mirror of
https://github.com/apple/swift.git
synced 2025-12-14 20:36:38 +01:00
799 lines
26 KiB
C++
799 lines
26 KiB
C++
//===--- RewriteSystem.cpp - Generics with term rewriting -----------------===//
|
|
//
|
|
// This source file is part of the Swift.org open source project
|
|
//
|
|
// Copyright (c) 2021 Apple Inc. and the Swift project authors
|
|
// Licensed under Apache License v2.0 with Runtime Library Exception
|
|
//
|
|
// See https://swift.org/LICENSE.txt for license information
|
|
// See https://swift.org/CONTRIBUTORS.txt for the list of Swift project authors
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "swift/AST/Decl.h"
|
|
#include "swift/AST/Types.h"
|
|
#include "llvm/Support/raw_ostream.h"
|
|
#include <algorithm>
|
|
#include <vector>
|
|
#include "RewriteContext.h"
|
|
#include "RewriteLoop.h"
|
|
#include "RewriteSystem.h"
|
|
#include "Rule.h"
|
|
#include "Trie.h"
|
|
|
|
using namespace swift;
|
|
using namespace rewriting;
|
|
|
|
RewriteSystem::RewriteSystem(RewriteContext &ctx)
|
|
: Context(ctx), Debug(ctx.getDebugOptions()) {
|
|
Initialized = 0;
|
|
Complete = 0;
|
|
Minimized = 0;
|
|
Frozen = 0;
|
|
RecordLoops = 0;
|
|
LongestInitialRule = 0;
|
|
}
|
|
|
|
RewriteSystem::~RewriteSystem() {
|
|
Trie.updateHistograms(Context.RuleTrieHistogram,
|
|
Context.RuleTrieRootHistogram);
|
|
}
|
|
|
|
/// Initialize the rewrite system using rewrite rules built by the RuleBuilder.
|
|
///
|
|
/// - recordLoops: Whether this is a rewrite system built from user-written
|
|
/// requirements, in which case we will perform minimization using rewrite
|
|
/// loops recorded during completion.
|
|
///
|
|
/// - protos: If this is a rewrite system built from a protocol connected
|
|
/// component, this contains the members of the protocol. For a rewrite
|
|
/// system built from a generic signature, this is empty. Used by
|
|
/// RewriteSystem::isInMinimizationDomain().
|
|
///
|
|
/// These parameters should be populated from the corresponding fields of the
|
|
/// RuleBuilder instance:
|
|
///
|
|
/// - writtenRequirements: The user-written requirements, if any, used to
|
|
/// track source locations for redundancy diagnostics.
|
|
///
|
|
/// - importedRules: Rewrite rules for referenced protocols. These come from
|
|
/// the Requirement Machine instances for these protocols' connected
|
|
/// components, so they are already confluent and can be imported verbatim.
|
|
///
|
|
/// - permanentRules: Permanent rules, such as associated type introduction
|
|
/// rules for associated types defined in protocols in this connected
|
|
/// component.
|
|
///
|
|
/// - requirementRules: Rules corresponding to generic requirements written
|
|
/// by the user.
|
|
///
|
|
/// This can only be called once. It adds the rules to the rewrite system,
|
|
/// allowing computeConfluentCompletion() to be called to compute the
|
|
/// complete rewrite system.
|
|
void RewriteSystem::initialize(
|
|
bool recordLoops, ArrayRef<const ProtocolDecl *> protos,
|
|
std::vector<StructuralRequirement> &&writtenRequirements,
|
|
std::vector<Rule> &&importedRules,
|
|
std::vector<std::pair<MutableTerm, MutableTerm>> &&permanentRules,
|
|
std::vector<std::tuple<MutableTerm, MutableTerm, Optional<unsigned>>>
|
|
&&requirementRules) {
|
|
assert(!Initialized);
|
|
Initialized = 1;
|
|
|
|
RecordLoops = recordLoops;
|
|
Protos = protos;
|
|
WrittenRequirements = std::move(writtenRequirements);
|
|
|
|
addRules(std::move(importedRules),
|
|
std::move(permanentRules),
|
|
std::move(requirementRules));
|
|
|
|
for (const auto &rule : getLocalRules()) {
|
|
LongestInitialRule = std::max(LongestInitialRule, rule.getDepth());
|
|
}
|
|
}
|
|
|
|
/// Reduce a term by applying all rewrite rules until fixed point.
|
|
///
|
|
/// If \p path is non-null, records the series of rewrite steps taken.
|
|
bool RewriteSystem::simplify(MutableTerm &term, RewritePath *path) const {
|
|
bool changed = false;
|
|
|
|
MutableTerm original;
|
|
RewritePath subpath;
|
|
|
|
bool debug = false;
|
|
if (Debug.contains(DebugFlags::Simplify)) {
|
|
original = term;
|
|
debug = true;
|
|
}
|
|
|
|
while (true) {
|
|
bool tryAgain = false;
|
|
|
|
auto from = term.begin();
|
|
auto end = term.end();
|
|
while (from < end) {
|
|
auto ruleID = Trie.find(from, end);
|
|
if (ruleID) {
|
|
const auto &rule = getRule(*ruleID);
|
|
|
|
auto to = from + rule.getLHS().size();
|
|
assert(std::equal(from, to, rule.getLHS().begin()));
|
|
|
|
unsigned startOffset = (unsigned)(from - term.begin());
|
|
unsigned endOffset = term.size() - rule.getLHS().size() - startOffset;
|
|
|
|
term.rewriteSubTerm(from, to, rule.getRHS());
|
|
|
|
if (path || debug) {
|
|
subpath.add(RewriteStep::forRewriteRule(startOffset, endOffset, *ruleID,
|
|
/*inverse=*/false));
|
|
}
|
|
|
|
changed = true;
|
|
tryAgain = true;
|
|
break;
|
|
}
|
|
|
|
++from;
|
|
}
|
|
|
|
if (!tryAgain)
|
|
break;
|
|
}
|
|
|
|
if (debug) {
|
|
if (changed) {
|
|
llvm::dbgs() << "= Simplified " << original << " to " << term << " via ";
|
|
subpath.dump(llvm::dbgs(), original, *this);
|
|
llvm::dbgs() << "\n";
|
|
} else {
|
|
llvm::dbgs() << "= Irreducible term: " << term << "\n";
|
|
}
|
|
}
|
|
|
|
if (path != nullptr) {
|
|
assert(changed != subpath.empty());
|
|
path->append(subpath);
|
|
}
|
|
|
|
return changed;
|
|
}
|
|
|
|
/// Adds a rewrite rule, returning true if the new rule was non-trivial.
|
|
///
|
|
/// If both sides simplify to the same term, the rule is trivial and discarded,
|
|
/// and this method returns false.
|
|
///
|
|
/// If \p path is non-null, the new rule is derived from existing rules in the
|
|
/// rewrite system; the path records a series of rewrite steps which transform
|
|
/// \p lhs to \p rhs.
|
|
bool RewriteSystem::addRule(MutableTerm lhs, MutableTerm rhs,
|
|
const RewritePath *path) {
|
|
assert(!Frozen);
|
|
|
|
assert(!lhs.empty());
|
|
assert(!rhs.empty());
|
|
|
|
if (Debug.contains(DebugFlags::Add)) {
|
|
llvm::dbgs() << "# Adding rule " << lhs << " == " << rhs << "\n\n";
|
|
}
|
|
|
|
// Now simplify both sides as much as possible with the rules we have so far.
|
|
//
|
|
// This avoids unnecessary work in the completion algorithm.
|
|
RewritePath lhsPath;
|
|
RewritePath rhsPath;
|
|
|
|
simplify(lhs, &lhsPath);
|
|
simplify(rhs, &rhsPath);
|
|
|
|
RewritePath loop;
|
|
if (path) {
|
|
// Produce a path from the simplified lhs to the simplified rhs.
|
|
|
|
// (1) First, apply lhsPath in reverse to produce the original lhs.
|
|
lhsPath.invert();
|
|
loop.append(lhsPath);
|
|
|
|
// (2) Now, apply the path from the original lhs to the original rhs
|
|
// given to us by the completion procedure.
|
|
loop.append(*path);
|
|
|
|
// (3) Finally, apply rhsPath to produce the simplified rhs, which
|
|
// is the same as the simplified lhs.
|
|
loop.append(rhsPath);
|
|
}
|
|
|
|
// If the left hand side and right hand side are already equivalent, we're
|
|
// done.
|
|
Optional<int> result = lhs.compare(rhs, Context);
|
|
if (*result == 0) {
|
|
// If this rule is a consequence of existing rules, add a homotopy
|
|
// generator.
|
|
if (path) {
|
|
// We already have a loop, since the simplified lhs is identical to the
|
|
// simplified rhs.
|
|
recordRewriteLoop(lhs, loop);
|
|
|
|
if (Debug.contains(DebugFlags::Add)) {
|
|
llvm::dbgs() << "## Recorded trivial loop at " << lhs << ": ";
|
|
loop.dump(llvm::dbgs(), lhs, *this);
|
|
llvm::dbgs() << "\n\n";
|
|
}
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
// Orient the two terms so that the left hand side is greater than the
|
|
// right hand side.
|
|
if (*result < 0) {
|
|
std::swap(lhs, rhs);
|
|
loop.invert();
|
|
}
|
|
|
|
assert(*lhs.compare(rhs, Context) > 0);
|
|
|
|
if (Debug.contains(DebugFlags::Add)) {
|
|
llvm::dbgs() << "## Simplified and oriented rule " << lhs << " => " << rhs << "\n\n";
|
|
}
|
|
|
|
unsigned newRuleID = Rules.size();
|
|
|
|
Rules.emplace_back(Term::get(lhs, Context), Term::get(rhs, Context));
|
|
|
|
if (path) {
|
|
// We have a rewrite path from the simplified lhs to the simplified rhs;
|
|
// add a rewrite step applying the new rule in reverse to close the loop.
|
|
loop.add(RewriteStep::forRewriteRule(/*startOffset=*/0, /*endOffset=*/0,
|
|
newRuleID, /*inverse=*/true));
|
|
recordRewriteLoop(lhs, loop);
|
|
|
|
if (Debug.contains(DebugFlags::Add)) {
|
|
llvm::dbgs() << "## Recorded non-trivial loop at " << lhs << ": ";
|
|
loop.dump(llvm::dbgs(), lhs, *this);
|
|
llvm::dbgs() << "\n\n";
|
|
}
|
|
}
|
|
|
|
auto oldRuleID = Trie.insert(lhs.begin(), lhs.end(), newRuleID);
|
|
if (oldRuleID) {
|
|
llvm::errs() << "Duplicate rewrite rule!\n";
|
|
const auto &oldRule = getRule(*oldRuleID);
|
|
llvm::errs() << "Old rule #" << *oldRuleID << ": ";
|
|
oldRule.dump(llvm::errs());
|
|
llvm::errs() << "\nTrying to replay what happened when I simplified this term:\n";
|
|
Debug |= DebugFlags::Simplify;
|
|
MutableTerm term = lhs;
|
|
simplify(lhs);
|
|
|
|
dump(llvm::errs());
|
|
abort();
|
|
}
|
|
|
|
// Tell the caller that we added a new rule.
|
|
return true;
|
|
}
|
|
|
|
/// Add a new rule, marking it permanent.
|
|
bool RewriteSystem::addPermanentRule(MutableTerm lhs, MutableTerm rhs) {
|
|
bool added = addRule(std::move(lhs), std::move(rhs));
|
|
if (added)
|
|
Rules.back().markPermanent();
|
|
|
|
return added;
|
|
}
|
|
|
|
/// Add a new rule, marking it explicit.
|
|
bool RewriteSystem::addExplicitRule(MutableTerm lhs, MutableTerm rhs,
|
|
Optional<unsigned> requirementID) {
|
|
bool added = addRule(std::move(lhs), std::move(rhs));
|
|
if (added) {
|
|
Rules.back().markExplicit();
|
|
Rules.back().setRequirementID(requirementID);
|
|
}
|
|
|
|
return added;
|
|
}
|
|
|
|
/// Add a set of rules from a RuleBuilder.
|
|
///
|
|
/// This is used when building a rewrite system in initialize() above.
|
|
///
|
|
/// It is also used when conditional requirement inference pulls in additional
|
|
/// protocols after the fact.
|
|
void RewriteSystem::addRules(
|
|
std::vector<Rule> &&importedRules,
|
|
std::vector<std::pair<MutableTerm, MutableTerm>> &&permanentRules,
|
|
std::vector<std::tuple<MutableTerm, MutableTerm, Optional<unsigned>>> &&requirementRules) {
|
|
unsigned ruleCount = Rules.size();
|
|
|
|
if (ruleCount == 0) {
|
|
// Fast path if this is called from initialization; just steal the
|
|
// underlying storage of the imported rule vector.
|
|
Rules = std::move(importedRules);
|
|
}
|
|
else {
|
|
// Otherwise, copy the imported rules in.
|
|
Rules.insert(Rules.end(), importedRules.begin(), importedRules.end());
|
|
}
|
|
|
|
// If this is the initial call, note the first non-imported rule so that
|
|
// we can skip over imported rules later.
|
|
if (ruleCount == 0)
|
|
FirstLocalRule = Rules.size();
|
|
|
|
// Add the imported rules to the trie.
|
|
for (unsigned newRuleID = ruleCount, e = Rules.size();
|
|
newRuleID < e; ++newRuleID) {
|
|
const auto &newRule = Rules[newRuleID];
|
|
// Skip simplified rules. At the very least we need to skip RHS-simplified
|
|
// rules since their left hand sides might duplicate existing rules; the
|
|
// others are skipped purely as an optimization. We can't skip subst-
|
|
// simplified rules, since property map construction considers them.
|
|
if (newRule.isLHSSimplified() ||
|
|
newRule.isRHSSimplified())
|
|
continue;
|
|
|
|
auto oldRuleID = Trie.insert(newRule.getLHS().begin(),
|
|
newRule.getLHS().end(),
|
|
newRuleID);
|
|
if (oldRuleID) {
|
|
llvm::errs() << "Imported rules have duplicate left hand sides!\n";
|
|
llvm::errs() << "New rule #" << newRuleID << ": " << newRule << "\n";
|
|
const auto &oldRule = getRule(*oldRuleID);
|
|
llvm::errs() << "Old rule #" << *oldRuleID << ": " << oldRule << "\n\n";
|
|
dump(llvm::errs());
|
|
abort();
|
|
}
|
|
}
|
|
|
|
// Now add our own rules.
|
|
for (const auto &rule : permanentRules)
|
|
addPermanentRule(rule.first, rule.second);
|
|
|
|
for (const auto &rule : requirementRules) {
|
|
auto lhs = std::get<0>(rule);
|
|
auto rhs = std::get<1>(rule);
|
|
auto requirementID = std::get<2>(rule);
|
|
|
|
// When this is called while adding conditional requirements, there
|
|
// shouldn't be any new structural requirement IDs.
|
|
assert(ruleCount == 0 || !requirementID.hasValue());
|
|
|
|
addExplicitRule(lhs, rhs, requirementID);
|
|
}
|
|
}
|
|
|
|
/// Delete any rules whose left hand sides can be reduced by other rules.
|
|
///
|
|
/// Must be run after the completion procedure, since the deletion of
|
|
/// rules is only valid to perform if the rewrite system is confluent.
|
|
void RewriteSystem::simplifyLeftHandSides() {
|
|
assert(Complete);
|
|
|
|
for (unsigned ruleID = FirstLocalRule, e = Rules.size(); ruleID < e; ++ruleID) {
|
|
auto &rule = getRule(ruleID);
|
|
if (rule.isLHSSimplified())
|
|
continue;
|
|
|
|
// First, see if the left hand side of this rule can be reduced using
|
|
// some other rule.
|
|
auto lhs = rule.getLHS();
|
|
auto begin = lhs.begin();
|
|
auto end = lhs.end();
|
|
while (begin < end) {
|
|
if (auto otherRuleID = Trie.find(begin++, end)) {
|
|
// A rule does not obsolete itself.
|
|
if (*otherRuleID == ruleID)
|
|
continue;
|
|
|
|
// Ignore other deleted rules.
|
|
const auto &otherRule = getRule(*otherRuleID);
|
|
if (otherRule.isLHSSimplified())
|
|
continue;
|
|
|
|
if (Debug.contains(DebugFlags::Completion)) {
|
|
const auto &otherRule = getRule(*otherRuleID);
|
|
llvm::dbgs() << "$ Deleting rule " << rule << " because "
|
|
<< "its left hand side contains " << otherRule
|
|
<< "\n";
|
|
}
|
|
|
|
rule.markLHSSimplified();
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Reduce the right hand sides of all remaining rules as much as
|
|
/// possible.
|
|
///
|
|
/// Must be run after the completion procedure, since the deletion of
|
|
/// rules is only valid to perform if the rewrite system is confluent.
|
|
void RewriteSystem::simplifyRightHandSides() {
|
|
assert(Complete);
|
|
|
|
for (unsigned ruleID = FirstLocalRule, e = Rules.size(); ruleID < e; ++ruleID) {
|
|
auto &rule = getRule(ruleID);
|
|
if (rule.isRHSSimplified())
|
|
continue;
|
|
|
|
// Now, try to reduce the right hand side.
|
|
RewritePath rhsPath;
|
|
MutableTerm rhs(rule.getRHS());
|
|
if (!simplify(rhs, &rhsPath))
|
|
continue;
|
|
|
|
auto lhs = rule.getLHS();
|
|
|
|
// We're adding a new rule, so the old rule won't apply anymore.
|
|
rule.markRHSSimplified();
|
|
|
|
unsigned newRuleID = Rules.size();
|
|
|
|
if (Debug.contains(DebugFlags::Add)) {
|
|
llvm::dbgs() << "## RHS simplification adds a rule " << lhs << " => " << rhs << "\n\n";
|
|
}
|
|
|
|
// Add a new rule with the simplified right hand side.
|
|
Rules.emplace_back(lhs, Term::get(rhs, Context));
|
|
auto oldRuleID = Trie.insert(lhs.begin(), lhs.end(), newRuleID);
|
|
assert(oldRuleID == ruleID);
|
|
(void) oldRuleID;
|
|
|
|
// Produce a loop at the original lhs.
|
|
RewritePath loop;
|
|
|
|
// (1) First, apply the original rule to produce the original rhs.
|
|
loop.add(RewriteStep::forRewriteRule(/*startOffset=*/0, /*endOffset=*/0,
|
|
ruleID, /*inverse=*/false));
|
|
|
|
// (2) Next, apply rhsPath to produce the simplified rhs.
|
|
loop.append(rhsPath);
|
|
|
|
// (3) Finally, apply the new rule in reverse to produce the original lhs.
|
|
loop.add(RewriteStep::forRewriteRule(/*startOffset=*/0, /*endOffset=*/0,
|
|
newRuleID, /*inverse=*/true));
|
|
|
|
if (Debug.contains(DebugFlags::Completion)) {
|
|
llvm::dbgs() << "$ Right hand side simplification recorded a loop at ";
|
|
llvm::dbgs() << lhs << ": ";
|
|
loop.dump(llvm::dbgs(), MutableTerm(lhs), *this);
|
|
llvm::dbgs() << "\n";
|
|
}
|
|
|
|
recordRewriteLoop(MutableTerm(lhs), loop);
|
|
}
|
|
}
|
|
|
|
/// When minimizing a generic signature, we only care about loops where the
|
|
/// basepoint is a generic parameter symbol.
|
|
///
|
|
/// When minimizing protocol requirement signatures, we only care about loops
|
|
/// where the basepoint is a protocol symbol or associated type symbol whose
|
|
/// protocol is part of the connected component.
|
|
///
|
|
/// All other loops can be discarded since they do not encode redundancies
|
|
/// that are relevant to us.
|
|
bool RewriteSystem::isInMinimizationDomain(const ProtocolDecl *proto) const {
|
|
assert(Protos.empty() || proto != nullptr);
|
|
|
|
if (proto == nullptr && Protos.empty())
|
|
return true;
|
|
|
|
if (std::find(Protos.begin(), Protos.end(), proto) != Protos.end())
|
|
return true;
|
|
|
|
return false;
|
|
}
|
|
|
|
void RewriteSystem::recordRewriteLoop(MutableTerm basepoint,
|
|
RewritePath path) {
|
|
assert(!Frozen);
|
|
|
|
RewriteLoop loop(basepoint, path);
|
|
loop.verify(*this);
|
|
|
|
if (!RecordLoops)
|
|
return;
|
|
|
|
// Ignore the rewrite loop if it is not part of our minimization domain.
|
|
if (!isInMinimizationDomain(basepoint.getRootProtocol()))
|
|
return;
|
|
|
|
Loops.push_back(loop);
|
|
}
|
|
|
|
void RewriteSystem::verifyRewriteRules(ValidityPolicy policy) const {
|
|
#define ASSERT_RULE(expr) \
|
|
if (!(expr)) { \
|
|
llvm::errs() << "&&& Malformed rewrite rule: " << rule << "\n"; \
|
|
llvm::errs() << "&&& " << #expr << "\n\n"; \
|
|
dump(llvm::errs()); \
|
|
abort(); \
|
|
}
|
|
|
|
for (const auto &rule : getLocalRules()) {
|
|
const auto &lhs = rule.getLHS();
|
|
const auto &rhs = rule.getRHS();
|
|
|
|
for (unsigned index : indices(lhs)) {
|
|
auto symbol = lhs[index];
|
|
|
|
// The left hand side can contain a single name symbol if it has the form
|
|
// T.N or T.N.[p], where T is some prefix that does not contain name
|
|
// symbols, N is a name symbol, and [p] is an optional property symbol.
|
|
//
|
|
// In the latter case, we have a protocol typealias, or a rule derived
|
|
// via resolving a critical pair involving a protocol typealias.
|
|
//
|
|
// Any other valid occurrence of a name symbol should have been reduced by
|
|
// an associated type introduction rule [P].N, marking the rule as
|
|
// LHS-simplified.
|
|
if (!rule.isLHSSimplified() &&
|
|
(rule.isPropertyRule()
|
|
? index != lhs.size() - 2
|
|
: index != lhs.size() - 1)) {
|
|
// This is only true if the input requirements were valid.
|
|
if (policy == DisallowInvalidRequirements) {
|
|
ASSERT_RULE(symbol.getKind() != Symbol::Kind::Name);
|
|
} else {
|
|
// FIXME: Assert that we diagnosed an error
|
|
}
|
|
}
|
|
|
|
if (index != lhs.size() - 1) {
|
|
ASSERT_RULE(symbol.getKind() != Symbol::Kind::Layout);
|
|
ASSERT_RULE(!symbol.hasSubstitutions());
|
|
}
|
|
|
|
if (index != 0) {
|
|
ASSERT_RULE(symbol.getKind() != Symbol::Kind::GenericParam);
|
|
}
|
|
|
|
if (index != 0 && index != lhs.size() - 1) {
|
|
ASSERT_RULE(symbol.getKind() != Symbol::Kind::Protocol);
|
|
}
|
|
}
|
|
|
|
for (unsigned index : indices(rhs)) {
|
|
auto symbol = rhs[index];
|
|
|
|
// The right hand side can contain a single name symbol if it has the form
|
|
// T.N, where T is some prefix that does not contain name symbols, and
|
|
// N is a name symbol.
|
|
//
|
|
// In this case, we have a protocol typealias, or a rule derived via
|
|
// resolving a critical pair involving a protocol typealias.
|
|
//
|
|
// Any other valid occurrence of a name symbol should have been reduced by
|
|
// an associated type introduction rule [P].N, marking the rule as
|
|
// RHS-simplified.
|
|
if (!rule.isRHSSimplified() &&
|
|
index != rhs.size() - 1) {
|
|
// This is only true if the input requirements were valid.
|
|
if (policy == DisallowInvalidRequirements) {
|
|
ASSERT_RULE(symbol.getKind() != Symbol::Kind::Name);
|
|
} else {
|
|
// FIXME: Assert that we diagnosed an error
|
|
}
|
|
}
|
|
|
|
ASSERT_RULE(symbol.getKind() != Symbol::Kind::Layout);
|
|
ASSERT_RULE(!symbol.hasSubstitutions());
|
|
|
|
if (index != 0) {
|
|
ASSERT_RULE(symbol.getKind() != Symbol::Kind::GenericParam);
|
|
}
|
|
|
|
// Completion can produce rules like [P:T].[Q:R] => [P:T].[Q]
|
|
// which are immediately simplified away.
|
|
if (!rule.isRHSSimplified() &&
|
|
index != 0) {
|
|
ASSERT_RULE(symbol.getKind() != Symbol::Kind::Protocol);
|
|
}
|
|
}
|
|
|
|
auto lhsDomain = lhs.getRootProtocol();
|
|
auto rhsDomain = rhs.getRootProtocol();
|
|
|
|
ASSERT_RULE(lhsDomain == rhsDomain);
|
|
}
|
|
|
|
#undef ASSERT_RULE
|
|
}
|
|
|
|
/// Computes the set of explicit redundant requirements to
|
|
/// emit warnings for in the source code.
|
|
void RewriteSystem::computeRedundantRequirementDiagnostics(
|
|
SmallVectorImpl<RequirementError> &errors) {
|
|
// Collect all rule IDs for each unique requirement ID.
|
|
llvm::SmallDenseMap<unsigned, llvm::SmallVector<unsigned, 2>>
|
|
rulesPerRequirement;
|
|
|
|
// Collect non-explicit requirements that are not redundant.
|
|
llvm::SmallDenseSet<unsigned, 2> nonExplicitNonRedundantRules;
|
|
|
|
for (unsigned ruleID = FirstLocalRule, e = Rules.size();
|
|
ruleID < e; ++ruleID) {
|
|
auto &rule = getRules()[ruleID];
|
|
|
|
if (rule.isPermanent())
|
|
continue;
|
|
|
|
if (!isInMinimizationDomain(rule.getLHS().getRootProtocol()))
|
|
continue;
|
|
|
|
auto requirementID = rule.getRequirementID();
|
|
|
|
if (!requirementID.hasValue()) {
|
|
if (!rule.isRedundant())
|
|
nonExplicitNonRedundantRules.insert(ruleID);
|
|
|
|
continue;
|
|
}
|
|
|
|
rulesPerRequirement[*requirementID].push_back(ruleID);
|
|
}
|
|
|
|
// Compute the set of redundant rules which transitively reference a
|
|
// non-explicit non-redundant rule. This updates nonExplicitNonRedundantRules.
|
|
//
|
|
// Since earlier redundant paths might reference rules which appear later in
|
|
// the list but not vice versa, walk the redundant paths in reverse order.
|
|
for (const auto &pair : llvm::reverse(RedundantRules)) {
|
|
// Pre-condition: the replacement path only references redundant rules
|
|
// which we have already seen. If any of those rules transitively reference
|
|
// a non-explicit, non-redundant rule, they have been inserted into the
|
|
// nonExplicitNonRedundantRules set on previous iterations.
|
|
unsigned ruleID = pair.first;
|
|
const auto &rewritePath = pair.second;
|
|
|
|
// Check if this rewrite path references a rule that is already known to
|
|
// either be non-explicit and non-redundant, or reference such a rule via
|
|
// it's redundancy path.
|
|
for (auto step : rewritePath) {
|
|
switch (step.Kind) {
|
|
case RewriteStep::Rule: {
|
|
if (nonExplicitNonRedundantRules.count(step.getRuleID())) {
|
|
nonExplicitNonRedundantRules.insert(ruleID);
|
|
continue;
|
|
}
|
|
|
|
break;
|
|
}
|
|
|
|
case RewriteStep::LeftConcreteProjection:
|
|
case RewriteStep::Decompose:
|
|
case RewriteStep::PrefixSubstitutions:
|
|
case RewriteStep::Shift:
|
|
case RewriteStep::Relation:
|
|
case RewriteStep::DecomposeConcrete:
|
|
case RewriteStep::RightConcreteProjection:
|
|
break;
|
|
}
|
|
}
|
|
|
|
// Post-condition: If the current replacement path transitively references
|
|
// any non-explicit, non-redundant rules, then nonExplicitNonRedundantRules
|
|
// contains the current rule.
|
|
}
|
|
|
|
// We diagnose a redundancy if the rule is redundant, and if its replacement
|
|
// path does not transitively involve any non-explicit, non-redundant rules.
|
|
auto isRedundantRule = [&](unsigned ruleID) {
|
|
const auto &rule = getRules()[ruleID];
|
|
|
|
return (rule.isRedundant() &&
|
|
nonExplicitNonRedundantRules.count(ruleID) == 0);
|
|
};
|
|
|
|
// Finally walk through the written requirements, diagnosing any that are
|
|
// redundant.
|
|
for (auto requirementID : indices(WrittenRequirements)) {
|
|
auto requirement = WrittenRequirements[requirementID];
|
|
|
|
// Inferred requirements can be re-stated without warning.
|
|
if (requirement.inferred)
|
|
continue;
|
|
|
|
auto pairIt = rulesPerRequirement.find(requirementID);
|
|
|
|
// If there are no rules for this structural requirement, then the
|
|
// requirement is unnecessary in the source code.
|
|
//
|
|
// This means the requirement was determined to be vacuous by
|
|
// requirement lowering and produced no rules, or the rewrite rules were
|
|
// trivially simplified by RewriteSystem::addRule().
|
|
if (pairIt == rulesPerRequirement.end()) {
|
|
errors.push_back(
|
|
RequirementError::forRedundantRequirement(requirement.req,
|
|
requirement.loc));
|
|
continue;
|
|
}
|
|
|
|
// If all rules derived from this structural requirement are redundant,
|
|
// then the requirement is unnecessary in the source code.
|
|
const auto &ruleIDs = pairIt->second;
|
|
if (llvm::all_of(ruleIDs, isRedundantRule)) {
|
|
auto requirement = WrittenRequirements[requirementID];
|
|
errors.push_back(
|
|
RequirementError::forRedundantRequirement(requirement.req,
|
|
requirement.loc));
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Free up memory by purging unused data structures after completion
|
|
/// (for a rewrite system built from a generic signature) or minimization
|
|
/// (for a rewrite system built from user-written requirements).
|
|
void RewriteSystem::freeze() {
|
|
assert(Complete);
|
|
assert(!Frozen);
|
|
|
|
for (unsigned ruleID = FirstLocalRule, e = Rules.size();
|
|
ruleID < e; ++ruleID) {
|
|
getRule(ruleID).freeze();
|
|
}
|
|
|
|
WrittenRequirements.clear();
|
|
CheckedOverlaps.clear();
|
|
RelationMap.clear();
|
|
Relations.clear();
|
|
DifferenceMap.clear();
|
|
Differences.clear();
|
|
CheckedDifferences.clear();
|
|
Loops.clear();
|
|
RedundantRules.clear();
|
|
ConflictingRules.clear();
|
|
}
|
|
|
|
void RewriteSystem::dump(llvm::raw_ostream &out) const {
|
|
out << "Rewrite system: {\n";
|
|
for (const auto &rule : Rules) {
|
|
out << "- " << rule;
|
|
if (auto ID = rule.getRequirementID()) {
|
|
auto requirement = WrittenRequirements[*ID];
|
|
out << " [ID: " << *ID << " - ";
|
|
requirement.req.dump(out);
|
|
out << " at ";
|
|
requirement.loc.print(out, Context.getASTContext().SourceMgr);
|
|
out << "]";
|
|
}
|
|
out << "\n";
|
|
}
|
|
out << "}\n";
|
|
if (!Relations.empty()) {
|
|
out << "Relations: {\n";
|
|
for (const auto &relation : Relations) {
|
|
out << "- " << relation.first << " =>> " << relation.second << "\n";
|
|
}
|
|
out << "}\n";
|
|
}
|
|
if (!Differences.empty()) {
|
|
out << "Type differences: {\n";
|
|
for (const auto &difference : Differences) {
|
|
difference.dump(out);
|
|
out << "\n";
|
|
}
|
|
out << "}\n";
|
|
}
|
|
if (!Loops.empty()) {
|
|
out << "Rewrite loops: {\n";
|
|
for (unsigned loopID : indices(Loops)) {
|
|
const auto &loop = Loops[loopID];
|
|
if (loop.isDeleted())
|
|
continue;
|
|
|
|
out << "- (#" << loopID << ") ";
|
|
loop.dump(out, *this);
|
|
out << "\n";
|
|
}
|
|
}
|
|
out << "}\n";
|
|
}
|