module_path!()) { println!(...) } } Notably, the first layer of checking is still intended to be "super fast" in that it's just a load of a global word and a compare. The second layer of checking is executed to determine if the current module does indeed have logging turned on. This means that if any module has a debug log level turned on, all modules with debug log levels get a little bit slower (they all do more expensive dynamic checks to determine if they're turned on or not). Semantically, this migration brings no change in this respect, but runtime-wise, this will have a perf impact on some code. * A `RUST_LOG=::help` directive will no longer print out a list of all modules that can be logged. This is because the crate map will no longer specify the log levels of all modules, so the list of modules is not known. Additionally, warnings can no longer be provided if a malformed logging directive was supplied. The new "hello world" for logging looks like: #[phase(syntax, link)] extern crate log; fn main() { debug!("Hello, world!"); }
2014-03-09 06:11:44 +00:00
use crateid::CrateId;
2012-09-04 18:37:29 +00:00
use ext::base::*;
2014-07-03 06:16:01 +00:00
use fold;
use fold::*;
use parse;
use parse::token::{fresh_mark, fresh_name, intern};
use parse::token;
2013-06-04 21:56:33 +00:00
use visit;
2013-06-08 16:21:11 +00:00
use visit::Visitor;
use util::small_vector::SmallVector;
2011-08-05 20:06:11 +00:00
use std::gc::{Gc, GC};
2014-05-16 07:16:13 +00:00
2014-06-25 22:20:01 +00:00
2014-05-16 07:16:13 +00:00
pub fn expand_expr(e: Gc<ast::Expr>, fld: &mut MacroExpander) -> Gc<ast::Expr> {
match e.node {
// expr_mac should really be expr_ext or something; it's the
// entry-point for all syntax extensions.
ExprMac(ref mac) => {
match (*mac).node {
2013-07-08 22:55:14 +00:00
// 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:
MacInvocTT(ref pth, ref tts, _) => {
2014-01-19 08:21:14 +00:00
if pth.segments.len() > 1u {
fld.cx.span_err(pth.span,
"expected macro name without module \
separators");
// let compilation continue
return DummyResult::raw_expr(e.span);
}
let extname = pth.segments.get(0).identifier;
let extnamestr = token::get_ident(extname);
let marked_after = match fld.extsbox.find(&extname.name) {
None => {
fld.cx.span_err(
pth.span,
2014-01-30 23:53:09 +00:00
format!("macro undefined: '{}'",
extnamestr.get()).as_slice());
// let compilation continue
return DummyResult::raw_expr(e.span);
}
Some(&NormalTT(ref expandfun, exp_span)) => {
fld.cx.bt_push(ExpnInfo {
call_site: e.span,
callee: NameAndSpan {
name: extnamestr.get().to_string(),
format: MacroBang,
span: exp_span,
},
});
let fm = fresh_mark();
// mark before:
let marked_before = mark_tts(tts.as_slice(), fm);
// The span that we pass to the expanders we want to
// be the root of the call stack. That's the most
// relevant span and it's the actual invocation of
// the macro.
let mac_span = original_span(fld.cx);
let expanded = match expandfun.expand(fld.cx,
mac_span.call_site,
marked_before.as_slice()).make_expr() {
Some(e) => e,
None => {
fld.cx.span_err(
pth.span,
2014-05-19 20:59:35 +00:00
format!("non-expression macro in expression position: {}",
extnamestr.get().as_slice()
).as_slice());
return DummyResult::raw_expr(e.span);
}
};
// mark after:
mark_expr(expanded,fm)
}
_ => {
fld.cx.span_err(
pth.span,
2014-01-30 23:53:09 +00:00
format!("'{}' is not a tt-style macro",
extnamestr.get()).as_slice());
return DummyResult::raw_expr(e.span);
}
};
// Keep going, outside-in.
//
// FIXME(pcwalton): Is it necessary to clone the
// node here?
let fully_expanded =
fld.fold_expr(marked_after).node.clone();
fld.cx.bt_pop();
2014-05-16 07:16:13 +00:00
box(GC) ast::Expr {
id: ast::DUMMY_NODE_ID,
node: fully_expanded,
span: e.span,
}
}
}
}
// Desugar expr_for_loop
// From: `['<ident>:] for <src_pat> in <src_expr> <src_loop_block>`
2014-02-06 22:38:33 +00:00
// FIXME #6993: change type of opt_ident to Option<Name>
ast::ExprForLoop(src_pat, src_expr, src_loop_block, opt_ident) => {
let span = e.span;
// to:
//
// match &mut <src_expr> {
// i => {
// ['<ident>:] loop {
// match i.next() {
// None => break ['<ident>],
// Some(mut value) => {
// let <src_pat> = value;
// <src_loop_block>
// }
// }
// }
// }
// }
//
// (The use of the `let` is to give better error messages
// when the pattern is refutable.)
let local_ident = token::gensym_ident("i");
let next_ident = fld.cx.ident_of("next");
let none_ident = fld.cx.ident_of("None");
let local_path = fld.cx.path_ident(span, local_ident);
let some_path = fld.cx.path_ident(span, fld.cx.ident_of("Some"));
// `None => break ['<ident>],`
let none_arm = {
let break_expr = fld.cx.expr(span, ast::ExprBreak(opt_ident));
let none_pat = fld.cx.pat_ident(span, none_ident);
fld.cx.arm(span, vec!(none_pat), break_expr)
};
// let <src_pat> = value;
// use underscore to suppress lint error:
let value_ident = token::gensym_ident("_value");
// this is careful to use src_pat.span so that error
// messages point exact at that.
2014-05-16 17:15:33 +00:00
let local = box(GC) ast::Local {
ty: fld.cx.ty_infer(src_pat.span),
pat: src_pat,
init: Some(fld.cx.expr_ident(src_pat.span, value_ident)),
id: ast::DUMMY_NODE_ID,
span: src_pat.span,
source: ast::LocalFor
};
let local = codemap::respan(src_pat.span, ast::DeclLocal(local));
2014-05-16 17:15:33 +00:00
let local = box(GC) codemap::respan(span, ast::StmtDecl(box(GC) local,
ast::DUMMY_NODE_ID));
// { let ...; <src_loop_block> }
let block = fld.cx.block(span, vec![local],
Some(fld.cx.expr_block(src_loop_block)));
// `Some(mut value) => { ... }`
// Note the _'s in the name will stop any unused mutability warnings.
let value_pat = fld.cx.pat_ident_binding_mode(span, value_ident,
ast::BindByValue(ast::MutMutable));
let some_arm =
fld.cx.arm(span,
vec!(fld.cx.pat_enum(span, some_path, vec!(value_pat))),
fld.cx.expr_block(block));
// `match i.next() { ... }`
let match_expr = {
let next_call_expr =
fld.cx.expr_method_call(span,
fld.cx.expr_path(local_path),
next_ident,
Vec::new());
fld.cx.expr_match(span, next_call_expr, vec!(none_arm, some_arm))
};
// ['ident:] loop { ... }
let loop_expr = fld.cx.expr(span,
ast::ExprLoop(fld.cx.block_expr(match_expr),
opt_ident));
// `i => loop { ... }`
// `match &mut <src_expr> { i => loop { ... } }`
let discrim = fld.cx.expr_mut_addr_of(span, src_expr);
let i_pattern = fld.cx.pat_ident(span, local_ident);
let arm = fld.cx.arm(span, vec!(i_pattern), loop_expr);
// why these clone()'s everywhere? I guess I'll follow the pattern....
let match_expr = fld.cx.expr_match(span, discrim, vec!(arm));
fld.fold_expr(match_expr).clone()
}
ast::ExprLoop(loop_block, opt_ident) => {
let (loop_block, opt_ident) = expand_loop_block(loop_block, opt_ident, fld);
fld.cx.expr(e.span, ast::ExprLoop(loop_block, opt_ident))
}
ast::ExprFnBlock(fn_decl, block) => {
let (rewritten_fn_decl, rewritten_block)
= expand_and_rename_fn_decl_and_block(fn_decl, block, fld);
let new_node = ast::ExprFnBlock(rewritten_fn_decl, rewritten_block);
box(GC) ast::Expr{id:e.id, node: new_node, span: fld.new_span(e.span)}
}
ast::ExprProc(fn_decl, block) => {
let (rewritten_fn_decl, rewritten_block)
= expand_and_rename_fn_decl_and_block(fn_decl, block, fld);
let new_node = ast::ExprProc(rewritten_fn_decl, rewritten_block);
box(GC) ast::Expr{id:e.id, node: new_node, span: fld.new_span(e.span)}
}
_ => noop_fold_expr(e, fld)
}
}
// Rename loop label and expand its loop body
//
// The renaming procedure for loop is different in the sense that the loop
// body is in a block enclosed by loop head so the renaming of loop label
// must be propagated to the enclosed context.
fn expand_loop_block(loop_block: P<Block>,
opt_ident: Option<Ident>,
fld: &mut MacroExpander) -> (P<Block>, Option<Ident>) {
match opt_ident {
Some(label) => {
let new_label = fresh_name(&label);
let rename = (label, new_label);
// The rename *must not* be added to the pending list of current
// syntax context otherwise an unrelated `break` or `continue` in
// the same context will pick that up in the deferred renaming pass
// and be renamed incorrectly.
let mut rename_list = vec!(rename);
2014-06-26 20:18:52 +00:00
let mut rename_fld = IdentRenamer{renames: &mut rename_list};
let renamed_ident = rename_fld.fold_ident(label);
// The rename *must* be added to the enclosed syntax context for
// `break` or `continue` to pick up because by definition they are
// in a block enclosed by loop head.
fld.extsbox.push_frame();
fld.extsbox.info().pending_renames.push(rename);
2014-05-16 07:16:13 +00:00
let expanded_block = expand_block_elts(&*loop_block, fld);
fld.extsbox.pop_frame();
(expanded_block, Some(renamed_ident))
}
None => (fld.fold_block(loop_block), opt_ident)
}
}
// eval $e with a new exts frame.
// must be a macro so that $e isn't evaluated too early.
macro_rules! with_exts_frame (
($extsboxexpr:expr,$macros_escape:expr,$e:expr) =>
({$extsboxexpr.push_frame();
$extsboxexpr.info().macros_escape = $macros_escape;
let result = $e;
$extsboxexpr.pop_frame();
result
})
)
// When we enter a module, record it, for the sake of `module!`
2014-06-25 22:20:01 +00:00
fn expand_item(it: Gc<ast::Item>, fld: &mut MacroExpander)
2014-05-16 07:16:13 +00:00
-> SmallVector<Gc<ast::Item>> {
let it = expand_item_modifiers(it, fld);
let mut decorator_items = SmallVector::zero();
let mut new_attrs = Vec::new();
for attr in it.attrs.iter() {
let mname = attr.name();
match fld.extsbox.find(&intern(mname.get())) {
Some(&ItemDecorator(dec_fn)) => {
attr::mark_used(attr);
fld.cx.bt_push(ExpnInfo {
call_site: attr.span,
callee: NameAndSpan {
name: mname.get().to_string(),
format: MacroAttribute,
span: None
}
});
// we'd ideally decorator_items.push_all(expand_item(item, fld)),
// but that double-mut-borrows fld
2014-05-16 07:16:13 +00:00
let mut items: SmallVector<Gc<ast::Item>> = SmallVector::zero();
dec_fn(fld.cx, attr.span, attr.node.value, it,
|item| items.push(item));
decorator_items.extend(items.move_iter()
.flat_map(|item| expand_item(item, fld).move_iter()));
fld.cx.bt_pop();
}
_ => new_attrs.push((*attr).clone()),
}
}
let mut new_items = match it.node {
ast::ItemMac(..) => expand_item_mac(it, fld),
ast::ItemMod(_) | ast::ItemForeignMod(_) => {
fld.cx.mod_push(it.ident);
let macro_escape = contains_macro_escape(new_attrs.as_slice());
let result = with_exts_frame!(fld.extsbox,
macro_escape,
2014-05-16 07:16:13 +00:00
noop_fold_item(&*it, fld));
fld.cx.mod_pop();
result
},
_ => {
2014-05-16 07:16:13 +00:00
let it = box(GC) ast::Item {
attrs: new_attrs,
..(*it).clone()
};
2014-05-16 07:16:13 +00:00
noop_fold_item(&*it, fld)
}
};
new_items.push_all(decorator_items);
new_items
}
2014-05-16 07:16:13 +00:00
fn expand_item_modifiers(mut it: Gc<ast::Item>, fld: &mut MacroExpander)
-> Gc<ast::Item> {
// partition the attributes into ItemModifiers and others
let (modifiers, other_attrs) = it.attrs.partitioned(|attr| {
match fld.extsbox.find(&intern(attr.name().get())) {
Some(&ItemModifier(_)) => true,
_ => false
}
});
// update the attrs, leave everything else alone. Is this mutation really a good idea?
2014-05-16 07:16:13 +00:00
it = box(GC) ast::Item {
attrs: other_attrs,
..(*it).clone()
};
if modifiers.is_empty() {
return it;
}
for attr in modifiers.iter() {
let mname = attr.name();
match fld.extsbox.find(&intern(mname.get())) {
Some(&ItemModifier(dec_fn)) => {
attr::mark_used(attr);
fld.cx.bt_push(ExpnInfo {
call_site: attr.span,
callee: NameAndSpan {
name: mname.get().to_string(),
format: MacroAttribute,
span: None,
}
});
it = dec_fn(fld.cx, attr.span, attr.node.value, it);
fld.cx.bt_pop();
}
_ => unreachable!()
}
}
// expansion may have added new ItemModifiers
expand_item_modifiers(it, fld)
}
/// Expand item_underscore
fn expand_item_underscore(item: &ast::Item_, fld: &mut MacroExpander) -> ast::Item_ {
match *item {
ast::ItemFn(decl, fn_style, abi, ref generics, body) => {
let (rewritten_fn_decl, rewritten_body)
= expand_and_rename_fn_decl_and_block(decl,body,fld);
let expanded_generics = fold::fold_generics(generics,fld);
ast::ItemFn(rewritten_fn_decl, fn_style, abi, expanded_generics, rewritten_body)
}
_ => noop_fold_item_underscore(&*item, fld)
}
}
// does this attribute list contain "macro_escape" ?
2014-06-25 22:20:01 +00:00
fn contains_macro_escape(attrs: &[ast::Attribute]) -> bool {
attr::contains_name(attrs, "macro_escape")
}
// Support for item-position macro invocations, exactly the same
// logic as for expression-position macro invocations.
2014-06-25 22:20:01 +00:00
fn expand_item_mac(it: Gc<ast::Item>, fld: &mut MacroExpander)
2014-05-16 07:16:13 +00:00
-> SmallVector<Gc<ast::Item>> {
let (pth, tts) = match it.node {
ItemMac(codemap::Spanned {
node: MacInvocTT(ref pth, ref tts, _),
2013-11-28 20:22:53 +00:00
..
}) => {
(pth, (*tts).clone())
}
_ => fld.cx.span_bug(it.span, "invalid item macro invocation")
};
let extname = pth.segments.get(0).identifier;
let extnamestr = token::get_ident(extname);
let fm = fresh_mark();
let expanded = match fld.extsbox.find(&extname.name) {
None => {
fld.cx.span_err(pth.span,
2014-01-30 23:53:09 +00:00
format!("macro undefined: '{}!'",
extnamestr).as_slice());
// let compilation continue
return SmallVector::zero();
}
Some(&NormalTT(ref expander, span)) => {
if it.ident.name != parse::token::special_idents::invalid.name {
fld.cx
.span_err(pth.span,
format!("macro {}! expects no ident argument, \
given '{}'",
extnamestr,
token::get_ident(it.ident)).as_slice());
return SmallVector::zero();
}
fld.cx.bt_push(ExpnInfo {
call_site: it.span,
callee: NameAndSpan {
name: extnamestr.get().to_string(),
format: MacroBang,
span: span
}
});
// mark before expansion:
let marked_before = mark_tts(tts.as_slice(), fm);
expander.expand(fld.cx, it.span, marked_before.as_slice())
}
Some(&IdentTT(ref expander, span)) => {
if it.ident.name == parse::token::special_idents::invalid.name {
fld.cx.span_err(pth.span,
2014-01-30 23:53:09 +00:00
format!("macro {}! expects an ident argument",
extnamestr.get()).as_slice());
return SmallVector::zero();
}
fld.cx.bt_push(ExpnInfo {
call_site: it.span,
callee: NameAndSpan {
name: extnamestr.get().to_string(),
format: MacroBang,
span: span
}
});
// mark before expansion:
let marked_tts = mark_tts(tts.as_slice(), fm);
expander.expand(fld.cx, it.span, it.ident, marked_tts)
}
_ => {
2014-01-30 23:53:09 +00:00
fld.cx.span_err(it.span,
format!("{}! is not legal in item position",
extnamestr.get()).as_slice());
return SmallVector::zero();
}
};
let items = match expanded.make_def() {
Some(MacroDef { name, ext }) => {
// 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.
fld.extsbox.insert(intern(name.as_slice()), ext);
if attr::contains_name(it.attrs.as_slice(), "macro_export") {
2013-12-25 18:10:33 +00:00
SmallVector::one(it)
} else {
SmallVector::zero()
}
}
None => {
match expanded.make_items() {
Some(items) => {
items.move_iter()
.flat_map(|i| mark_item(i, fm).move_iter())
.flat_map(|i| fld.fold_item(i).move_iter())
.collect()
}
None => {
fld.cx.span_err(pth.span,
2014-05-19 20:59:35 +00:00
format!("non-item macro in item position: {}",
extnamestr.get()).as_slice());
return SmallVector::zero();
}
}
}
};
fld.cx.bt_pop();
return items;
}
// expand a stmt
2014-06-25 22:20:01 +00:00
fn expand_stmt(s: &Stmt, fld: &mut MacroExpander) -> SmallVector<Gc<Stmt>> {
// why the copying here and not in expand_expr?
// looks like classic changed-in-only-one-place
let (pth, tts, semi) = match s.node {
StmtMac(ref mac, semi) => {
match mac.node {
MacInvocTT(ref pth, ref tts, _) => {
(pth, (*tts).clone(), semi)
}
}
}
_ => return expand_non_macro_stmt(s, fld)
};
2014-01-19 08:21:14 +00:00
if pth.segments.len() > 1u {
fld.cx.span_err(pth.span, "expected macro name without module separators");
return SmallVector::zero();
}
let extname = pth.segments.get(0).identifier;
let extnamestr = token::get_ident(extname);
let marked_after = match fld.extsbox.find(&extname.name) {
None => {
fld.cx.span_err(pth.span,
format!("macro undefined: '{}'",
extnamestr).as_slice());
return SmallVector::zero();
}
Some(&NormalTT(ref expandfun, exp_span)) => {
fld.cx.bt_push(ExpnInfo {
call_site: s.span,
callee: NameAndSpan {
name: extnamestr.get().to_string(),
format: MacroBang,
span: exp_span,
}
});
let fm = fresh_mark();
// mark before expansion:
let marked_tts = mark_tts(tts.as_slice(), fm);
// See the comment in expand_expr for why we want the original span,
// not the current mac.span.
let mac_span = original_span(fld.cx);
let expanded = match expandfun.expand(fld.cx,
mac_span.call_site,
marked_tts.as_slice()).make_stmt() {
Some(stmt) => stmt,
None => {
fld.cx.span_err(pth.span,
2014-05-19 20:59:35 +00:00
format!("non-statement macro in statement position: {}",
extnamestr).as_slice());
return SmallVector::zero();
}
};
2014-05-16 07:16:13 +00:00
mark_stmt(&*expanded,fm)
}
_ => {
2014-01-30 23:53:09 +00:00
fld.cx.span_err(pth.span, format!("'{}' is not a tt-style macro",
extnamestr).as_slice());
return SmallVector::zero();
}
};
// Keep going, outside-in.
2014-05-16 07:16:13 +00:00
let fully_expanded = fld.fold_stmt(&*marked_after);
if fully_expanded.is_empty() {
fld.cx.span_err(pth.span, "macro didn't expand to a statement");
return SmallVector::zero();
}
fld.cx.bt_pop();
2014-05-16 07:16:13 +00:00
let fully_expanded: SmallVector<Gc<Stmt>> = fully_expanded.move_iter()
.map(|s| box(GC) Spanned { span: s.span, node: s.node.clone() })
.collect();
fully_expanded.move_iter().map(|s| {
match s.node {
StmtExpr(e, stmt_id) if semi => {
2014-05-16 07:16:13 +00:00
box(GC) Spanned {
span: s.span,
node: StmtSemi(e, stmt_id)
}
}
_ => s /* might already have a semi */
}
}).collect()
}
// expand a non-macro stmt. this is essentially the fallthrough for
// expand_stmt, above.
2013-12-28 03:34:51 +00:00
fn expand_non_macro_stmt(s: &Stmt, fld: &mut MacroExpander)
2014-05-16 07:16:13 +00:00
-> SmallVector<Gc<Stmt>> {
// is it a let?
match s.node {
StmtDecl(decl, node_id) => {
match *decl {
Spanned {
node: DeclLocal(ref local),
span: stmt_span
} => {
// take it apart:
let Local {
ty: _,
pat: pat,
init: init,
id: id,
span: span,
source: source,
} = **local;
// expand the pat (it might contain macro uses):
let expanded_pat = fld.fold_pat(pat);
// find the PatIdents 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.
// generate fresh names, push them to a new pending list
let idents = pattern_bindings(expanded_pat);
let mut new_pending_renames =
idents.iter().map(|ident| (*ident, fresh_name(ident))).collect();
// rewrite the pattern using the new names (the old
// ones have already been applied):
let rewritten_pat = {
// nested binding to allow borrow to expire:
let mut rename_fld = IdentRenamer{renames: &mut new_pending_renames};
rename_fld.fold_pat(expanded_pat)
};
// add them to the existing pending renames:
fld.extsbox.info().pending_renames.push_all_move(new_pending_renames);
// also, don't forget to expand the init:
let new_init_opt = init.map(|e| fld.fold_expr(e));
let rewritten_local =
2014-05-16 07:16:13 +00:00
box(GC) Local {
ty: local.ty,
pat: rewritten_pat,
init: new_init_opt,
id: id,
span: span,
source: source
};
2014-05-16 07:16:13 +00:00
SmallVector::one(box(GC) Spanned {
node: StmtDecl(box(GC) Spanned {
node: DeclLocal(rewritten_local),
span: stmt_span
},
node_id),
span: span
})
}
_ => noop_fold_stmt(s, fld),
}
},
_ => noop_fold_stmt(s, fld),
}
}
// expand the arm of a 'match', renaming for macro hygiene
2014-06-25 22:20:01 +00:00
fn expand_arm(arm: &ast::Arm, fld: &mut MacroExpander) -> ast::Arm {
// expand pats... they might contain macro uses:
let expanded_pats : Vec<Gc<ast::Pat>> = arm.pats.iter().map(|pat| fld.fold_pat(*pat)).collect();
if expanded_pats.len() == 0 {
2014-06-25 22:20:01 +00:00
fail!("encountered match arm with 0 patterns");
}
2014-06-26 20:19:34 +00:00
// all of the pats must have the same set of bindings, so use the
// first one to extract them and generate new names:
let first_pat = expanded_pats.get(0);
let idents = pattern_bindings(*first_pat);
let new_renames =
idents.iter().map(|id| (*id,fresh_name(id))).collect();
2014-07-04 19:15:24 +00:00
// apply the renaming, but only to the PatIdents:
let mut rename_pats_fld = PatIdentRenamer{renames:&new_renames};
let rewritten_pats =
2014-07-04 19:15:24 +00:00
expanded_pats.iter().map(|pat| rename_pats_fld.fold_pat(*pat)).collect();
// apply renaming and then expansion to the guard and the body:
let mut rename_fld = IdentRenamer{renames:&new_renames};
let rewritten_guard =
arm.guard.map(|g| fld.fold_expr(rename_fld.fold_expr(g)));
let rewritten_body = fld.fold_expr(rename_fld.fold_expr(arm.body));
2014-06-26 20:19:34 +00:00
ast::Arm {
attrs: arm.attrs.iter().map(|x| fld.fold_attribute(*x)).collect(),
pats: rewritten_pats,
guard: rewritten_guard,
body: rewritten_body,
2014-06-26 23:53:33 +00:00
}
2014-06-25 22:20:01 +00:00
}
/// A visitor that extracts the PatIdent (binding) paths
/// from a given thingy and puts them in a mutable
/// array
#[deriving(Clone)]
struct PatIdentFinder {
ident_accumulator: Vec<ast::Ident> ,
}
impl Visitor<()> for PatIdentFinder {
2013-11-25 18:22:21 +00:00
fn visit_pat(&mut self, pattern: &ast::Pat, _: ()) {
match *pattern {
ast::Pat { id: _, node: ast::PatIdent(_, ref path1, ref inner), span: _ } => {
self.ident_accumulator.push(path1.node);
// visit optional subpattern of PatIdent:
for subpat in inner.iter() {
2014-05-16 07:16:13 +00:00
self.visit_pat(&**subpat, ())
2013-06-04 21:56:33 +00:00
}
}
// use the default traversal for non-PatIdents
_ => visit::walk_pat(self, pattern, ())
}
}
}
/// find the PatIdent paths in a pattern
fn pattern_bindings(pat : &ast::Pat) -> Vec<ast::Ident> {
let mut name_finder = PatIdentFinder{ident_accumulator:Vec::new()};
name_finder.visit_pat(pat,());
name_finder.ident_accumulator
}
/// find the PatIdent paths in a
fn fn_decl_arg_bindings(fn_decl: &ast::FnDecl) -> Vec<ast::Ident> {
let mut pat_idents = PatIdentFinder{ident_accumulator:Vec::new()};
for arg in fn_decl.inputs.iter() {
pat_idents.visit_pat(arg.pat,());
}
pat_idents.ident_accumulator
}
2013-07-05 20:57:53 +00:00
// expand a block. pushes a new exts_frame, then calls expand_block_elts
2014-06-25 22:20:01 +00:00
fn expand_block(blk: &Block, fld: &mut MacroExpander) -> P<Block> {
// see note below about treatment of exts table
with_exts_frame!(fld.extsbox,false,
expand_block_elts(blk, fld))
}
2013-07-05 20:57:53 +00:00
// expand the elements of a block.
2014-06-25 22:20:01 +00:00
fn expand_block_elts(b: &Block, fld: &mut MacroExpander) -> P<Block> {
let new_view_items = b.view_items.iter().map(|x| fld.fold_view_item(x)).collect();
let new_stmts =
b.stmts.iter().flat_map(|x| {
2014-06-26 20:18:52 +00:00
// perform all pending renames
let renamed_stmt = {
let pending_renames = &mut fld.extsbox.info().pending_renames;
2014-06-26 20:18:52 +00:00
let mut rename_fld = IdentRenamer{renames:pending_renames};
2014-05-16 07:16:13 +00:00
rename_fld.fold_stmt(&**x).expect_one("rename_fold didn't return one value")
};
2014-06-26 20:18:52 +00:00
// expand macros in the statement
2014-05-16 07:16:13 +00:00
fld.fold_stmt(&*renamed_stmt).move_iter()
}).collect();
let new_expr = b.expr.map(|x| {
let expr = {
let pending_renames = &mut fld.extsbox.info().pending_renames;
2014-06-26 20:18:52 +00:00
let mut rename_fld = IdentRenamer{renames:pending_renames};
rename_fld.fold_expr(x)
};
fld.fold_expr(expr)
});
P(Block {
view_items: new_view_items,
stmts: new_stmts,
expr: new_expr,
id: fld.new_id(b.id),
rules: b.rules,
span: b.span,
})
2013-05-17 16:41:04 +00:00
}
2014-06-25 22:20:01 +00:00
fn expand_pat(p: Gc<ast::Pat>, fld: &mut MacroExpander) -> Gc<ast::Pat> {
2014-05-19 20:59:35 +00:00
let (pth, tts) = match p.node {
PatMac(ref mac) => {
match mac.node {
MacInvocTT(ref pth, ref tts, _) => {
(pth, (*tts).clone())
}
}
}
_ => return noop_fold_pat(p, fld),
};
if pth.segments.len() > 1u {
fld.cx.span_err(pth.span, "expected macro name without module separators");
return DummyResult::raw_pat(p.span);
}
let extname = pth.segments.get(0).identifier;
let extnamestr = token::get_ident(extname);
let marked_after = match fld.extsbox.find(&extname.name) {
None => {
fld.cx.span_err(pth.span,
format!("macro undefined: '{}!'",
extnamestr).as_slice());
// let compilation continue
return DummyResult::raw_pat(p.span);
}
Some(&NormalTT(ref expander, span)) => {
fld.cx.bt_push(ExpnInfo {
call_site: p.span,
callee: NameAndSpan {
name: extnamestr.get().to_string(),
format: MacroBang,
span: span
}
});
let fm = fresh_mark();
let marked_before = mark_tts(tts.as_slice(), fm);
let mac_span = original_span(fld.cx);
let expanded = match expander.expand(fld.cx,
mac_span.call_site,
marked_before.as_slice()).make_pat() {
Some(e) => e,
None => {
fld.cx.span_err(
pth.span,
format!(
"non-pattern macro in pattern position: {}",
extnamestr.get()
).as_slice()
);
return DummyResult::raw_pat(p.span);
}
};
// mark after:
mark_pat(expanded,fm)
}
_ => {
fld.cx.span_err(p.span,
format!("{}! is not legal in pattern position",
extnamestr.get()).as_slice());
return DummyResult::raw_pat(p.span);
}
};
let fully_expanded =
fld.fold_pat(marked_after).node.clone();
fld.cx.bt_pop();
2014-05-16 17:15:33 +00:00
box(GC) ast::Pat {
2014-05-19 20:59:35 +00:00
id: ast::DUMMY_NODE_ID,
node: fully_expanded,
span: p.span,
}
}
2014-07-03 06:17:10 +00:00
/// A tree-folder that applies every rename in its (mutable) list
/// to every identifier, including both bindings and varrefs
/// (and lots of things that will turn out to be neither)
2014-03-02 23:26:39 +00:00
pub struct IdentRenamer<'a> {
renames: &'a mtwt::RenameList,
}
impl<'a> Folder for IdentRenamer<'a> {
fn fold_ident(&mut self, id: Ident) -> Ident {
Ident {
name: id.name,
ctxt: mtwt::apply_renames(self.renames, id.ctxt),
}
}
}
2014-07-03 06:16:01 +00:00
/// A tree-folder that applies every rename in its list to
/// the idents that are in PatIdent patterns. This is more narrowly
/// focused than IdentRenamer, and is needed for FnDecl,
/// where we want to rename the args but not the fn name or the generics etc.
pub struct PatIdentRenamer<'a> {
renames: &'a mtwt::RenameList,
}
impl<'a> Folder for PatIdentRenamer<'a> {
fn fold_pat(&mut self, pat: Gc<ast::Pat>) -> Gc<ast::Pat> {
match pat.node {
ast::PatIdent(binding_mode, Spanned{span: ref sp, node: id}, ref sub) => {
let new_ident = Ident{name: id.name,
ctxt: mtwt::apply_renames(self.renames, id.ctxt)};
2014-07-03 06:16:01 +00:00
let new_node =
ast::PatIdent(binding_mode,
Spanned{span: self.new_span(*sp), node: new_ident},
sub.map(|p| self.fold_pat(p)));
box(GC) ast::Pat {
id: pat.id,
span: self.new_span(pat.span),
node: new_node,
}
},
_ => noop_fold_pat(pat, self)
}
}
}
/// Given a fn_decl and a block and a MacroExpander, expand the fn_decl, then use the
/// PatIdents in its arguments to perform renaming in the FnDecl and
/// the block, returning both the new FnDecl and the new Block.
fn expand_and_rename_fn_decl_and_block(fn_decl: &ast::FnDecl, block: Gc<ast::Block>,
fld: &mut MacroExpander)
-> (Gc<ast::FnDecl>, Gc<ast::Block>) {
let expanded_decl = fld.fold_fn_decl(fn_decl);
let idents = fn_decl_arg_bindings(expanded_decl);
let renames =
idents.iter().map(|id : &ast::Ident| (*id,fresh_name(id))).collect();
// first, a renamer for the PatIdents, for the fn_decl:
let mut rename_pat_fld = PatIdentRenamer{renames: &renames};
let rewritten_fn_decl = rename_pat_fld.fold_fn_decl(expanded_decl);
// now, a renamer for *all* idents, for the body:
let mut rename_fld = IdentRenamer{renames: &renames};
let rewritten_body = fld.fold_block(rename_fld.fold_block(block));
(rewritten_fn_decl,rewritten_body)
}
2014-07-03 06:16:01 +00:00
/// A tree-folder that performs macro expansion
2014-03-09 14:54:34 +00:00
pub struct MacroExpander<'a, 'b> {
pub extsbox: SyntaxEnv,
pub cx: &'a mut ExtCtxt<'b>,
}
2014-03-09 14:54:34 +00:00
impl<'a, 'b> Folder for MacroExpander<'a, 'b> {
2014-05-16 07:16:13 +00:00
fn fold_expr(&mut self, expr: Gc<ast::Expr>) -> Gc<ast::Expr> {
expand_expr(expr, self)
}
2014-05-16 07:16:13 +00:00
fn fold_pat(&mut self, pat: Gc<ast::Pat>) -> Gc<ast::Pat> {
2014-05-19 20:59:35 +00:00
expand_pat(pat, self)
}
2014-05-16 07:16:13 +00:00
fn fold_item(&mut self, item: Gc<ast::Item>) -> SmallVector<Gc<ast::Item>> {
expand_item(item, self)
}
fn fold_item_underscore(&mut self, item: &ast::Item_) -> ast::Item_ {
expand_item_underscore(item, self)
}
2014-05-16 07:16:13 +00:00
fn fold_stmt(&mut self, stmt: &ast::Stmt) -> SmallVector<Gc<ast::Stmt>> {
expand_stmt(stmt, self)
}
2013-12-28 03:34:51 +00:00
fn fold_block(&mut self, block: P<Block>) -> P<Block> {
2014-05-16 07:16:13 +00:00
expand_block(&*block, self)
}
2014-06-25 22:20:01 +00:00
fn fold_arm(&mut self, arm: &ast::Arm) -> ast::Arm {
expand_arm(arm, self)
}
2013-12-28 03:34:51 +00:00
fn new_span(&mut self, span: Span) -> Span {
new_span(self.cx, span)
}
}
2014-07-03 06:16:01 +00:00
fn new_span(cx: &ExtCtxt, sp: Span) -> Span {
/* this discards information in the case of macro-defining macros */
Span {
lo: sp.lo,
hi: sp.hi,
expn_info: cx.backtrace(),
}
}
pub struct ExpansionConfig {
pub deriving_hash_type_parameter: bool,
pub crate_id: CrateId,
}
pub struct ExportedMacros {
pub crate_name: Ident,
pub macros: Vec<String>,
}
2014-03-09 14:54:34 +00:00
pub fn expand_crate(parse_sess: &parse::ParseSess,
cfg: ExpansionConfig,
macros: Vec<ExportedMacros>,
user_exts: Vec<NamedSyntaxExtension>,
2013-09-28 02:46:09 +00:00
c: Crate) -> Crate {
let mut cx = ExtCtxt::new(parse_sess, c.config.clone(), cfg);
2013-12-28 03:34:51 +00:00
let mut expander = MacroExpander {
extsbox: syntax_expander_table(),
2013-12-29 05:06:22 +00:00
cx: &mut cx,
2013-12-28 00:17:36 +00:00
};
for ExportedMacros { crate_name, macros } in macros.move_iter() {
let name = format!("<{} macros>", token::get_ident(crate_name))
.into_string();
for source in macros.move_iter() {
let item = parse::parse_item_from_source_str(name.clone(),
source,
expander.cx.cfg(),
expander.cx.parse_sess())
.expect("expected a serialized item");
expand_item_mac(item, &mut expander);
}
}
for (name, extension) in user_exts.move_iter() {
expander.extsbox.insert(name, extension);
}
2013-09-28 02:46:09 +00:00
let ret = expander.fold_crate(c);
parse_sess.span_diagnostic.handler().abort_if_errors();
return ret;
}
// 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:
// Marker - add a mark to a context
2014-02-06 22:38:33 +00:00
// A Marker adds the given mark to the syntax context
struct Marker { mark: Mrk }
2014-02-06 22:38:33 +00:00
impl Folder for Marker {
fn fold_ident(&mut self, id: Ident) -> Ident {
ast::Ident {
2014-02-06 22:38:33 +00:00
name: id.name,
ctxt: mtwt::apply_mark(self.mark, id.ctxt)
}
}
fn fold_mac(&mut self, m: &ast::Mac) -> ast::Mac {
let macro = match m.node {
MacInvocTT(ref path, ref tts, ctxt) => {
MacInvocTT(self.fold_path(path),
fold_tts(tts.as_slice(), self),
mtwt::apply_mark(self.mark, ctxt))
}
};
Spanned {
node: macro,
span: m.span,
}
}
}
// apply a given mark to the given token trees. Used prior to expansion of a macro.
fn mark_tts(tts: &[TokenTree], m: Mrk) -> Vec<TokenTree> {
2014-06-26 22:37:18 +00:00
fold_tts(tts, &mut Marker{mark:m})
}
// apply a given mark to the given expr. Used following the expansion of a macro.
2014-05-16 07:16:13 +00:00
fn mark_expr(expr: Gc<ast::Expr>, m: Mrk) -> Gc<ast::Expr> {
2014-06-26 22:37:18 +00:00
Marker{mark:m}.fold_expr(expr)
}
2014-05-19 20:59:35 +00:00
// apply a given mark to the given pattern. Used following the expansion of a macro.
2014-05-16 17:15:33 +00:00
fn mark_pat(pat: Gc<ast::Pat>, m: Mrk) -> Gc<ast::Pat> {
2014-06-26 22:37:18 +00:00
Marker{mark:m}.fold_pat(pat)
2014-05-19 20:59:35 +00:00
}
// apply a given mark to the given stmt. Used following the expansion of a macro.
2014-05-16 07:16:13 +00:00
fn mark_stmt(expr: &ast::Stmt, m: Mrk) -> Gc<ast::Stmt> {
2014-06-26 22:37:18 +00:00
Marker{mark:m}.fold_stmt(expr)
2013-11-26 04:02:15 +00:00
.expect_one("marking a stmt didn't return a stmt")
}
// apply a given mark to the given item. Used following the expansion of a macro.
2014-05-16 07:16:13 +00:00
fn mark_item(expr: Gc<ast::Item>, m: Mrk) -> SmallVector<Gc<ast::Item>> {
2014-06-26 22:37:18 +00:00
Marker{mark:m}.fold_item(expr)
}
2014-05-16 07:16:13 +00:00
fn original_span(cx: &ExtCtxt) -> Gc<codemap::ExpnInfo> {
let mut relevant_info = cx.backtrace();
let mut einfo = relevant_info.unwrap();
loop {
match relevant_info {
None => { break }
Some(e) => {
einfo = e;
relevant_info = einfo.call_site.expn_info;
}
}
}
return einfo;
}
#[cfg(test)]
mod test {
use super::{pattern_bindings, expand_crate, contains_macro_escape};
use super::{PatIdentFinder, IdentRenamer, PatIdentRenamer};
use ast;
use ast::{Attribute_, AttrOuter, MetaWord};
2014-05-20 07:07:24 +00:00
use attr;
use codemap;
use codemap::Spanned;
use ext::mtwt;
use fold::Folder;
use parse;
use parse::token;
2014-03-09 14:54:34 +00:00
use util::parser_testing::{string_to_parser};
use util::parser_testing::{string_to_pat, string_to_crate, strs_to_idents};
use visit;
use visit::Visitor;
use std::gc::GC;
// 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 PathExprFinderContext {
path_accumulator: Vec<ast::Path> ,
}
impl Visitor<()> for PathExprFinderContext {
2014-01-06 12:00:46 +00:00
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, but it should be fine.
}
_ => visit::walk_expr(self,expr,())
}
}
}
// find the variable references in a crate
fn crate_varrefs(the_crate : &ast::Crate) -> Vec<ast::Path> {
let mut path_finder = PathExprFinderContext{path_accumulator:Vec::new()};
visit::walk_crate(&mut path_finder, the_crate, ());
path_finder.path_accumulator
}
/// A Visitor that extracts the identifiers from a thingy.
// as a side note, I'm starting to want to abstract over these....
struct IdentFinder{
ident_accumulator: Vec<ast::Ident>
}
impl Visitor<()> for IdentFinder {
fn visit_ident(&mut self, _: codemap::Span, id: ast::Ident, _: ()){
self.ident_accumulator.push(id);
}
}
/// Find the idents in a crate
fn crate_idents(the_crate: &ast::Crate) -> Vec<ast::Ident> {
let mut ident_finder = IdentFinder{ident_accumulator: Vec::new()};
visit::walk_crate(&mut ident_finder, the_crate, ());
ident_finder.ident_accumulator
}
// these following tests are quite fragile, in that they don't test what
// *kind* of failure occurs.
2014-06-24 01:45:51 +00:00
// make sure that macros can't escape fns
#[should_fail]
#[test] fn macros_cant_escape_fns_test () {
2014-04-16 01:17:48 +00:00
let src = "fn bogus() {macro_rules! z (() => (3+4))}\
fn inty() -> int { z!() }".to_string();
2014-02-06 22:38:33 +00:00
let sess = parse::new_parse_sess();
let crate_ast = parse::parse_crate_from_source_str(
"<test>".to_string(),
src,
2014-03-17 07:55:41 +00:00
Vec::new(), &sess);
// should fail:
let cfg = ::syntax::ext::expand::ExpansionConfig {
deriving_hash_type_parameter: false,
log: Introduce liblog, the old std::logging This commit moves all logging out of the standard library into an external crate. This crate is the new crate which is responsible for all logging macros and logging implementation. A few reasons for this change are: * The crate map has always been a bit of a code smell among rust programs. It has difficulty being loaded on almost all platforms, and it's used almost exclusively for logging and only logging. Removing the crate map is one of the end goals of this movement. * The compiler has a fair bit of special support for logging. It has the __log_level() expression as well as generating a global word per module specifying the log level. This is unfairly favoring the built-in logging system, and is much better done purely in libraries instead of the compiler itself. * Initialization of logging is much easier to do if there is no reliance on a magical crate map being available to set module log levels. * If the logging library can be written outside of the standard library, there's no reason that it shouldn't be. It's likely that we're not going to build the highest quality logging library of all time, so third-party libraries should be able to provide just as high-quality logging systems as the default one provided in the rust distribution. With a migration such as this, the change does not come for free. There are some subtle changes in the behavior of liblog vs the previous logging macros: * The core change of this migration is that there is no longer a physical log-level per module. This concept is still emulated (it is quite useful), but there is now only a global log level, not a local one. This global log level is a reflection of the maximum of all log levels specified. The previously generated logging code looked like: if specified_level <= __module_log_level() { println!(...) } The newly generated code looks like: if specified_level <= ::log::LOG_LEVEL { if ::log::module_enabled(module_path!()) { println!(...) } } Notably, the first layer of checking is still intended to be "super fast" in that it's just a load of a global word and a compare. The second layer of checking is executed to determine if the current module does indeed have logging turned on. This means that if any module has a debug log level turned on, all modules with debug log levels get a little bit slower (they all do more expensive dynamic checks to determine if they're turned on or not). Semantically, this migration brings no change in this respect, but runtime-wise, this will have a perf impact on some code. * A `RUST_LOG=::help` directive will no longer print out a list of all modules that can be logged. This is because the crate map will no longer specify the log levels of all modules, so the list of modules is not known. Additionally, warnings can no longer be provided if a malformed logging directive was supplied. The new "hello world" for logging looks like: #[phase(syntax, link)] extern crate log; fn main() { debug!("Hello, world!"); }
2014-03-09 06:11:44 +00:00
crate_id: from_str("test").unwrap(),
};
expand_crate(&sess,cfg,vec!(),vec!(),crate_ast);
}
2014-06-24 01:45:51 +00:00
// make sure that macros can't escape modules
#[should_fail]
#[test] fn macros_cant_escape_mods_test () {
2014-04-16 01:17:48 +00:00
let src = "mod foo {macro_rules! z (() => (3+4))}\
fn inty() -> int { z!() }".to_string();
2014-02-06 22:38:33 +00:00
let sess = parse::new_parse_sess();
let crate_ast = parse::parse_crate_from_source_str(
"<test>".to_string(),
src,
2014-03-17 07:55:41 +00:00
Vec::new(), &sess);
let cfg = ::syntax::ext::expand::ExpansionConfig {
deriving_hash_type_parameter: false,
log: Introduce liblog, the old std::logging This commit moves all logging out of the standard library into an external crate. This crate is the new crate which is responsible for all logging macros and logging implementation. A few reasons for this change are: * The crate map has always been a bit of a code smell among rust programs. It has difficulty being loaded on almost all platforms, and it's used almost exclusively for logging and only logging. Removing the crate map is one of the end goals of this movement. * The compiler has a fair bit of special support for logging. It has the __log_level() expression as well as generating a global word per module specifying the log level. This is unfairly favoring the built-in logging system, and is much better done purely in libraries instead of the compiler itself. * Initialization of logging is much easier to do if there is no reliance on a magical crate map being available to set module log levels. * If the logging library can be written outside of the standard library, there's no reason that it shouldn't be. It's likely that we're not going to build the highest quality logging library of all time, so third-party libraries should be able to provide just as high-quality logging systems as the default one provided in the rust distribution. With a migration such as this, the change does not come for free. There are some subtle changes in the behavior of liblog vs the previous logging macros: * The core change of this migration is that there is no longer a physical log-level per module. This concept is still emulated (it is quite useful), but there is now only a global log level, not a local one. This global log level is a reflection of the maximum of all log levels specified. The previously generated logging code looked like: if specified_level <= __module_log_level() { println!(...) } The newly generated code looks like: if specified_level <= ::log::LOG_LEVEL { if ::log::module_enabled(module_path!()) { println!(...) } } Notably, the first layer of checking is still intended to be "super fast" in that it's just a load of a global word and a compare. The second layer of checking is executed to determine if the current module does indeed have logging turned on. This means that if any module has a debug log level turned on, all modules with debug log levels get a little bit slower (they all do more expensive dynamic checks to determine if they're turned on or not). Semantically, this migration brings no change in this respect, but runtime-wise, this will have a perf impact on some code. * A `RUST_LOG=::help` directive will no longer print out a list of all modules that can be logged. This is because the crate map will no longer specify the log levels of all modules, so the list of modules is not known. Additionally, warnings can no longer be provided if a malformed logging directive was supplied. The new "hello world" for logging looks like: #[phase(syntax, link)] extern crate log; fn main() { debug!("Hello, world!"); }
2014-03-09 06:11:44 +00:00
crate_id: from_str("test").unwrap(),
};
expand_crate(&sess,cfg,vec!(),vec!(),crate_ast);
}
2014-06-24 01:45:51 +00:00
// macro_escape modules should allow macros to escape
#[test] fn macros_can_escape_flattened_mods_test () {
2014-04-16 01:17:48 +00:00
let src = "#[macro_escape] mod foo {macro_rules! z (() => (3+4))}\
fn inty() -> int { z!() }".to_string();
2014-02-06 22:38:33 +00:00
let sess = parse::new_parse_sess();
let crate_ast = parse::parse_crate_from_source_str(
"<test>".to_string(),
src,
2014-03-17 07:55:41 +00:00
Vec::new(), &sess);
let cfg = ::syntax::ext::expand::ExpansionConfig {
deriving_hash_type_parameter: false,
log: Introduce liblog, the old std::logging This commit moves all logging out of the standard library into an external crate. This crate is the new crate which is responsible for all logging macros and logging implementation. A few reasons for this change are: * The crate map has always been a bit of a code smell among rust programs. It has difficulty being loaded on almost all platforms, and it's used almost exclusively for logging and only logging. Removing the crate map is one of the end goals of this movement. * The compiler has a fair bit of special support for logging. It has the __log_level() expression as well as generating a global word per module specifying the log level. This is unfairly favoring the built-in logging system, and is much better done purely in libraries instead of the compiler itself. * Initialization of logging is much easier to do if there is no reliance on a magical crate map being available to set module log levels. * If the logging library can be written outside of the standard library, there's no reason that it shouldn't be. It's likely that we're not going to build the highest quality logging library of all time, so third-party libraries should be able to provide just as high-quality logging systems as the default one provided in the rust distribution. With a migration such as this, the change does not come for free. There are some subtle changes in the behavior of liblog vs the previous logging macros: * The core change of this migration is that there is no longer a physical log-level per module. This concept is still emulated (it is quite useful), but there is now only a global log level, not a local one. This global log level is a reflection of the maximum of all log levels specified. The previously generated logging code looked like: if specified_level <= __module_log_level() { println!(...) } The newly generated code looks like: if specified_level <= ::log::LOG_LEVEL { if ::log::module_enabled(module_path!()) { println!(...) } } Notably, the first layer of checking is still intended to be "super fast" in that it's just a load of a global word and a compare. The second layer of checking is executed to determine if the current module does indeed have logging turned on. This means that if any module has a debug log level turned on, all modules with debug log levels get a little bit slower (they all do more expensive dynamic checks to determine if they're turned on or not). Semantically, this migration brings no change in this respect, but runtime-wise, this will have a perf impact on some code. * A `RUST_LOG=::help` directive will no longer print out a list of all modules that can be logged. This is because the crate map will no longer specify the log levels of all modules, so the list of modules is not known. Additionally, warnings can no longer be provided if a malformed logging directive was supplied. The new "hello world" for logging looks like: #[phase(syntax, link)] extern crate log; fn main() { debug!("Hello, world!"); }
2014-03-09 06:11:44 +00:00
crate_id: from_str("test").unwrap(),
};
expand_crate(&sess, cfg, vec!(), vec!(), crate_ast);
}
#[test] fn test_contains_flatten (){
2014-01-31 02:46:19 +00:00
let attr1 = make_dummy_attr ("foo");
let attr2 = make_dummy_attr ("bar");
let escape_attr = make_dummy_attr ("macro_escape");
let attrs1 = vec!(attr1, escape_attr, attr2);
assert_eq!(contains_macro_escape(attrs1.as_slice()),true);
let attrs2 = vec!(attr1,attr2);
assert_eq!(contains_macro_escape(attrs2.as_slice()),false);
}
// make a MetaWord outer attribute with the given name
2014-01-31 02:46:19 +00:00
fn make_dummy_attr(s: &str) -> ast::Attribute {
Spanned {
span:codemap::DUMMY_SP,
node: Attribute_ {
2014-05-20 07:07:24 +00:00
id: attr::mk_attr_id(),
style: AttrOuter,
2014-05-16 07:16:13 +00:00
value: box(GC) Spanned {
2014-01-31 02:46:19 +00:00
node: MetaWord(token::intern_and_get_ident(s)),
span: codemap::DUMMY_SP,
},
is_sugared_doc: false,
}
}
}
fn expand_crate_str(crate_str: String) -> ast::Crate {
2014-03-09 14:54:34 +00:00
let ps = parse::new_parse_sess();
2014-03-17 07:55:41 +00:00
let crate_ast = string_to_parser(&ps, crate_str).parse_crate_mod();
// the cfg argument actually does matter, here...
let cfg = ::syntax::ext::expand::ExpansionConfig {
deriving_hash_type_parameter: false,
log: Introduce liblog, the old std::logging This commit moves all logging out of the standard library into an external crate. This crate is the new crate which is responsible for all logging macros and logging implementation. A few reasons for this change are: * The crate map has always been a bit of a code smell among rust programs. It has difficulty being loaded on almost all platforms, and it's used almost exclusively for logging and only logging. Removing the crate map is one of the end goals of this movement. * The compiler has a fair bit of special support for logging. It has the __log_level() expression as well as generating a global word per module specifying the log level. This is unfairly favoring the built-in logging system, and is much better done purely in libraries instead of the compiler itself. * Initialization of logging is much easier to do if there is no reliance on a magical crate map being available to set module log levels. * If the logging library can be written outside of the standard library, there's no reason that it shouldn't be. It's likely that we're not going to build the highest quality logging library of all time, so third-party libraries should be able to provide just as high-quality logging systems as the default one provided in the rust distribution. With a migration such as this, the change does not come for free. There are some subtle changes in the behavior of liblog vs the previous logging macros: * The core change of this migration is that there is no longer a physical log-level per module. This concept is still emulated (it is quite useful), but there is now only a global log level, not a local one. This global log level is a reflection of the maximum of all log levels specified. The previously generated logging code looked like: if specified_level <= __module_log_level() { println!(...) } The newly generated code looks like: if specified_level <= ::log::LOG_LEVEL { if ::log::module_enabled(module_path!()) { println!(...) } } Notably, the first layer of checking is still intended to be "super fast" in that it's just a load of a global word and a compare. The second layer of checking is executed to determine if the current module does indeed have logging turned on. This means that if any module has a debug log level turned on, all modules with debug log levels get a little bit slower (they all do more expensive dynamic checks to determine if they're turned on or not). Semantically, this migration brings no change in this respect, but runtime-wise, this will have a perf impact on some code. * A `RUST_LOG=::help` directive will no longer print out a list of all modules that can be logged. This is because the crate map will no longer specify the log levels of all modules, so the list of modules is not known. Additionally, warnings can no longer be provided if a malformed logging directive was supplied. The new "hello world" for logging looks like: #[phase(syntax, link)] extern crate log; fn main() { debug!("Hello, world!"); }
2014-03-09 06:11:44 +00:00
crate_id: from_str("test").unwrap(),
};
expand_crate(&ps,cfg,vec!(),vec!(),crate_ast)
}
// find the pat_ident paths in a crate
fn crate_bindings(the_crate : &ast::Crate) -> Vec<ast::Ident> {
let mut name_finder = PatIdentFinder{ident_accumulator:Vec::new()};
visit::walk_crate(&mut name_finder, the_crate, ());
name_finder.ident_accumulator
}
//fn expand_and_resolve(crate_str: @str) -> ast::crate {
//let expanded_ast = expand_crate_str(crate_str);
// println!("expanded: {:?}\n",expanded_ast);
2013-06-07 19:28:03 +00:00
//mtwt_resolve_crate(expanded_ast)
//}
//fn expand_and_resolve_and_pretty_print (crate_str: @str) -> String {
2013-06-07 19:28:03 +00:00
//let resolved_ast = expand_and_resolve(crate_str);
2013-06-07 01:09:31 +00:00
//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);}".to_string());
}
// 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.
//
2014-06-24 01:45:51 +00:00
// Put differently; this is a sparse representation of a boolean matrix
// indicating which bindings capture which identifiers.
//
// Note also that this matrix is dependent on the implicit ordering of
// the bindings and the varrefs discovered by the name-finder and the path-finder.
//
// The comparisons are done post-mtwt-resolve, so we're comparing renamed
// names; differences in marks don't matter any more.
2013-07-13 01:35:47 +00:00
//
// oog... I also want tests that check "bound-identifier-=?". That is,
2013-07-13 01:35:47 +00:00
// 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 RenamingTest = (&'static str, Vec<Vec<uint>>, bool);
#[test]
fn automatic_renaming () {
let tests: Vec<RenamingTest> =
vec!(// b & c should get new names throughout, in the expr too:
("fn a() -> int { let b = 13; let c = b; b+c }",
vec!(vec!(0,1),vec!(2)), false),
2013-06-07 22:01:28 +00:00
// both x's should be renamed (how is this causing a bug?)
2014-02-06 22:38:33 +00:00
("fn main () {let x: int = 13;x;}",
vec!(vec!(0)), false),
// the use of b after the + should be renamed, the other one not:
("macro_rules! f (($x:ident) => (b + $x)) fn a() -> int { let b = 13; f!(b)}",
vec!(vec!(1)), false),
2013-06-06 18:21:02 +00:00
// the b before the plus should not be renamed (requires marks)
("macro_rules! f (($x:ident) => ({let b=9; ($x + b)})) fn a() -> int { f!(b)}",
vec!(vec!(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); $x}))
fn main() -> int {user!(z)}",
vec!(vec!(0)), false)
);
for (idx,s) in tests.iter().enumerate() {
run_renaming_test(s,idx);
}
}
// no longer a fixme #8062: this test exposes a *potential* bug; our system does
// not behave exactly like MTWT, but a conversation with Matthew Flatt
// suggests that this can only occur in the presence of local-expand, which
// we have no plans to support. ... unless it's needed for item hygiene....
#[ignore]
#[test] fn issue_8062(){
run_renaming_test(
&("fn main() {let hrcoo = 19; macro_rules! getx(()=>(hrcoo)); getx!();}",
vec!(vec!(0)), true), 0)
}
// FIXME #6994:
// the z flows into and out of two macros (g & f) along one path, and one
// (just g) along the other, so the result of the whole thing should
// be "let z_123 = 3; z_123"
#[ignore]
#[test] fn issue_6994(){
run_renaming_test(
&("macro_rules! g (($x:ident) =>
({macro_rules! f(($y:ident)=>({let $y=3;$x}));f!($x)}))
fn a(){g!(z)}",
vec!(vec!(0)),false),
0)
2014-06-24 23:34:03 +00:00
}
// match variable hygiene. Should expand into
// fn z() {match 8 {x_1 => {match 9 {x_2 | x_2 if x_2 == x_1 => x_2 + x_1}}}}
2014-06-24 23:34:03 +00:00
#[test] fn issue_9384(){
run_renaming_test(
&("macro_rules! bad_macro (($ex:expr) => ({match 9 {x | x if x == $ex => x + $ex}}))
fn z() {match 8 {x => bad_macro!(x)}}",
2014-06-24 23:34:03 +00:00
// NB: the third "binding" is the repeat of the second one.
vec!(vec!(1,3),vec!(0,2),vec!(0,2)),
2014-06-24 23:34:03 +00:00
true),
0)
}
// interpolated nodes weren't getting labeled.
// should expand into
// fn main(){let g1_1 = 13; g1_1}}
#[test] fn pat_expand_issue_15221(){
run_renaming_test(
&("macro_rules! inner ( ($e:pat ) => ($e))
macro_rules! outer ( ($e:pat ) => (inner!($e)))
fn main() { let outer!(g) = 13; g;}",
vec!(vec!(0)),
true),
0)
}
// create a really evil test case where a $x appears inside a binding of $x
2014-07-03 01:27:07 +00:00
// but *shouldn't* bind because it was inserted by a different macro....
// can't write this test case until we have macro-generating macros.
2014-07-04 17:57:28 +00:00
// item fn hygiene
// expands to fn q(x_1:int){fn g(x_2:int){x_2 + x_1};}
2014-07-04 17:57:28 +00:00
#[test] fn issue_9383(){
run_renaming_test(
&("macro_rules! bad_macro (($ex:expr) => (fn g(x:int){ x + $ex }))
fn q(x:int) { bad_macro!(x); }",
vec!(vec!(1),vec!(0)),true),
0)
}
// closure arg hygiene (ExprFnBlock)
// expands to fn f(){(|x_1 : int| {(x_2 + x_1)})(3);}
#[test] fn closure_arg_hygiene(){
run_renaming_test(
&("macro_rules! inject_x (()=>(x))
fn f(){(|x : int| {(inject_x!() + x)})(3);}",
vec!(vec!(1)),
true),
0)
}
// closure arg hygiene (ExprProc)
// expands to fn f(){(proc(x_1 : int) {(x_2 + x_1)})(3);}
#[test] fn closure_arg_hygiene_2(){
run_renaming_test(
&("macro_rules! inject_x (()=>(x))
fn f(){ (proc(x : int){(inject_x!() + x)})(3); }",
vec!(vec!(1)),
true),
0)
}
2013-07-13 01:35:47 +00:00
// run one of the renaming tests
fn run_renaming_test(t: &RenamingTest, test_idx: uint) {
2013-09-05 16:29:31 +00:00
let invalid_name = token::special_idents::invalid.name;
2013-07-13 01:35:47 +00:00
let (teststr, bound_connections, bound_ident_check) = match *t {
(ref str,ref conns, bic) => (str.to_owned(), conns.clone(), bic)
};
let cr = expand_crate_str(teststr.to_string());
let bindings = crate_bindings(&cr);
let varrefs = crate_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.get(binding_idx));
let binding_marks = mtwt::marksof(bindings.get(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
2013-07-13 01:35:47 +00:00
// be free-identifier=? or bound-identifier=? to the given binding
assert_eq!(varref.segments.len(),1);
let varref_name = mtwt::resolve(varref.segments
.get(0)
.identifier);
let varref_marks = mtwt::marksof(varref.segments
.get(0)
.identifier
.ctxt,
invalid_name);
if !(varref_name==binding_name) {
let varref_idents : Vec<ast::Ident>
= varref.segments.iter().map(|s|
s.identifier)
.collect();
println!("uh oh, should match but doesn't:");
println!("varref #{}: {}",idx, varref_idents);
println!("binding #{}: {}", binding_idx, *bindings.get(binding_idx));
mtwt::with_sctable(|x| mtwt::display_sctable(x));
}
assert_eq!(varref_name,binding_name);
if bound_ident_check {
2013-09-05 16:29:31 +00:00
// we're checking bound-identifier=?, and the marks
// should be the same, too:
2013-07-13 01:35:47 +00:00
assert_eq!(varref_marks,binding_marks.clone());
}
} else {
let varref_name = mtwt::resolve(varref.segments.get(0).identifier);
let fail = (varref.segments.len() == 1)
&& (varref_name == binding_name);
// temp debugging:
if fail {
let varref_idents : Vec<ast::Ident>
= varref.segments.iter().map(|s|
s.identifier)
.collect();
println!("failure on test {}",test_idx);
println!("text of test case: \"{}\"", teststr);
println!("");
println!("uh oh, matches but shouldn't:");
println!("varref #{}: {}, resolves to {}",idx, varref_idents,
varref_name);
// good lord, you can't make a path with 0 segments, can you?
let string = token::get_ident(varref.segments
.get(0)
.identifier);
println!("varref's first segment's uint: {}, and string: \"{}\"",
varref.segments.get(0).identifier.name,
string.get());
println!("binding #{}: {}, resolves to {}",
binding_idx, *bindings.get(binding_idx),
binding_name);
mtwt::with_sctable(|x| mtwt::display_sctable(x));
}
assert!(!fail);
}
}
}
}
2013-07-13 01:35:47 +00:00
#[test] fn fmt_in_macro_used_inside_module_macro() {
2014-04-16 01:17:48 +00:00
let crate_str = "macro_rules! fmt_wrap(($b:expr)=>($b.to_str()))
2013-07-13 01:35:47 +00:00
macro_rules! foo_module (() => (mod generated { fn a() { let xx = 147; fmt_wrap!(xx);}}))
foo_module!()
".to_string();
2013-07-13 01:35:47 +00:00
let cr = expand_crate_str(crate_str);
// find the xx binding
let bindings = crate_bindings(&cr);
let cxbinds: Vec<&ast::Ident> =
bindings.iter().filter(|b| {
let ident = token::get_ident(**b);
let string = ident.get();
"xx" == string
}).collect();
let cxbinds: &[&ast::Ident] = cxbinds.as_slice();
2013-07-13 01:35:47 +00:00
let cxbind = match cxbinds {
[b] => b,
_ => fail!("expected just one binding for ext_cx")
2013-07-13 01:35:47 +00:00
};
let resolved_binding = mtwt::resolve(*cxbind);
let varrefs = crate_varrefs(&cr);
2013-07-13 01:35:47 +00:00
// the xx binding should bind all of the xx varrefs:
for (idx,v) in varrefs.iter().filter(|p| {
p.segments.len() == 1
&& "xx" == token::get_ident(p.segments.get(0).identifier).get()
}).enumerate() {
if mtwt::resolve(v.segments.get(0).identifier) != resolved_binding {
println!("uh oh, xx binding didn't match xx varref:");
println!("this is xx varref \\# {:?}",idx);
println!("binding: {:?}",cxbind);
println!("resolves to: {:?}",resolved_binding);
println!("varref: {:?}",v.segments.get(0).identifier);
println!("resolves to: {:?}",
mtwt::resolve(v.segments.get(0).identifier));
mtwt::with_sctable(|x| mtwt::display_sctable(x));
2013-07-13 01:35:47 +00:00
}
assert_eq!(mtwt::resolve(v.segments.get(0).identifier),
resolved_binding);
2013-07-13 01:35:47 +00:00
};
}
2013-06-04 21:56:33 +00:00
#[test]
fn pat_idents(){
let pat = string_to_pat(
"(a,Foo{x:c @ (b,9),y:Bar(4,d)})".to_string());
let idents = pattern_bindings(pat);
assert_eq!(idents, strs_to_idents(vec!("a","c","b","d")));
2013-06-04 21:56:33 +00:00
}
2013-06-07 19:28:03 +00:00
// test the list of identifier patterns gathered by the visitor. Note that
// 'None' is listed as an identifier pattern because we don't yet know that
// it's the name of a 0-ary variant, and that 'i' appears twice in succession.
#[test]
fn crate_bindings_test(){
let the_crate = string_to_crate("fn main (a : int) -> int {|b| {
match 34 {None => 3, Some(i) | i => j, Foo{k:z,l:y} => \"banana\"}} }".to_string());
let idents = crate_bindings(&the_crate);
assert_eq!(idents, strs_to_idents(vec!("a","b","None","i","i","z","y")));
}
// test the IdentRenamer directly
#[test]
fn ident_renamer_test () {
let the_crate = string_to_crate("fn f(x : int){let x = x; x}".to_string());
let f_ident = token::str_to_ident("f");
let x_ident = token::str_to_ident("x");
let int_ident = token::str_to_ident("int");
let renames = vec!((x_ident,16));
let mut renamer = IdentRenamer{renames: &renames};
let renamed_crate = renamer.fold_crate(the_crate);
let idents = crate_idents(&renamed_crate);
let resolved : Vec<ast::Name> = idents.iter().map(|id| mtwt::resolve(*id)).collect();
assert_eq!(resolved,vec!(f_ident.name,16,int_ident.name,16,16,16));
}
// test the PatIdentRenamer; only PatIdents get renamed
#[test]
fn pat_ident_renamer_test () {
let the_crate = string_to_crate("fn f(x : int){let x = x; x}".to_string());
let f_ident = token::str_to_ident("f");
let x_ident = token::str_to_ident("x");
let int_ident = token::str_to_ident("int");
let renames = vec!((x_ident,16));
let mut renamer = PatIdentRenamer{renames: &renames};
let renamed_crate = renamer.fold_crate(the_crate);
let idents = crate_idents(&renamed_crate);
let resolved : Vec<ast::Name> = idents.iter().map(|id| mtwt::resolve(*id)).collect();
let x_name = x_ident.name;
assert_eq!(resolved,vec!(f_ident.name,16,int_ident.name,16,x_name,x_name));
}
}