rust/src/librustc/middle/region.rs

971 lines
33 KiB
Rust
Raw Normal View History

// Copyright 2012 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.
/*!
This file actually contains two passes related to regions. The first
pass builds up the `scope_map`, which describes the parent links in
the region hierarchy. The second pass infers which types must be
region parameterized.
*/
2012-03-09 18:39:54 -06:00
use driver::session::Session;
2012-09-04 13:54:36 -05:00
use metadata::csearch;
use middle::resolve;
use middle::ty::{region_variance, rv_covariant, rv_invariant};
use middle::ty::{rv_contravariant, FreeRegion};
use middle::ty;
2012-09-04 13:54:36 -05:00
use core::hashmap::{HashMap, HashSet};
use syntax::ast_map;
use syntax::codemap::span;
use syntax::print::pprust;
use syntax::parse::token::special_idents;
use syntax::{ast, visit};
2012-03-09 18:39:54 -06:00
pub type parent = Option<ast::node_id>;
2012-03-09 18:39:54 -06:00
/**
The region maps encode information about region relationships.
- `scope_map` maps from:
- an expression to the expression or block encoding the maximum
(static) lifetime of a value produced by that expression. This is
generally the innermost call, statement, match, or block.
- a variable or binding id to the block in which that variable is declared.
- `free_region_map` maps from:
- a free region `a` to a list of free regions `bs` such that
`a <= b for all b in bs`
- the free region map is populated during type check as we check
each function. See the function `relate_free_regions` for
more information.
2013-03-15 14:24:24 -05:00
- `cleanup_scopes` includes scopes where trans cleanups occur
- this is intended to reflect the current state of trans, not
necessarily how I think things ought to work
*/
pub struct RegionMaps {
priv scope_map: HashMap<ast::node_id, ast::node_id>,
priv free_region_map: HashMap<FreeRegion, ~[FreeRegion]>,
2013-03-15 14:24:24 -05:00
priv cleanup_scopes: HashSet<ast::node_id>
}
2013-03-15 14:24:24 -05:00
pub struct Context {
sess: Session,
def_map: resolve::DefMap,
// Generated maps:
region_maps: @mut RegionMaps,
2013-03-15 14:24:24 -05:00
// Scope where variables should be parented to
var_parent: parent,
// Innermost enclosing expression
parent: parent,
}
pub impl RegionMaps {
fn relate_free_regions(&mut self,
sub: FreeRegion,
sup: FreeRegion)
{
match self.free_region_map.find_mut(&sub) {
Some(sups) => {
if !sups.contains(&sup) {
sups.push(sup);
}
return;
}
None => {}
}
debug!("relate_free_regions(sub=%?, sup=%?)", sub, sup);
self.free_region_map.insert(sub, ~[sup]);
}
fn record_parent(&mut self,
sub: ast::node_id,
sup: ast::node_id)
{
debug!("record_parent(sub=%?, sup=%?)", sub, sup);
2013-03-15 14:24:24 -05:00
assert!(sub != sup);
self.scope_map.insert(sub, sup);
}
2013-03-15 14:24:24 -05:00
pub fn record_cleanup_scope(&mut self,
scope_id: ast::node_id)
{
//! Records that a scope is a CLEANUP SCOPE. This is invoked
//! from within regionck. We wait until regionck because we do
//! not know which operators are overloaded until that point,
//! and only overloaded operators result in cleanup scopes.
self.cleanup_scopes.insert(scope_id);
}
fn opt_encl_scope(&self,
id: ast::node_id) -> Option<ast::node_id>
{
//! Returns the narrowest scope that encloses `id`, if any.
self.scope_map.find(&id).map(|&x| *x)
}
fn encl_scope(&self,
id: ast::node_id) -> ast::node_id
{
//! Returns the narrowest scope that encloses `id`, if any.
match self.scope_map.find(&id) {
Some(&r) => r,
None => { fail!(fmt!("No enclosing scope for id %?", id)); }
}
}
2013-03-15 14:24:24 -05:00
fn is_cleanup_scope(&self, scope_id: ast::node_id) -> bool {
self.cleanup_scopes.contains(&scope_id)
}
fn cleanup_scope(&self,
expr_id: ast::node_id) -> ast::node_id
{
//! Returns the scope when temps in expr will be cleaned up
let mut id = self.encl_scope(expr_id);
while !self.cleanup_scopes.contains(&id) {
id = self.encl_scope(id);
}
return id;
}
fn encl_region(&self,
id: ast::node_id) -> ty::Region
{
//! Returns the narrowest scope region that encloses `id`, if any.
ty::re_scope(self.encl_scope(id))
}
2013-03-15 14:24:24 -05:00
pub fn scopes_intersect(&self,
scope1: ast::node_id,
scope2: ast::node_id) -> bool
{
self.is_subscope_of(scope1, scope2) || self.is_subscope_of(scope2, scope1)
}
fn is_subscope_of(&self,
subscope: ast::node_id,
superscope: ast::node_id) -> bool
{
/*!
2013-03-15 14:24:24 -05:00
* Returns true if `subscope` is equal to or is lexically
* nested inside `superscope` and false otherwise.
*/
2013-03-15 14:24:24 -05:00
let mut s = subscope;
while superscope != s {
match self.scope_map.find(&s) {
None => {
debug!("is_subscope_of(%?, %?, s=%?)=false",
subscope, superscope, s);
return false;
}
Some(&scope) => s = scope
}
}
2013-03-15 14:24:24 -05:00
debug!("is_subscope_of(%?, %?)=true",
subscope, superscope);
return true;
}
fn sub_free_region(&self,
sub: FreeRegion,
sup: FreeRegion) -> bool
{
/*!
* Determines whether two free regions have a subregion relationship
* by walking the graph encoded in `free_region_map`. Note that
* it is possible that `sub != sup` and `sub <= sup` and `sup <= sub`
* (that is, the user can give two different names to the same lifetime).
*/
if sub == sup {
return true;
}
2012-03-26 17:07:15 -05:00
// Do a little breadth-first-search here. The `queue` list
// doubles as a way to detect if we've seen a particular FR
// before. Note that we expect this graph to be an *extremely
// shallow* tree.
let mut queue = ~[sub];
let mut i = 0;
while i < queue.len() {
match self.free_region_map.find(&queue[i]) {
Some(parents) => {
for parents.each |parent| {
if *parent == sup {
return true;
}
if !queue.contains(parent) {
queue.push(*parent);
}
}
2012-03-26 17:07:15 -05:00
}
None => {}
2012-03-26 17:07:15 -05:00
}
i += 1;
2012-03-26 17:07:15 -05:00
}
return false;
2012-03-26 17:07:15 -05:00
}
fn is_subregion_of(&self,
sub_region: ty::Region,
super_region: ty::Region) -> bool
{
/*!
* Determines whether one region is a subregion of another. This is
* intended to run *after inference* and sadly the logic is somewhat
* duplicated with the code in infer.rs.
*/
debug!("is_subregion_of(sub_region=%?, super_region=%?)",
sub_region, super_region);
sub_region == super_region || {
match (sub_region, super_region) {
(_, ty::re_static) => {
true
}
(ty::re_scope(sub_scope), ty::re_scope(super_scope)) => {
2013-03-15 14:24:24 -05:00
self.is_subscope_of(sub_scope, super_scope)
}
2012-03-26 17:07:15 -05:00
(ty::re_scope(sub_scope), ty::re_free(ref fr)) => {
2013-03-15 14:24:24 -05:00
self.is_subscope_of(sub_scope, fr.scope_id)
}
(ty::re_free(sub_fr), ty::re_free(super_fr)) => {
self.sub_free_region(sub_fr, super_fr)
}
_ => {
false
}
}
}
}
fn nearest_common_ancestor(&self,
scope_a: ast::node_id,
scope_b: ast::node_id) -> Option<ast::node_id>
{
/*!
* Finds the nearest common ancestor (if any) of two scopes. That
* is, finds the smallest scope which is greater than or equal to
* both `scope_a` and `scope_b`.
*/
if scope_a == scope_b { return Some(scope_a); }
let a_ancestors = ancestors_of(self, scope_a);
let b_ancestors = ancestors_of(self, scope_b);
let mut a_index = vec::len(a_ancestors) - 1u;
let mut b_index = vec::len(b_ancestors) - 1u;
// Here, ~[ab]_ancestors is a vector going from narrow to broad.
// The end of each vector will be the item where the scope is
// defined; if there are any common ancestors, then the tails of
// the vector will be the same. So basically we want to walk
// backwards from the tail of each vector and find the first point
// where they diverge. If one vector is a suffix of the other,
// then the corresponding scope is a superscope of the other.
if a_ancestors[a_index] != b_ancestors[b_index] {
return None;
}
loop {
// Loop invariant: a_ancestors[a_index] == b_ancestors[b_index]
// for all indices between a_index and the end of the array
if a_index == 0u { return Some(scope_a); }
if b_index == 0u { return Some(scope_b); }
a_index -= 1u;
b_index -= 1u;
if a_ancestors[a_index] != b_ancestors[b_index] {
return Some(a_ancestors[a_index + 1u]);
}
}
fn ancestors_of(this: &RegionMaps, scope: ast::node_id)
-> ~[ast::node_id]
{
2013-03-15 14:24:24 -05:00
// debug!("ancestors_of(scope=%d)", scope);
let mut result = ~[scope];
let mut scope = scope;
loop {
match this.scope_map.find(&scope) {
None => return result,
Some(&superscope) => {
result.push(superscope);
scope = superscope;
}
}
2013-03-15 14:24:24 -05:00
// debug!("ancestors_of_loop(scope=%d)", scope);
}
}
2012-03-26 17:07:15 -05:00
}
}
/// Extracts that current parent from cx, failing if there is none.
2013-03-15 14:24:24 -05:00
pub fn parent_id(cx: Context, span: span) -> ast::node_id {
2012-08-06 14:34:08 -05:00
match cx.parent {
2012-08-20 14:23:37 -05:00
None => {
2013-04-30 11:47:52 -05:00
cx.sess.span_bug(span, "crate should not be parent here");
}
2012-08-20 14:23:37 -05:00
Some(parent_id) => {
parent_id
}
}
}
/// Records the current parent (if any) as the parent of `child_id`.
2013-03-15 14:24:24 -05:00
pub fn parent_to_expr(cx: Context, child_id: ast::node_id) {
for cx.parent.each |parent_id| {
cx.region_maps.record_parent(child_id, *parent_id);
2012-03-09 18:39:54 -06:00
}
}
2013-03-15 14:24:24 -05:00
pub fn resolve_block(blk: &ast::blk, cx: Context, visitor: visit::vt<Context>) {
// Record the parent of this block.
2013-03-15 14:24:24 -05:00
parent_to_expr(cx, blk.node.id);
2012-03-09 18:39:54 -06:00
// Descend.
2013-03-15 14:24:24 -05:00
let new_cx = Context {var_parent: Some(blk.node.id),
parent: Some(blk.node.id),
..cx};
2012-03-09 18:39:54 -06:00
visit::visit_block(blk, new_cx, visitor);
}
2013-03-15 14:24:24 -05:00
pub fn resolve_arm(arm: &ast::arm, cx: Context, visitor: visit::vt<Context>) {
visit::visit_arm(arm, cx, visitor);
}
2013-03-15 14:24:24 -05:00
pub fn resolve_pat(pat: @ast::pat, cx: Context, visitor: visit::vt<Context>) {
assert!(cx.var_parent == cx.parent);
parent_to_expr(cx, pat.id);
visit::visit_pat(pat, cx, visitor);
}
2013-03-15 14:24:24 -05:00
pub fn resolve_stmt(stmt: @ast::stmt, cx: Context, visitor: visit::vt<Context>) {
match stmt.node {
2013-03-15 14:24:24 -05:00
ast::stmt_decl(*) => {
visit::visit_stmt(stmt, cx, visitor);
}
ast::stmt_expr(_, stmt_id) |
ast::stmt_semi(_, stmt_id) => {
parent_to_expr(cx, stmt_id);
let expr_cx = Context {parent: Some(stmt_id), ..cx};
visit::visit_stmt(stmt, expr_cx, visitor);
}
ast::stmt_mac(*) => cx.sess.bug(~"unexpanded macro")
}
}
2013-03-15 14:24:24 -05:00
pub fn resolve_expr(expr: @ast::expr, cx: Context, visitor: visit::vt<Context>) {
parent_to_expr(cx, expr.id);
let mut new_cx = cx;
2013-03-15 14:24:24 -05:00
new_cx.parent = Some(expr.id);
2012-08-06 14:34:08 -05:00
match expr.node {
2013-03-15 14:24:24 -05:00
ast::expr_assign_op(*) | ast::expr_index(*) | ast::expr_binary(*) |
ast::expr_unary(*) | ast::expr_call(*) | ast::expr_method_call(*) => {
2013-05-06 13:02:28 -05:00
// FIXME(#6268) Nested method calls
2013-03-15 14:24:24 -05:00
//
// The lifetimes for a call or method call look as follows:
//
// call.id
// - arg0.id
// - ...
// - argN.id
// - call.callee_id
//
// The idea is that call.callee_id represents *the time when
// the invoked function is actually running* and call.id
// represents *the time to prepare the arguments and make the
// call*. See the section "Borrows in Calls" borrowck/doc.rs
// for an extended explanantion of why this distinction is
// important.
//
// parent_to_expr(new_cx, expr.callee_id);
}
ast::expr_match(*) => {
new_cx.var_parent = Some(expr.id);
}
_ => {}
};
visit::visit_expr(expr, new_cx, visitor);
}
pub fn resolve_local(local: @ast::local,
2013-03-15 14:24:24 -05:00
cx: Context,
visitor: visit::vt<Context>) {
assert!(cx.var_parent == cx.parent);
parent_to_expr(cx, local.node.id);
visit::visit_local(local, cx, visitor);
}
2013-03-15 14:24:24 -05:00
pub fn resolve_item(item: @ast::item, cx: Context, visitor: visit::vt<Context>) {
2012-03-09 18:39:54 -06:00
// Items create a new outer block scope as far as we're concerned.
2013-03-15 14:24:24 -05:00
let new_cx = Context {var_parent: None, parent: None, ..cx};
2012-03-09 18:39:54 -06:00
visit::visit_item(item, new_cx, visitor);
}
pub fn resolve_fn(fk: &visit::fn_kind,
decl: &ast::fn_decl,
body: &ast::blk,
2013-03-15 14:24:24 -05:00
_sp: span,
id: ast::node_id,
2013-03-15 14:24:24 -05:00
cx: Context,
visitor: visit::vt<Context>) {
debug!("region::resolve_fn(id=%?, body.node.id=%?, cx.parent=%?)",
id, body.node.id, cx.parent);
// The arguments and `self` are parented to the body of the fn.
let decl_cx = Context {parent: Some(body.node.id),
var_parent: Some(body.node.id),
..cx};
match *fk {
visit::fk_method(_, _, method) => {
cx.region_maps.record_parent(method.self_id, body.node.id);
}
_ => {}
}
2013-03-15 14:24:24 -05:00
visit::visit_fn_decl(decl, decl_cx, visitor);
2013-03-15 14:24:24 -05:00
// The body of the fn itself is either a root scope (top-level fn)
// or it continues with the inherited scope (closures).
let body_cx = match *fk {
visit::fk_item_fn(*) |
visit::fk_method(*) => {
2013-03-15 14:24:24 -05:00
Context {parent: None, var_parent: None, ..cx}
}
visit::fk_anon(*) |
visit::fk_fn_block(*) => {
cx
}
};
(visitor.visit_block)(body, body_cx, visitor);
}
pub fn resolve_crate(sess: Session,
def_map: resolve::DefMap,
crate: @ast::crate) -> @mut RegionMaps
{
let region_maps = @mut RegionMaps {
scope_map: HashMap::new(),
2013-03-15 14:24:24 -05:00
free_region_map: HashMap::new(),
cleanup_scopes: HashSet::new(),
};
2013-03-15 14:24:24 -05:00
let cx = Context {sess: sess,
def_map: def_map,
region_maps: region_maps,
parent: None,
var_parent: None};
let visitor = visit::mk_vt(@visit::Visitor {
2012-03-09 18:39:54 -06:00
visit_block: resolve_block,
visit_item: resolve_item,
visit_fn: resolve_fn,
visit_arm: resolve_arm,
visit_pat: resolve_pat,
visit_stmt: resolve_stmt,
visit_expr: resolve_expr,
2012-09-04 15:29:32 -05:00
visit_local: resolve_local,
.. *visit::default_visitor()
2012-03-09 18:39:54 -06:00
});
visit::visit_crate(crate, cx, visitor);
return region_maps;
2012-03-09 18:39:54 -06:00
}
// ___________________________________________________________________________
// Determining region parameterization
//
// Infers which type defns must be region parameterized---this is done
// by scanning their contents to see whether they reference a region
// type, directly or indirectly. This is a fixed-point computation.
//
// We do it in two passes. First we walk the AST and construct a map
// from each type defn T1 to other defns which make use of it. For example,
// if we have a type like:
//
// type S = *int;
// type T = S;
//
// Then there would be a map entry from S to T. During the same walk,
// we also construct add any types that reference regions to a set and
// a worklist. We can then process the worklist, propagating indirect
// dependencies until a fixed point is reached.
pub type region_paramd_items = @mut HashMap<ast::node_id, region_variance>;
#[deriving(Eq)]
pub struct region_dep {
ambient_variance: region_variance,
id: ast::node_id
}
2012-08-27 18:26:35 -05:00
pub type dep_map = @mut HashMap<ast::node_id, @mut ~[region_dep]>;
pub struct DetermineRpCtxt {
sess: Session,
ast_map: ast_map::map,
2012-08-29 15:26:26 -05:00
def_map: resolve::DefMap,
region_paramd_items: region_paramd_items,
dep_map: dep_map,
worklist: ~[ast::node_id],
2012-07-12 16:01:23 -05:00
// the innermost enclosing item id
item_id: ast::node_id,
2012-07-12 16:01:23 -05:00
// true when we are within an item but not within a method.
// see long discussion on region_is_relevant().
anon_implies_rp: bool,
// encodes the context of the current type; invariant if
// mutable, covariant otherwise
ambient_variance: region_variance,
}
pub fn join_variance(variance1: region_variance,
variance2: region_variance)
-> region_variance {
match (variance1, variance2) {
(rv_invariant, _) => {rv_invariant}
(_, rv_invariant) => {rv_invariant}
(rv_covariant, rv_contravariant) => {rv_invariant}
(rv_contravariant, rv_covariant) => {rv_invariant}
(rv_covariant, rv_covariant) => {rv_covariant}
(rv_contravariant, rv_contravariant) => {rv_contravariant}
}
}
/// Combines the ambient variance with the variance of a
/// particular site to yield the final variance of the reference.
///
/// Example: if we are checking function arguments then the ambient
/// variance is contravariant. If we then find a `&'r T` pointer, `r`
/// appears in a co-variant position. This implies that this
/// occurrence of `r` is contra-variant with respect to the current
/// item, and hence the function returns `rv_contravariant`.
pub fn add_variance(ambient_variance: region_variance,
variance: region_variance)
-> region_variance {
match (ambient_variance, variance) {
(rv_invariant, _) => rv_invariant,
(_, rv_invariant) => rv_invariant,
(rv_covariant, c) => c,
(c, rv_covariant) => c,
(rv_contravariant, rv_contravariant) => rv_covariant
}
}
pub impl DetermineRpCtxt {
fn add_variance(&self, variance: region_variance) -> region_variance {
add_variance(self.ambient_variance, variance)
}
/// Records that item `id` is region-parameterized with the
/// variance `variance`. If `id` was already parameterized, then
/// the new variance is joined with the old variance.
fn add_rp(&mut self, id: ast::node_id, variance: region_variance) {
2013-03-28 20:39:09 -05:00
assert!(id != 0);
2013-03-22 21:26:41 -05:00
let old_variance = self.region_paramd_items.find(&id).
map_consume(|x| *x);
let joined_variance = match old_variance {
2012-08-20 14:23:37 -05:00
None => variance,
Some(v) => join_variance(v, variance)
};
2012-08-22 19:24:52 -05:00
debug!("add_rp() variance for %s: %? == %? ^ %?",
ast_map::node_id_to_str(self.ast_map, id,
self.sess.parse_sess.interner),
2012-08-22 19:24:52 -05:00
joined_variance, old_variance, variance);
2012-08-20 14:23:37 -05:00
if Some(joined_variance) != old_variance {
let region_paramd_items = self.region_paramd_items;
region_paramd_items.insert(id, joined_variance);
self.worklist.push(id);
}
}
/// Indicates that the region-parameterization of the current item
/// is dependent on the region-parameterization of the item
/// `from`. Put another way, it indicates that the current item
/// contains a value of type `from`, so if `from` is
/// region-parameterized, so is the current item.
fn add_dep(&mut self, from: ast::node_id) {
2012-08-22 19:24:52 -05:00
debug!("add dependency from %d -> %d (%s -> %s) with variance %?",
from, self.item_id,
2012-07-18 18:18:02 -05:00
ast_map::node_id_to_str(self.ast_map, from,
self.sess.parse_sess.interner),
ast_map::node_id_to_str(self.ast_map, self.item_id,
self.sess.parse_sess.interner),
2012-08-22 19:24:52 -05:00
copy self.ambient_variance);
let vec = match self.dep_map.find(&from) {
2013-03-22 21:26:41 -05:00
Some(&vec) => vec,
2012-08-20 14:23:37 -05:00
None => {
2013-03-07 17:37:14 -06:00
let vec = @mut ~[];
let dep_map = self.dep_map;
dep_map.insert(from, vec);
vec
}
};
let dep = region_dep {
ambient_variance: self.ambient_variance,
id: self.item_id
};
2012-09-25 19:39:22 -05:00
if !vec.contains(&dep) { vec.push(dep); }
}
2012-07-12 16:01:23 -05:00
// Determines whether a reference to a region that appears in the
// AST implies that the enclosing type is region-parameterized (RP).
// This point is subtle. Here are some examples to make it more
2012-07-12 16:01:23 -05:00
// concrete.
//
// 1. impl foo for &int { ... }
// 2. impl foo for &'self int { ... }
// 3. impl foo for bar { fn m(@self) -> &'self int { ... } }
// 4. impl foo for bar { fn m(&self) -> &'self int { ... } }
// 5. impl foo for bar { fn m(&self) -> &int { ... } }
2012-07-12 16:01:23 -05:00
//
// In case 1, the anonymous region is being referenced,
// but it appears in a context where the anonymous region
// resolves to self, so the impl foo is RP.
2012-07-12 16:01:23 -05:00
//
// In case 2, the self parameter is written explicitly.
//
// In case 3, the method refers to the region `self`, so that
// implies that the impl must be region parameterized. (If the
// type bar is not region parameterized, that is an error, because
// the self region is effectively unconstrained, but that is
// detected elsewhere).
//
// In case 4, the method refers to the region `self`, but the
// `self` region is bound by the `&self` receiver, and so this
// does not require that `bar` be RP.
2012-07-12 16:01:23 -05:00
//
// In case 5, the anonymous region is referenced, but it
2012-07-12 16:01:23 -05:00
// bound by the method, so it does not refer to self. This impl
// need not be region parameterized.
//
// Normally, & or &self implies that the enclosing item is RP.
// However, within a function, & is always bound. Within a method
// with &self type, &self is also bound. We detect those last two
// cases via flags (anon_implies_rp and self_implies_rp) that are
// true when the anon or self region implies RP.
fn region_is_relevant(&self, r: Option<@ast::Lifetime>) -> bool {
match r {
None => {
self.anon_implies_rp
}
Some(ref l) if l.ident == special_idents::statik => {
false
}
Some(ref l) if l.ident == special_idents::self_ => {
true
}
Some(_) => {
false
}
}
}
fn with(@mut self,
item_id: ast::node_id,
anon_implies_rp: bool,
f: &fn()) {
let old_item_id = self.item_id;
let old_anon_implies_rp = self.anon_implies_rp;
self.item_id = item_id;
self.anon_implies_rp = anon_implies_rp;
debug!("with_item_id(%d, %b)",
item_id,
anon_implies_rp);
let _i = ::util::common::indenter();
f();
self.item_id = old_item_id;
self.anon_implies_rp = old_anon_implies_rp;
}
fn with_ambient_variance(@mut self, variance: region_variance, f: &fn()) {
let old_ambient_variance = self.ambient_variance;
self.ambient_variance = self.add_variance(variance);
f();
self.ambient_variance = old_ambient_variance;
}
}
pub fn determine_rp_in_item(item: @ast::item,
cx: @mut DetermineRpCtxt,
visitor: visit::vt<@mut DetermineRpCtxt>) {
do cx.with(item.id, true) {
visit::visit_item(item, cx, visitor);
}
}
pub fn determine_rp_in_fn(fk: &visit::fn_kind,
decl: &ast::fn_decl,
body: &ast::blk,
_: span,
_: ast::node_id,
cx: @mut DetermineRpCtxt,
visitor: visit::vt<@mut DetermineRpCtxt>) {
do cx.with(cx.item_id, false) {
do cx.with_ambient_variance(rv_contravariant) {
for decl.inputs.each |a| {
(visitor.visit_ty)(a.ty, cx, visitor);
}
}
(visitor.visit_ty)(decl.output, cx, visitor);
let generics = visit::generics_of_fn(fk);
(visitor.visit_generics)(&generics, cx, visitor);
(visitor.visit_block)(body, cx, visitor);
}
}
pub fn determine_rp_in_ty_method(ty_m: &ast::ty_method,
cx: @mut DetermineRpCtxt,
visitor: visit::vt<@mut DetermineRpCtxt>) {
do cx.with(cx.item_id, false) {
visit::visit_ty_method(ty_m, cx, visitor);
}
}
pub fn determine_rp_in_ty(ty: @ast::Ty,
cx: @mut DetermineRpCtxt,
visitor: visit::vt<@mut DetermineRpCtxt>) {
// we are only interested in types that will require an item to
// be region-parameterized. if cx.item_id is zero, then this type
// is not a member of a type defn nor is it a constitutent of an
// impl etc. So we can ignore it and its components.
2012-08-01 19:30:05 -05:00
if cx.item_id == 0 { return; }
// if this type directly references a region pointer like &'r ty,
// add to the worklist/set. Note that &'r ty is contravariant with
// respect to &r, because &'r ty can be used whereever a *smaller*
// region is expected (and hence is a supertype of those
// locations)
let sess = cx.sess;
2012-08-06 14:34:08 -05:00
match ty.node {
ast::ty_rptr(r, _) => {
debug!("referenced rptr type %s",
pprust::ty_to_str(ty, sess.intr()));
if cx.region_is_relevant(r) {
2013-03-15 14:24:24 -05:00
let rv = cx.add_variance(rv_contravariant);
cx.add_rp(cx.item_id, rv)
}
}
ast::ty_closure(ref f) => {
debug!("referenced fn type: %s",
pprust::ty_to_str(ty, sess.intr()));
match f.region {
Some(_) => {
if cx.region_is_relevant(f.region) {
2013-03-15 14:24:24 -05:00
let rv = cx.add_variance(rv_contravariant);
cx.add_rp(cx.item_id, rv)
}
}
None => {
if f.sigil == ast::BorrowedSigil && cx.anon_implies_rp {
2013-03-15 14:24:24 -05:00
let rv = cx.add_variance(rv_contravariant);
cx.add_rp(cx.item_id, rv)
}
}
}
}
_ => {}
}
// if this references another named type, add the dependency
// to the dep_map. If the type is not defined in this crate,
// then check whether it is region-parameterized and consider
// that as a direct dependency.
2012-08-06 14:34:08 -05:00
match ty.node {
ast::ty_path(path, id) => {
match cx.def_map.find(&id) {
Some(&ast::def_ty(did)) |
Some(&ast::def_trait(did)) |
Some(&ast::def_struct(did)) => {
if did.crate == ast::local_crate {
if cx.region_is_relevant(path.rp) {
cx.add_dep(did.node);
}
} else {
let cstore = sess.cstore;
match csearch::get_region_param(cstore, did) {
2012-08-20 14:23:37 -05:00
None => {}
Some(variance) => {
2012-08-22 19:24:52 -05:00
debug!("reference to external, rp'd type %s",
pprust::ty_to_str(ty, sess.intr()));
if cx.region_is_relevant(path.rp) {
2013-03-15 14:24:24 -05:00
let rv = cx.add_variance(variance);
cx.add_rp(cx.item_id, rv)
}
}
}
}
}
2012-08-03 21:59:04 -05:00
_ => {}
}
}
2012-08-03 21:59:04 -05:00
_ => {}
}
2012-08-06 14:34:08 -05:00
match ty.node {
ast::ty_box(mt) | ast::ty_uniq(mt) | ast::ty_vec(mt) |
ast::ty_rptr(_, mt) | ast::ty_ptr(mt) => {
visit_mt(mt, cx, visitor);
}
ast::ty_path(path, _) => {
// type parameters are---for now, anyway---always invariant
do cx.with_ambient_variance(rv_invariant) {
for path.types.each |tp| {
(visitor.visit_ty)(*tp, cx, visitor);
}
}
}
ast::ty_closure(@ast::TyClosure {decl: ref decl, _}) |
ast::ty_bare_fn(@ast::TyBareFn {decl: ref decl, _}) => {
// fn() binds the & region, so do not consider &T types that
// appear *inside* a fn() type to affect the enclosing item:
do cx.with(cx.item_id, false) {
// parameters are contravariant
do cx.with_ambient_variance(rv_contravariant) {
for decl.inputs.each |a| {
(visitor.visit_ty)(a.ty, cx, visitor);
}
}
(visitor.visit_ty)(decl.output, cx, visitor);
}
}
_ => {
visit::visit_ty(ty, cx, visitor);
}
}
fn visit_mt(mt: ast::mt,
cx: @mut DetermineRpCtxt,
visitor: visit::vt<@mut DetermineRpCtxt>) {
// mutability is invariant
if mt.mutbl == ast::m_mutbl {
do cx.with_ambient_variance(rv_invariant) {
(visitor.visit_ty)(mt.ty, cx, visitor);
}
} else {
(visitor.visit_ty)(mt.ty, cx, visitor);
}
}
}
pub fn determine_rp_in_struct_field(
cm: @ast::struct_field,
cx: @mut DetermineRpCtxt,
visitor: visit::vt<@mut DetermineRpCtxt>) {
visit::visit_struct_field(cm, cx, visitor);
}
pub fn determine_rp_in_crate(sess: Session,
ast_map: ast_map::map,
def_map: resolve::DefMap,
crate: @ast::crate)
-> region_paramd_items {
let cx = @mut DetermineRpCtxt {
sess: sess,
ast_map: ast_map,
def_map: def_map,
region_paramd_items: @mut HashMap::new(),
dep_map: @mut HashMap::new(),
worklist: ~[],
item_id: 0,
anon_implies_rp: false,
ambient_variance: rv_covariant
};
// Gather up the base set, worklist and dep_map
let visitor = visit::mk_vt(@visit::Visitor {
visit_fn: determine_rp_in_fn,
visit_item: determine_rp_in_item,
visit_ty: determine_rp_in_ty,
visit_ty_method: determine_rp_in_ty_method,
visit_struct_field: determine_rp_in_struct_field,
2012-09-04 15:29:32 -05:00
.. *visit::default_visitor()
});
visit::visit_crate(crate, cx, visitor);
// Propagate indirect dependencies
//
// Each entry in the worklist is the id of an item C whose region
// parameterization has been updated. So we pull ids off of the
// worklist, find the current variance, and then iterate through
// all of the dependent items (that is, those items that reference
// C). For each dependent item D, we combine the variance of C
// with the ambient variance where the reference occurred and then
// update the region-parameterization of D to reflect the result.
{
let cx = &mut *cx;
while cx.worklist.len() != 0 {
let c_id = cx.worklist.pop();
let c_variance = cx.region_paramd_items.get_copy(&c_id);
debug!("popped %d from worklist", c_id);
match cx.dep_map.find(&c_id) {
None => {}
Some(deps) => {
for deps.each |dep| {
let v = add_variance(dep.ambient_variance, c_variance);
cx.add_rp(dep.id, v);
}
}
}
}
}
debug!("%s", {
debug!("Region variance results:");
let region_paramd_items = cx.region_paramd_items;
for region_paramd_items.each |&key, &value| {
debug!("item %? (%s) is parameterized with variance %?",
key,
ast_map::node_id_to_str(ast_map, key,
sess.parse_sess.interner),
value);
}
"----"
});
// return final set
2012-08-01 19:30:05 -05:00
return cx.region_paramd_items;
}