1 | //===- Parser.h - MLIR Base Parser 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 | #ifndef MLIR_LIB_ASMPARSER_PARSER_H |
10 | #define MLIR_LIB_ASMPARSER_PARSER_H |
11 | |
12 | #include "ParserState.h" |
13 | #include "mlir/IR/Builders.h" |
14 | #include "mlir/IR/OpImplementation.h" |
15 | #include <optional> |
16 | |
17 | namespace mlir { |
18 | namespace detail { |
19 | //===----------------------------------------------------------------------===// |
20 | // Parser |
21 | //===----------------------------------------------------------------------===// |
22 | |
23 | /// This class implement support for parsing global entities like attributes and |
24 | /// types. It is intended to be subclassed by specialized subparsers that |
25 | /// include state. |
26 | class Parser { |
27 | public: |
28 | using Delimiter = OpAsmParser::Delimiter; |
29 | |
30 | Builder builder; |
31 | |
32 | Parser(ParserState &state) |
33 | : builder(state.config.getContext()), state(state) {} |
34 | |
35 | // Helper methods to get stuff from the parser-global state. |
36 | ParserState &getState() const { return state; } |
37 | MLIRContext *getContext() const { return state.config.getContext(); } |
38 | const llvm::SourceMgr &getSourceMgr() { return state.lex.getSourceMgr(); } |
39 | |
40 | /// Parse a comma-separated list of elements up until the specified end token. |
41 | ParseResult |
42 | parseCommaSeparatedListUntil(Token::Kind rightToken, |
43 | function_ref<ParseResult()> parseElement, |
44 | bool allowEmptyList = true); |
45 | |
46 | /// Parse a list of comma-separated items with an optional delimiter. If a |
47 | /// delimiter is provided, then an empty list is allowed. If not, then at |
48 | /// least one element will be parsed. |
49 | ParseResult |
50 | parseCommaSeparatedList(Delimiter delimiter, |
51 | function_ref<ParseResult()> parseElementFn, |
52 | StringRef contextMessage = StringRef()); |
53 | |
54 | /// Parse a comma separated list of elements that must have at least one entry |
55 | /// in it. |
56 | ParseResult |
57 | parseCommaSeparatedList(function_ref<ParseResult()> parseElementFn) { |
58 | return parseCommaSeparatedList(delimiter: Delimiter::None, parseElementFn); |
59 | } |
60 | |
61 | /// Parse the body of a dialect symbol, which starts and ends with <>'s, and |
62 | /// may be recursive. Return with the 'body' StringRef encompassing the entire |
63 | /// body. `isCodeCompletion` is set to true if the body contained a code |
64 | /// completion location, in which case the body is only populated up to the |
65 | /// completion. |
66 | ParseResult parseDialectSymbolBody(StringRef &body, bool &isCodeCompletion); |
67 | ParseResult parseDialectSymbolBody(StringRef &body) { |
68 | bool isCodeCompletion = false; |
69 | return parseDialectSymbolBody(body, isCodeCompletion); |
70 | } |
71 | |
72 | // We have two forms of parsing methods - those that return a non-null |
73 | // pointer on success, and those that return a ParseResult to indicate whether |
74 | // they returned a failure. The second class fills in by-reference arguments |
75 | // as the results of their action. |
76 | |
77 | //===--------------------------------------------------------------------===// |
78 | // Error Handling |
79 | //===--------------------------------------------------------------------===// |
80 | |
81 | /// Emit an error and return failure. |
82 | InFlightDiagnostic emitError(const Twine &message = {}); |
83 | InFlightDiagnostic emitError(SMLoc loc, const Twine &message = {}); |
84 | |
85 | /// Emit an error about a "wrong token". If the current token is at the |
86 | /// start of a source line, this will apply heuristics to back up and report |
87 | /// the error at the end of the previous line, which is where the expected |
88 | /// token is supposed to be. |
89 | InFlightDiagnostic emitWrongTokenError(const Twine &message = {}); |
90 | |
91 | /// Encode the specified source location information into an attribute for |
92 | /// attachment to the IR. |
93 | Location getEncodedSourceLocation(SMLoc loc) { |
94 | return state.lex.getEncodedSourceLocation(loc); |
95 | } |
96 | |
97 | //===--------------------------------------------------------------------===// |
98 | // Token Parsing |
99 | //===--------------------------------------------------------------------===// |
100 | |
101 | /// Return the current token the parser is inspecting. |
102 | const Token &getToken() const { return state.curToken; } |
103 | StringRef getTokenSpelling() const { return state.curToken.getSpelling(); } |
104 | |
105 | /// Return the last parsed token. |
106 | const Token &getLastToken() const { return state.lastToken; } |
107 | |
108 | /// If the current token has the specified kind, consume it and return true. |
109 | /// If not, return false. |
110 | bool consumeIf(Token::Kind kind) { |
111 | if (state.curToken.isNot(k: kind)) |
112 | return false; |
113 | consumeToken(kind); |
114 | return true; |
115 | } |
116 | |
117 | /// Advance the current lexer onto the next token. |
118 | void consumeToken() { |
119 | assert(state.curToken.isNot(Token::eof, Token::error) && |
120 | "shouldn't advance past EOF or errors" ); |
121 | state.lastToken = state.curToken; |
122 | state.curToken = state.lex.lexToken(); |
123 | } |
124 | |
125 | /// Advance the current lexer onto the next token, asserting what the expected |
126 | /// current token is. This is preferred to the above method because it leads |
127 | /// to more self-documenting code with better checking. |
128 | void consumeToken(Token::Kind kind) { |
129 | assert(state.curToken.is(kind) && "consumed an unexpected token" ); |
130 | consumeToken(); |
131 | } |
132 | |
133 | /// Reset the parser to the given lexer position. |
134 | void resetToken(const char *tokPos) { |
135 | state.lex.resetPointer(newPointer: tokPos); |
136 | state.lastToken = state.curToken; |
137 | state.curToken = state.lex.lexToken(); |
138 | } |
139 | |
140 | /// Consume the specified token if present and return success. On failure, |
141 | /// output a diagnostic and return failure. |
142 | ParseResult parseToken(Token::Kind expectedToken, const Twine &message); |
143 | |
144 | /// Parse an optional integer value from the stream. |
145 | OptionalParseResult parseOptionalInteger(APInt &result); |
146 | |
147 | /// Parse a floating point value from an integer literal token. |
148 | ParseResult parseFloatFromIntegerLiteral(std::optional<APFloat> &result, |
149 | const Token &tok, bool isNegative, |
150 | const llvm::fltSemantics &semantics, |
151 | size_t typeSizeInBits); |
152 | |
153 | /// Returns true if the current token corresponds to a keyword. |
154 | bool isCurrentTokenAKeyword() const { |
155 | return getToken().isAny(k1: Token::bare_identifier, k2: Token::inttype) || |
156 | getToken().isKeyword(); |
157 | } |
158 | |
159 | /// Parse a keyword, if present, into 'keyword'. |
160 | ParseResult parseOptionalKeyword(StringRef *keyword); |
161 | |
162 | //===--------------------------------------------------------------------===// |
163 | // Resource Parsing |
164 | //===--------------------------------------------------------------------===// |
165 | |
166 | /// Parse a handle to a dialect resource within the assembly format. |
167 | FailureOr<AsmDialectResourceHandle> |
168 | parseResourceHandle(const OpAsmDialectInterface *dialect, StringRef &name); |
169 | FailureOr<AsmDialectResourceHandle> parseResourceHandle(Dialect *dialect); |
170 | |
171 | //===--------------------------------------------------------------------===// |
172 | // Type Parsing |
173 | //===--------------------------------------------------------------------===// |
174 | |
175 | /// Invoke the `getChecked` method of the given Attribute or Type class, using |
176 | /// the provided location to emit errors in the case of failure. Note that |
177 | /// unlike `OpBuilder::getType`, this method does not implicitly insert a |
178 | /// context parameter. |
179 | template <typename T, typename... ParamsT> |
180 | T getChecked(SMLoc loc, ParamsT &&...params) { |
181 | return T::getChecked([&] { return emitError(loc); }, |
182 | std::forward<ParamsT>(params)...); |
183 | } |
184 | |
185 | ParseResult parseFunctionResultTypes(SmallVectorImpl<Type> &elements); |
186 | ParseResult parseTypeListNoParens(SmallVectorImpl<Type> &elements); |
187 | ParseResult parseTypeListParens(SmallVectorImpl<Type> &elements); |
188 | |
189 | /// Optionally parse a type. |
190 | OptionalParseResult parseOptionalType(Type &type); |
191 | |
192 | /// Parse an arbitrary type. |
193 | Type parseType(); |
194 | |
195 | /// Parse a complex type. |
196 | Type parseComplexType(); |
197 | |
198 | /// Parse an extended type. |
199 | Type parseExtendedType(); |
200 | |
201 | /// Parse a function type. |
202 | Type parseFunctionType(); |
203 | |
204 | /// Parse a memref type. |
205 | Type parseMemRefType(); |
206 | |
207 | /// Parse a non function type. |
208 | Type parseNonFunctionType(); |
209 | |
210 | /// Parse a tensor type. |
211 | Type parseTensorType(); |
212 | |
213 | /// Parse a tuple type. |
214 | Type parseTupleType(); |
215 | |
216 | /// Parse a vector type. |
217 | VectorType parseVectorType(); |
218 | ParseResult parseVectorDimensionList(SmallVectorImpl<int64_t> &dimensions, |
219 | SmallVectorImpl<bool> &scalableDims); |
220 | ParseResult parseDimensionListRanked(SmallVectorImpl<int64_t> &dimensions, |
221 | bool allowDynamic = true, |
222 | bool withTrailingX = true); |
223 | ParseResult parseIntegerInDimensionList(int64_t &value); |
224 | ParseResult parseXInDimensionList(); |
225 | |
226 | //===--------------------------------------------------------------------===// |
227 | // Attribute Parsing |
228 | //===--------------------------------------------------------------------===// |
229 | |
230 | /// Parse an arbitrary attribute with an optional type. |
231 | Attribute parseAttribute(Type type = {}); |
232 | |
233 | /// Parse an optional attribute with the provided type. |
234 | OptionalParseResult parseOptionalAttribute(Attribute &attribute, |
235 | Type type = {}); |
236 | OptionalParseResult parseOptionalAttribute(ArrayAttr &attribute, Type type); |
237 | OptionalParseResult parseOptionalAttribute(StringAttr &attribute, Type type); |
238 | OptionalParseResult parseOptionalAttribute(SymbolRefAttr &result, Type type); |
239 | |
240 | /// Parse an optional attribute that is demarcated by a specific token. |
241 | template <typename AttributeT> |
242 | OptionalParseResult parseOptionalAttributeWithToken(Token::Kind kind, |
243 | AttributeT &attr, |
244 | Type type = {}) { |
245 | if (getToken().isNot(k: kind)) |
246 | return std::nullopt; |
247 | |
248 | if (Attribute parsedAttr = parseAttribute(type)) { |
249 | attr = cast<AttributeT>(parsedAttr); |
250 | return success(); |
251 | } |
252 | return failure(); |
253 | } |
254 | |
255 | /// Parse an attribute dictionary. |
256 | ParseResult parseAttributeDict(NamedAttrList &attributes); |
257 | |
258 | /// Parse a distinct attribute. |
259 | Attribute parseDistinctAttr(Type type); |
260 | |
261 | /// Parse an extended attribute. |
262 | Attribute parseExtendedAttr(Type type); |
263 | |
264 | /// Parse a float attribute. |
265 | Attribute parseFloatAttr(Type type, bool isNegative); |
266 | |
267 | /// Parse a decimal or a hexadecimal literal, which can be either an integer |
268 | /// or a float attribute. |
269 | Attribute parseDecOrHexAttr(Type type, bool isNegative); |
270 | |
271 | /// Parse a dense elements attribute. |
272 | Attribute parseDenseElementsAttr(Type attrType); |
273 | ShapedType parseElementsLiteralType(Type type); |
274 | |
275 | /// Parse a dense resource elements attribute. |
276 | Attribute parseDenseResourceElementsAttr(Type attrType); |
277 | |
278 | /// Parse a DenseArrayAttr. |
279 | Attribute parseDenseArrayAttr(Type type); |
280 | |
281 | /// Parse a sparse elements attribute. |
282 | Attribute parseSparseElementsAttr(Type attrType); |
283 | |
284 | /// Parse a strided layout attribute. |
285 | Attribute parseStridedLayoutAttr(); |
286 | |
287 | //===--------------------------------------------------------------------===// |
288 | // Location Parsing |
289 | //===--------------------------------------------------------------------===// |
290 | |
291 | /// Parse a raw location instance. |
292 | ParseResult parseLocationInstance(LocationAttr &loc); |
293 | |
294 | /// Parse a callsite location instance. |
295 | ParseResult parseCallSiteLocation(LocationAttr &loc); |
296 | |
297 | /// Parse a fused location instance. |
298 | ParseResult parseFusedLocation(LocationAttr &loc); |
299 | |
300 | /// Parse a name or FileLineCol location instance. |
301 | ParseResult parseNameOrFileLineColLocation(LocationAttr &loc); |
302 | |
303 | //===--------------------------------------------------------------------===// |
304 | // Affine Parsing |
305 | //===--------------------------------------------------------------------===// |
306 | |
307 | /// Parse a reference to either an affine map, expr, or an integer set. |
308 | ParseResult parseAffineMapOrIntegerSetReference(AffineMap &map, |
309 | IntegerSet &set); |
310 | ParseResult parseAffineMapReference(AffineMap &map); |
311 | ParseResult |
312 | parseAffineExprReference(ArrayRef<std::pair<StringRef, AffineExpr>> symbolSet, |
313 | AffineExpr &expr); |
314 | ParseResult parseIntegerSetReference(IntegerSet &set); |
315 | |
316 | /// Parse an AffineMap where the dim and symbol identifiers are SSA ids. |
317 | ParseResult |
318 | parseAffineMapOfSSAIds(AffineMap &map, |
319 | function_ref<ParseResult(bool)> parseElement, |
320 | Delimiter delimiter); |
321 | |
322 | /// Parse an AffineExpr where dim and symbol identifiers are SSA ids. |
323 | ParseResult |
324 | parseAffineExprOfSSAIds(AffineExpr &expr, |
325 | function_ref<ParseResult(bool)> parseElement); |
326 | |
327 | //===--------------------------------------------------------------------===// |
328 | // Code Completion |
329 | //===--------------------------------------------------------------------===// |
330 | |
331 | /// The set of various code completion methods. Every completion method |
332 | /// returns `failure` to signal that parsing should abort after any desired |
333 | /// completions have been enqueued. Note that `failure` is does not mean |
334 | /// completion failed, it's just a signal to the parser to stop. |
335 | |
336 | ParseResult codeCompleteDialectName(); |
337 | ParseResult codeCompleteOperationName(StringRef dialectName); |
338 | ParseResult codeCompleteDialectOrElidedOpName(SMLoc loc); |
339 | ParseResult codeCompleteStringDialectOrOperationName(StringRef name); |
340 | ParseResult codeCompleteExpectedTokens(ArrayRef<StringRef> tokens); |
341 | ParseResult codeCompleteOptionalTokens(ArrayRef<StringRef> tokens); |
342 | |
343 | Attribute codeCompleteAttribute(); |
344 | Type codeCompleteType(); |
345 | Attribute |
346 | codeCompleteDialectSymbol(const llvm::StringMap<Attribute> &aliases); |
347 | Type codeCompleteDialectSymbol(const llvm::StringMap<Type> &aliases); |
348 | |
349 | protected: |
350 | /// The Parser is subclassed and reinstantiated. Do not add additional |
351 | /// non-trivial state here, add it to the ParserState class. |
352 | ParserState &state; |
353 | }; |
354 | } // namespace detail |
355 | } // namespace mlir |
356 | |
357 | #endif // MLIR_LIB_ASMPARSER_PARSER_H |
358 | |