1//===--- RewriteRule.h - RewriteRule class ----------------------*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8///
9///  \file
10///  Defines the RewriteRule class and related functions for creating,
11///  modifying and interpreting RewriteRules.
12///
13//===----------------------------------------------------------------------===//
14
15#ifndef LLVM_CLANG_TOOLING_TRANSFORMER_REWRITE_RULE_H_
16#define LLVM_CLANG_TOOLING_TRANSFORMER_REWRITE_RULE_H_
17
18#include "clang/ASTMatchers/ASTMatchFinder.h"
19#include "clang/ASTMatchers/ASTMatchers.h"
20#include "clang/ASTMatchers/ASTMatchersInternal.h"
21#include "clang/Tooling/Refactoring/AtomicChange.h"
22#include "clang/Tooling/Transformer/MatchConsumer.h"
23#include "clang/Tooling/Transformer/RangeSelector.h"
24#include "llvm/ADT/STLExtras.h"
25#include "llvm/ADT/SmallVector.h"
26#include "llvm/Support/Error.h"
27#include <functional>
28#include <string>
29#include <utility>
30
31namespace clang {
32namespace transformer {
33using TextGenerator = std::shared_ptr<MatchComputation<std::string>>;
34
35// Description of a source-code edit, expressed in terms of an AST node.
36// Includes: an ID for the (bound) node, a selector for source related to the
37// node, a replacement and, optionally, an explanation for the edit.
38//
39// * Target: the source code impacted by the rule. This identifies an AST node,
40//   or part thereof (\c Part), whose source range indicates the extent of the
41//   replacement applied by the replacement term.  By default, the extent is the
42//   node matched by the pattern term (\c NodePart::Node). Target's are typed
43//   (\c Kind), which guides the determination of the node extent.
44//
45// * Replacement: a function that produces a replacement string for the target,
46//   based on the match result.
47//
48// * Note: (optional) a note specifically for this edit, potentially referencing
49//   elements of the match.  This will be displayed to the user, where possible;
50//   for example, in clang-tidy diagnostics.  Use of notes should be rare --
51//   explanations of the entire rewrite should be set in the rule
52//   (`RewriteRule::Explanation`) instead.  Notes serve the rare cases wherein
53//   edit-specific diagnostics are required.
54//
55// `ASTEdit` should be built using the `change` convenience functions. For
56// example,
57// \code
58//   changeTo(name(fun), cat("Frodo"))
59// \endcode
60// Or, if we use Stencil for the TextGenerator:
61// \code
62//   using stencil::cat;
63//   changeTo(statement(thenNode), cat("{", thenNode, "}"))
64//   changeTo(callArgs(call), cat(x, ",", y))
65// \endcode
66// Or, if you are changing the node corresponding to the rule's matcher, you can
67// use the single-argument override of \c change:
68// \code
69//   changeTo(cat("different_expr"))
70// \endcode
71struct ASTEdit {
72  RangeSelector TargetRange;
73  TextGenerator Replacement;
74  TextGenerator Note;
75};
76
77/// Format of the path in an include directive -- angle brackets or quotes.
78enum class IncludeFormat {
79  Quoted,
80  Angled,
81};
82
83/// Description of a source-code transformation.
84//
85// A *rewrite rule* describes a transformation of source code. A simple rule
86// contains each of the following components:
87//
88// * Matcher: the pattern term, expressed as clang matchers (with Transformer
89//   extensions).
90//
91// * Edits: a set of Edits to the source code, described with ASTEdits.
92//
93// * Explanation: explanation of the rewrite.  This will be displayed to the
94//   user, where possible; for example, in clang-tidy diagnostics.
95//
96// However, rules can also consist of (sub)rules, where the first that matches
97// is applied and the rest are ignored.  So, the above components are gathered
98// as a `Case` and a rule is a list of cases.
99//
100// Rule cases have an additional, implicit, component: the parameters. These are
101// portions of the pattern which are left unspecified, yet bound in the pattern
102// so that we can reference them in the edits.
103//
104// The \c Transformer class can be used to apply the rewrite rule and obtain the
105// corresponding replacements.
106struct RewriteRule {
107  struct Case {
108    ast_matchers::internal::DynTypedMatcher Matcher;
109    SmallVector<ASTEdit, 1> Edits;
110    TextGenerator Explanation;
111    // Include paths to add to the file affected by this case.  These are
112    // bundled with the `Case`, rather than the `RewriteRule`, because each case
113    // might have different associated changes to the includes.
114    std::vector<std::pair<std::string, IncludeFormat>> AddedIncludes;
115  };
116  // We expect RewriteRules will most commonly include only one case.
117  SmallVector<Case, 1> Cases;
118
119  // ID used as the default target of each match. The node described by the
120  // matcher is should always be bound to this id.
121  static constexpr llvm::StringLiteral RootID = "___root___";
122};
123
124/// Convenience function for constructing a simple \c RewriteRule.
125RewriteRule makeRule(ast_matchers::internal::DynTypedMatcher M,
126                     SmallVector<ASTEdit, 1> Edits,
127                     TextGenerator Explanation = nullptr);
128
129/// Convenience overload of \c makeRule for common case of only one edit.
130inline RewriteRule makeRule(ast_matchers::internal::DynTypedMatcher M,
131                            ASTEdit Edit,
132                            TextGenerator Explanation = nullptr) {
133  SmallVector<ASTEdit, 1> Edits;
134  Edits.emplace_back(std::move(Edit));
135  return makeRule(std::move(M), std::move(Edits), std::move(Explanation));
136}
137
138/// For every case in Rule, adds an include directive for the given header. The
139/// common use is assumed to be a rule with only one case. For example, to
140/// replace a function call and add headers corresponding to the new code, one
141/// could write:
142/// \code
143///   auto R = makeRule(callExpr(callee(functionDecl(hasName("foo")))),
144///            changeTo(cat("bar()")));
145///   AddInclude(R, "path/to/bar_header.h");
146///   AddInclude(R, "vector", IncludeFormat::Angled);
147/// \endcode
148void addInclude(RewriteRule &Rule, llvm::StringRef Header,
149                IncludeFormat Format = IncludeFormat::Quoted);
150
151/// Applies the first rule whose pattern matches; other rules are ignored.  If
152/// the matchers are independent then order doesn't matter. In that case,
153/// `applyFirst` is simply joining the set of rules into one.
154//
155// `applyFirst` is like an `anyOf` matcher with an edit action attached to each
156// of its cases. Anywhere you'd use `anyOf(m1.bind("id1"), m2.bind("id2"))` and
157// then dispatch on those ids in your code for control flow, `applyFirst` lifts
158// that behavior to the rule level.  So, you can write `applyFirst({makeRule(m1,
159// action1), makeRule(m2, action2), ...});`
160//
161// For example, consider a type `T` with a deterministic serialization function,
162// `serialize()`.  For performance reasons, we would like to make it
163// non-deterministic.  Therefore, we want to drop the expectation that
164// `a.serialize() = b.serialize() iff a = b` (although we'll maintain
165// `deserialize(a.serialize()) = a`).
166//
167// We have three cases to consider (for some equality function, `eq`):
168// ```
169// eq(a.serialize(), b.serialize()) --> eq(a,b)
170// eq(a, b.serialize())             --> eq(deserialize(a), b)
171// eq(a.serialize(), b)             --> eq(a, deserialize(b))
172// ```
173//
174// `applyFirst` allows us to specify each independently:
175// ```
176// auto eq_fun = functionDecl(...);
177// auto method_call = cxxMemberCallExpr(...);
178//
179// auto two_calls = callExpr(callee(eq_fun), hasArgument(0, method_call),
180//                           hasArgument(1, method_call));
181// auto left_call =
182//     callExpr(callee(eq_fun), callExpr(hasArgument(0, method_call)));
183// auto right_call =
184//     callExpr(callee(eq_fun), callExpr(hasArgument(1, method_call)));
185//
186// RewriteRule R = applyFirst({makeRule(two_calls, two_calls_action),
187//                             makeRule(left_call, left_call_action),
188//                             makeRule(right_call, right_call_action)});
189// ```
190RewriteRule applyFirst(ArrayRef<RewriteRule> Rules);
191
192/// Replaces a portion of the source text with \p Replacement.
193ASTEdit changeTo(RangeSelector Target, TextGenerator Replacement);
194/// DEPRECATED: use \c changeTo.
195inline ASTEdit change(RangeSelector Target, TextGenerator Replacement) {
196  return changeTo(std::move(Target), std::move(Replacement));
197}
198
199/// Replaces the entirety of a RewriteRule's match with \p Replacement.  For
200/// example, to replace a function call, one could write:
201/// \code
202///   makeRule(callExpr(callee(functionDecl(hasName("foo")))),
203///            changeTo(cat("bar()")))
204/// \endcode
205inline ASTEdit changeTo(TextGenerator Replacement) {
206  return changeTo(node(RewriteRule::RootID), std::move(Replacement));
207}
208/// DEPRECATED: use \c changeTo.
209inline ASTEdit change(TextGenerator Replacement) {
210  return changeTo(std::move(Replacement));
211}
212
213/// Inserts \p Replacement before \p S, leaving the source selected by \S
214/// unchanged.
215inline ASTEdit insertBefore(RangeSelector S, TextGenerator Replacement) {
216  return changeTo(before(std::move(S)), std::move(Replacement));
217}
218
219/// Inserts \p Replacement after \p S, leaving the source selected by \S
220/// unchanged.
221inline ASTEdit insertAfter(RangeSelector S, TextGenerator Replacement) {
222  return changeTo(after(std::move(S)), std::move(Replacement));
223}
224
225/// Removes the source selected by \p S.
226ASTEdit remove(RangeSelector S);
227
228/// The following three functions are a low-level part of the RewriteRule
229/// API. We expose them for use in implementing the fixtures that interpret
230/// RewriteRule, like Transformer and TransfomerTidy, or for more advanced
231/// users.
232//
233// FIXME: These functions are really public, if advanced, elements of the
234// RewriteRule API.  Recast them as such.  Or, just declare these functions
235// public and well-supported and move them out of `detail`.
236namespace detail {
237/// Builds a single matcher for the rule, covering all of the rule's cases.
238/// Only supports Rules whose cases' matchers share the same base "kind"
239/// (`Stmt`, `Decl`, etc.)  Deprecated: use `buildMatchers` instead, which
240/// supports mixing matchers of different kinds.
241ast_matchers::internal::DynTypedMatcher buildMatcher(const RewriteRule &Rule);
242
243/// Builds a set of matchers that cover the rule (one for each distinct node
244/// matcher base kind: Stmt, Decl, etc.). Node-matchers for `QualType` and
245/// `Type` are not permitted, since such nodes carry no source location
246/// information and are therefore not relevant for rewriting. If any such
247/// matchers are included, will return an empty vector.
248std::vector<ast_matchers::internal::DynTypedMatcher>
249buildMatchers(const RewriteRule &Rule);
250
251/// Gets the beginning location of the source matched by a rewrite rule. If the
252/// match occurs within a macro expansion, returns the beginning of the
253/// expansion point. `Result` must come from the matching of a rewrite rule.
254SourceLocation
255getRuleMatchLoc(const ast_matchers::MatchFinder::MatchResult &Result);
256
257/// Returns the \c Case of \c Rule that was selected in the match result.
258/// Assumes a matcher built with \c buildMatcher.
259const RewriteRule::Case &
260findSelectedCase(const ast_matchers::MatchFinder::MatchResult &Result,
261                 const RewriteRule &Rule);
262
263/// A source "transformation," represented by a character range in the source to
264/// be replaced and a corresponding replacement string.
265struct Transformation {
266  CharSourceRange Range;
267  std::string Replacement;
268};
269
270/// Attempts to translate `Edits`, which are in terms of AST nodes bound in the
271/// match `Result`, into Transformations, which are in terms of the source code
272/// text.
273///
274/// Returns an empty vector if any of the edits apply to portions of the source
275/// that are ineligible for rewriting (certain interactions with macros, for
276/// example).  Fails if any invariants are violated relating to bound nodes in
277/// the match.  However, it does not fail in the case of conflicting edits --
278/// conflict handling is left to clients.  We recommend use of the \c
279/// AtomicChange or \c Replacements classes for assistance in detecting such
280/// conflicts.
281Expected<SmallVector<Transformation, 1>>
282translateEdits(const ast_matchers::MatchFinder::MatchResult &Result,
283               llvm::ArrayRef<ASTEdit> Edits);
284} // namespace detail
285} // namespace transformer
286
287namespace tooling {
288// DEPRECATED: These are temporary aliases supporting client migration to the
289// `transformer` namespace.
290/// Wraps a string as a TextGenerator.
291using TextGenerator = transformer::TextGenerator;
292
293TextGenerator text(std::string M);
294
295using transformer::addInclude;
296using transformer::applyFirst;
297using transformer::change;
298using transformer::insertAfter;
299using transformer::insertBefore;
300using transformer::makeRule;
301using transformer::remove;
302using transformer::RewriteRule;
303using transformer::IncludeFormat;
304namespace detail {
305using namespace transformer::detail;
306} // namespace detail
307} // namespace tooling
308} // namespace clang
309
310#endif // LLVM_CLANG_TOOLING_TRANSFORMER_REWRITE_RULE_H_
311