| 1 |
//===--- Parsing.cpp - Parsing function implementations ---------*- 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 |
#include "clang/Tooling/Transformer/Parsing.h" |
| 10 |
#include "clang/AST/Expr.h" |
| 11 |
#include "clang/ASTMatchers/ASTMatchFinder.h" |
| 12 |
#include "clang/Basic/CharInfo.h" |
| 13 |
#include "clang/Basic/SourceLocation.h" |
| 14 |
#include "clang/Lex/Lexer.h" |
| 15 |
#include "clang/Tooling/Transformer/RangeSelector.h" |
| 16 |
#include "clang/Tooling/Transformer/SourceCode.h" |
| 17 |
#include "llvm/ADT/StringMap.h" |
| 18 |
#include "llvm/ADT/StringRef.h" |
| 19 |
#include "llvm/Support/Errc.h" |
| 20 |
#include "llvm/Support/Error.h" |
| 21 |
#include <optional> |
| 22 |
#include <string> |
| 23 |
#include <utility> |
| 24 |
#include <vector> |
| 25 |
|
| 26 |
using namespace clang; |
| 27 |
using namespace transformer; |
| 28 |
|
| 29 |
// FIXME: This implementation is entirely separate from that of the AST |
| 30 |
// matchers. Given the similarity of the languages and uses of the two parsers, |
| 31 |
// the two should share a common parsing infrastructure, as should other |
| 32 |
// Transformer types. We intend to unify this implementation soon to share as |
| 33 |
// much as possible with the AST Matchers parsing. |
| 34 |
|
| 35 |
namespace { |
| 36 |
using llvm::Expected; |
| 37 |
|
| 38 |
template <typename... Ts> using RangeSelectorOp = RangeSelector (*)(Ts...); |
| 39 |
|
| 40 |
struct ParseState { |
| 41 |
// The remaining input to be processed. |
| 42 |
StringRef Input; |
| 43 |
// The original input. Not modified during parsing; only for reference in |
| 44 |
// error reporting. |
| 45 |
StringRef OriginalInput; |
| 46 |
}; |
| 47 |
|
| 48 |
// Represents an intermediate result returned by a parsing function. Functions |
| 49 |
// that don't generate values should use `std::nullopt` |
| 50 |
template <typename ResultType> struct ParseProgress { |
| 51 |
ParseState State; |
| 52 |
// Intermediate result generated by the Parser. |
| 53 |
ResultType Value; |
| 54 |
}; |
| 55 |
|
| 56 |
template <typename T> using ExpectedProgress = llvm::Expected<ParseProgress<T>>; |
| 57 |
template <typename T> using ParseFunction = ExpectedProgress<T> (*)(ParseState); |
| 58 |
|
| 59 |
class ParseError : public llvm::ErrorInfo<ParseError> { |
| 60 |
public: |
| 61 |
// Required field for all ErrorInfo derivatives. |
| 62 |
static char ID; |
| 63 |
|
| 64 |
ParseError(size_t Pos, std::string ErrorMsg, std::string InputExcerpt) |
| 65 |
: Pos(Pos), ErrorMsg(std::move(ErrorMsg)), |
| 66 |
Excerpt(std::move(InputExcerpt)) {} |
| 67 |
|
| 68 |
void log(llvm::raw_ostream &OS) const override { |
| 69 |
OS << "parse error at position (" << Pos << "): " << ErrorMsg |
| 70 |
<< ": " + Excerpt; |
| 71 |
} |
| 72 |
|
| 73 |
std::error_code convertToErrorCode() const override { |
| 74 |
return llvm::inconvertibleErrorCode(); |
| 75 |
} |
| 76 |
|
| 77 |
// Position of the error in the input string. |
| 78 |
size_t Pos; |
| 79 |
std::string ErrorMsg; |
| 80 |
// Excerpt of the input starting at the error position. |
| 81 |
std::string Excerpt; |
| 82 |
}; |
| 83 |
|
| 84 |
char ParseError::ID; |
| 85 |
} // namespace |
| 86 |
|
| 87 |
static const llvm::StringMap<RangeSelectorOp<std::string>> & |
| 88 |
getUnaryStringSelectors() { |
| 89 |
static const llvm::StringMap<RangeSelectorOp<std::string>> M = { |
| 90 |
{"name", name}, |
| 91 |
{"node", node}, |
| 92 |
{"statement", statement}, |
| 93 |
{"statements", statements}, |
| 94 |
{"member", member}, |
| 95 |
{"callArgs", callArgs}, |
| 96 |
{"elseBranch", elseBranch}, |
| 97 |
{"initListElements", initListElements}}; |
| 98 |
return M; |
| 99 |
} |
| 100 |
|
| 101 |
static const llvm::StringMap<RangeSelectorOp<RangeSelector>> & |
| 102 |
getUnaryRangeSelectors() { |
| 103 |
static const llvm::StringMap<RangeSelectorOp<RangeSelector>> M = { |
| 104 |
{"before", before}, {"after", after}, {"expansion", expansion}}; |
| 105 |
return M; |
| 106 |
} |
| 107 |
|
| 108 |
static const llvm::StringMap<RangeSelectorOp<std::string, std::string>> & |
| 109 |
getBinaryStringSelectors() { |
| 110 |
static const llvm::StringMap<RangeSelectorOp<std::string, std::string>> M = { |
| 111 |
{"encloseNodes", encloseNodes}}; |
| 112 |
return M; |
| 113 |
} |
| 114 |
|
| 115 |
static const llvm::StringMap<RangeSelectorOp<RangeSelector, RangeSelector>> & |
| 116 |
getBinaryRangeSelectors() { |
| 117 |
static const llvm::StringMap<RangeSelectorOp<RangeSelector, RangeSelector>> |
| 118 |
M = {{"enclose", enclose}, {"between", between}}; |
| 119 |
return M; |
| 120 |
} |
| 121 |
|
| 122 |
template <typename Element> |
| 123 |
std::optional<Element> findOptional(const llvm::StringMap<Element> &Map, |
| 124 |
llvm::StringRef Key) { |
| 125 |
auto it = Map.find(Key); |
| 126 |
if (it == Map.end()) |
| 127 |
return std::nullopt; |
| 128 |
return it->second; |
| 129 |
} |
| 130 |
|
| 131 |
template <typename ResultType> |
| 132 |
ParseProgress<ResultType> makeParseProgress(ParseState State, |
| 133 |
ResultType Result) { |
| 134 |
return ParseProgress<ResultType>{State, std::move(Result)}; |
| 135 |
} |
| 136 |
|
| 137 |
static llvm::Error makeParseError(const ParseState &S, std::string ErrorMsg) { |
| 138 |
size_t Pos = S.OriginalInput.size() - S.Input.size(); |
| 139 |
return llvm::make_error<ParseError>(Pos, std::move(ErrorMsg), |
| 140 |
S.OriginalInput.substr(Pos, 20).str()); |
| 141 |
} |
| 142 |
|
| 143 |
// Returns a new ParseState that advances \c S by \c N characters. |
| 144 |
static ParseState advance(ParseState S, size_t N) { |
| 145 |
S.Input = S.Input.drop_front(N); |
| 146 |
return S; |
| 147 |
} |
| 148 |
|
| 149 |
static StringRef consumeWhitespace(StringRef S) { |
| 150 |
return S.drop_while([](char c) { return isASCII(c) && isWhitespace(c); }); |
| 151 |
} |
| 152 |
|
| 153 |
// Parses a single expected character \c c from \c State, skipping preceding |
| 154 |
// whitespace. Error if the expected character isn't found. |
| 155 |
static ExpectedProgress<std::nullopt_t> parseChar(char c, ParseState State) { |
| 156 |
State.Input = consumeWhitespace(State.Input); |
| 157 |
if (State.Input.empty() || State.Input.front() != c) |
| 158 |
return makeParseError(State, |
| 159 |
("expected char not found: " + llvm::Twine(c)).str()); |
| 160 |
return makeParseProgress(advance(State, 1), std::nullopt); |
| 161 |
} |
| 162 |
|
| 163 |
// Parses an identitifer "token" -- handles preceding whitespace. |
| 164 |
static ExpectedProgress<std::string> parseId(ParseState State) { |
| 165 |
State.Input = consumeWhitespace(State.Input); |
| 166 |
auto Id = State.Input.take_while( |
| 167 |
[](char c) { return isASCII(c) && isAsciiIdentifierContinue(c); }); |
| 168 |
if (Id.empty()) |
| 169 |
return makeParseError(State, "failed to parse name"); |
| 170 |
return makeParseProgress(advance(State, Id.size()), Id.str()); |
| 171 |
} |
| 172 |
|
| 173 |
// For consistency with the AST matcher parser and C++ code, node ids are |
| 174 |
// written as strings. However, we do not support escaping in the string. |
| 175 |
static ExpectedProgress<std::string> parseStringId(ParseState State) { |
| 176 |
State.Input = consumeWhitespace(State.Input); |
| 177 |
if (State.Input.empty()) |
| 178 |
return makeParseError(State, "unexpected end of input"); |
| 179 |
if (!State.Input.consume_front("\"")) |
| 180 |
return makeParseError( |
| 181 |
State, |
| 182 |
"expecting string, but encountered other character or end of input"); |
| 183 |
|
| 184 |
StringRef Id = State.Input.take_until([](char c) { return c == '"'; }); |
| 185 |
if (State.Input.size() == Id.size()) |
| 186 |
return makeParseError(State, "unterminated string"); |
| 187 |
// Advance past the trailing quote as well. |
| 188 |
return makeParseProgress(advance(State, Id.size() + 1), Id.str()); |
| 189 |
} |
| 190 |
|
| 191 |
// Parses a single element surrounded by parens. `Op` is applied to the parsed |
| 192 |
// result to create the result of this function call. |
| 193 |
template <typename T> |
| 194 |
ExpectedProgress<RangeSelector> parseSingle(ParseFunction<T> ParseElement, |
| 195 |
RangeSelectorOp<T> Op, |
| 196 |
ParseState State) { |
| 197 |
auto P = parseChar('(', State); |
| 198 |
if (!P) |
| 199 |
return P.takeError(); |
| 200 |
|
| 201 |
auto E = ParseElement(P->State); |
| 202 |
if (!E) |
| 203 |
return E.takeError(); |
| 204 |
|
| 205 |
P = parseChar(')', E->State); |
| 206 |
if (!P) |
| 207 |
return P.takeError(); |
| 208 |
|
| 209 |
return makeParseProgress(P->State, Op(std::move(E->Value))); |
| 210 |
} |
| 211 |
|
| 212 |
// Parses a pair of elements surrounded by parens and separated by comma. `Op` |
| 213 |
// is applied to the parsed results to create the result of this function call. |
| 214 |
template <typename T> |
| 215 |
ExpectedProgress<RangeSelector> parsePair(ParseFunction<T> ParseElement, |
| 216 |
RangeSelectorOp<T, T> Op, |
| 217 |
ParseState State) { |
| 218 |
auto P = parseChar('(', State); |
| 219 |
if (!P) |
| 220 |
return P.takeError(); |
| 221 |
|
| 222 |
auto Left = ParseElement(P->State); |
| 223 |
if (!Left) |
| 224 |
return Left.takeError(); |
| 225 |
|
| 226 |
P = parseChar(',', Left->State); |
| 227 |
if (!P) |
| 228 |
return P.takeError(); |
| 229 |
|
| 230 |
auto Right = ParseElement(P->State); |
| 231 |
if (!Right) |
| 232 |
return Right.takeError(); |
| 233 |
|
| 234 |
P = parseChar(')', Right->State); |
| 235 |
if (!P) |
| 236 |
return P.takeError(); |
| 237 |
|
| 238 |
return makeParseProgress(P->State, |
| 239 |
Op(std::move(Left->Value), std::move(Right->Value))); |
| 240 |
} |
| 241 |
|
| 242 |
// Parses input for a stencil operator(single arg ops like AsValue, MemberOp or |
| 243 |
// Id operator). Returns StencilType representing the operator on success and |
| 244 |
// error if it fails to parse input for an operator. |
| 245 |
static ExpectedProgress<RangeSelector> |
| 246 |
parseRangeSelectorImpl(ParseState State) { |
| 247 |
auto Id = parseId(State); |
| 248 |
if (!Id) |
| 249 |
return Id.takeError(); |
| 250 |
|
| 251 |
std::string OpName = std::move(Id->Value); |
| 252 |
if (auto Op = findOptional(getUnaryStringSelectors(), OpName)) |
| 253 |
return parseSingle(parseStringId, *Op, Id->State); |
| 254 |
|
| 255 |
if (auto Op = findOptional(getUnaryRangeSelectors(), OpName)) |
| 256 |
return parseSingle(parseRangeSelectorImpl, *Op, Id->State); |
| 257 |
|
| 258 |
if (auto Op = findOptional(getBinaryStringSelectors(), OpName)) |
| 259 |
return parsePair(parseStringId, *Op, Id->State); |
| 260 |
|
| 261 |
if (auto Op = findOptional(getBinaryRangeSelectors(), OpName)) |
| 262 |
return parsePair(parseRangeSelectorImpl, *Op, Id->State); |
| 263 |
|
| 264 |
return makeParseError(State, "unknown selector name: " + OpName); |
| 265 |
} |
| 266 |
|
| 267 |
Expected<RangeSelector> transformer::parseRangeSelector(llvm::StringRef Input) { |
| 268 |
ParseState State = {Input, Input}; |
| 269 |
ExpectedProgress<RangeSelector> Result = parseRangeSelectorImpl(State); |
| 270 |
if (!Result) |
| 271 |
return Result.takeError(); |
| 272 |
State = Result->State; |
| 273 |
// Discard any potentially trailing whitespace. |
| 274 |
State.Input = consumeWhitespace(State.Input); |
| 275 |
if (State.Input.empty()) |
| 276 |
return Result->Value; |
| 277 |
return makeParseError(State, "unexpected input after selector"); |
| 278 |
} |