microsoft/qdk
Publicmirrored from https://github.com/microsoft/qdkAvailable
source/compiler/qsc_stim_parser/src/parser.rs
379lines · modecode
| 1 | // Copyright (c) Microsoft Corporation. |
| 2 | // Licensed under the MIT License. |
| 3 | |
| 4 | use crate::lex::Delim::Brace; |
| 5 | use crate::lex::Delim::Paren; |
| 6 | use crate::lex::Lexer; |
| 7 | use crate::lex::Token; |
| 8 | use crate::lex::TokenKind; |
| 9 | use qsc_data_structures::span::Span; |
| 10 | use std::{iter::Peekable, str::FromStr}; |
| 11 | |
| 12 | #[derive(Debug)] |
| 13 | pub struct Circuit { |
| 14 | pub span: Span, |
| 15 | pub items: Vec<Item>, |
| 16 | } |
| 17 | |
| 18 | #[derive(Debug)] |
| 19 | pub enum Item { |
| 20 | Line(Line), |
| 21 | Block(Block), |
| 22 | } |
| 23 | |
| 24 | #[derive(Debug)] |
| 25 | pub struct Line { |
| 26 | pub span: Span, |
| 27 | pub instruction: Instruction, |
| 28 | } |
| 29 | |
| 30 | #[derive(Debug)] |
| 31 | pub struct Block { |
| 32 | pub span: Span, |
| 33 | pub block_instruction: Instruction, // currently, only the "REPEAT" instruction is supported |
| 34 | pub items: Vec<Item>, |
| 35 | } |
| 36 | |
| 37 | #[derive(Debug)] |
| 38 | pub struct Instruction { |
| 39 | pub span: Span, |
| 40 | pub name: String, |
| 41 | pub tag: Option<String>, |
| 42 | pub args: Vec<f64>, |
| 43 | pub targets: Vec<Target>, |
| 44 | } |
| 45 | |
| 46 | #[derive(Debug)] |
| 47 | pub struct Target { |
| 48 | pub span: Span, |
| 49 | pub kind: TargetKind, |
| 50 | } |
| 51 | |
| 52 | #[derive(Debug)] |
| 53 | pub enum TargetKind { |
| 54 | Qubit { |
| 55 | negated: bool, |
| 56 | value: u32, |
| 57 | }, |
| 58 | MeasurementRecord { |
| 59 | value: u32, |
| 60 | }, |
| 61 | SweepBit { |
| 62 | value: u32, |
| 63 | }, |
| 64 | Pauli { |
| 65 | negated: bool, |
| 66 | pauli: Pauli, |
| 67 | value: u32, |
| 68 | }, |
| 69 | Combiner, |
| 70 | } |
| 71 | |
| 72 | #[derive(Debug)] |
| 73 | pub enum Pauli { |
| 74 | X, |
| 75 | Y, |
| 76 | Z, |
| 77 | } |
| 78 | |
| 79 | impl FromStr for Pauli { |
| 80 | type Err = (); |
| 81 | |
| 82 | fn from_str(s: &str) -> Result<Self, Self::Err> { |
| 83 | match s { |
| 84 | "X" => Ok(Pauli::X), |
| 85 | "Y" => Ok(Pauli::Y), |
| 86 | "Z" => Ok(Pauli::Z), |
| 87 | _ => Err(()), |
| 88 | } |
| 89 | } |
| 90 | } |
| 91 | |
| 92 | pub fn parse(input: &str) -> Circuit { |
| 93 | Parser::new(input).parse() |
| 94 | } |
| 95 | |
| 96 | struct Parser<'a> { |
| 97 | input: &'a str, |
| 98 | tokens: Peekable<Lexer<'a>>, |
| 99 | } |
| 100 | |
| 101 | impl<'a> Parser<'a> { |
| 102 | pub fn new(input: &'a str) -> Self { |
| 103 | Self { |
| 104 | input, |
| 105 | tokens: Lexer::new(input).peekable(), |
| 106 | } |
| 107 | } |
| 108 | |
| 109 | pub fn expect(&mut self, kind: TokenKind) -> Token { |
| 110 | let token = self.tokens.next().expect("expected token"); |
| 111 | if token.kind != kind { |
| 112 | panic!("expected token of kind {:?}", kind); |
| 113 | } |
| 114 | token |
| 115 | } |
| 116 | |
| 117 | fn expect_number(&mut self) -> Token { |
| 118 | let token = self.tokens.next().expect("expected number"); |
| 119 | if token.kind != TokenKind::Uint && token.kind != TokenKind::Double { |
| 120 | panic!("expected number, got {:?}", token.kind); |
| 121 | } |
| 122 | token |
| 123 | } |
| 124 | |
| 125 | pub fn parse(&mut self) -> Circuit { |
| 126 | let input_len = self |
| 127 | .input |
| 128 | .len() |
| 129 | .try_into() |
| 130 | .expect("input length should fit into u32"); |
| 131 | |
| 132 | let mut items = Vec::new(); |
| 133 | while let Some(item) = self.parse_item() { |
| 134 | items.push(item); |
| 135 | } |
| 136 | |
| 137 | Circuit { |
| 138 | span: Span { |
| 139 | lo: 0, |
| 140 | hi: input_len, |
| 141 | }, |
| 142 | items, |
| 143 | } |
| 144 | } |
| 145 | |
| 146 | fn parse_item(&mut self) -> Option<Item> { |
| 147 | // Skip any leading newlines |
| 148 | while self |
| 149 | .tokens |
| 150 | .peek() |
| 151 | .is_some_and(|t| t.kind == TokenKind::Newline) |
| 152 | { |
| 153 | self.tokens.next(); |
| 154 | } |
| 155 | |
| 156 | if let TokenKind::InstructionName = self.tokens.peek()?.kind { |
| 157 | // Could be the start of a block or of a line |
| 158 | let instruction = self.parse_instruction(); |
| 159 | if let Some(token) = self.tokens.peek() |
| 160 | && token.kind == TokenKind::Open(Brace) |
| 161 | { |
| 162 | return Some(Item::Block(self.parse_block(instruction))); |
| 163 | } |
| 164 | |
| 165 | Some(Item::Line(self.parse_line(instruction))) |
| 166 | } else { |
| 167 | // TODO error! The start of every item should be an instruction; |
| 168 | None |
| 169 | } |
| 170 | } |
| 171 | |
| 172 | fn parse_block(&mut self, instruction: Instruction) -> Block { |
| 173 | let lo = instruction.span.lo; |
| 174 | let mut items = Vec::new(); |
| 175 | self.expect(TokenKind::Open(Brace)); |
| 176 | self.expect(TokenKind::Newline); |
| 177 | loop { |
| 178 | if self |
| 179 | .tokens |
| 180 | .peek() |
| 181 | .is_some_and(|t| t.kind == TokenKind::Close(Brace)) |
| 182 | { |
| 183 | break; |
| 184 | } |
| 185 | match self.parse_item() { |
| 186 | Some(item) => items.push(item), |
| 187 | None => break, |
| 188 | } |
| 189 | } |
| 190 | let closing_brace = self.expect(TokenKind::Close(Brace)); |
| 191 | self.expect(TokenKind::Newline); |
| 192 | let hi = closing_brace.span.hi; |
| 193 | Block { |
| 194 | span: Span { lo, hi }, |
| 195 | block_instruction: instruction, |
| 196 | items, |
| 197 | } |
| 198 | } |
| 199 | |
| 200 | fn parse_line(&mut self, instruction: Instruction) -> Line { |
| 201 | self.expect(TokenKind::Newline); |
| 202 | Line { |
| 203 | span: instruction.span, |
| 204 | instruction, |
| 205 | } |
| 206 | } |
| 207 | |
| 208 | fn parse_instruction(&mut self) -> Instruction { |
| 209 | let name_token = self.expect(TokenKind::InstructionName); |
| 210 | let lo = name_token.span.lo; |
| 211 | let name = self.extract_string(name_token, None); |
| 212 | |
| 213 | let tag_token = self.tokens.next_if(|t| t.kind == TokenKind::Tag); |
| 214 | let tag: Option<String> = tag_token.map(|tag_token| { |
| 215 | self.extract_string( |
| 216 | tag_token, |
| 217 | Some(Span { |
| 218 | lo: tag_token.span.lo + 1, |
| 219 | hi: tag_token.span.hi - 1, |
| 220 | }), |
| 221 | ) |
| 222 | }); |
| 223 | |
| 224 | let mut args = Vec::new(); |
| 225 | let mut targets = Vec::new(); |
| 226 | |
| 227 | if self |
| 228 | .tokens |
| 229 | .peek() |
| 230 | .is_some_and(|t| t.kind == TokenKind::Open(Paren)) |
| 231 | { |
| 232 | self.expect(TokenKind::Open(Paren)); // consume '(' |
| 233 | // Parse first arg (no leading comma) |
| 234 | if self |
| 235 | .tokens |
| 236 | .peek() |
| 237 | .is_some_and(|t| t.kind != TokenKind::Close(Paren)) |
| 238 | { |
| 239 | let arg = self.expect_number(); |
| 240 | args.push(self.extract_double(arg, None)); |
| 241 | } |
| 242 | // Each subsequent arg must be preceded by a comma |
| 243 | while self |
| 244 | .tokens |
| 245 | .peek() |
| 246 | .is_some_and(|t| t.kind != TokenKind::Close(Paren)) |
| 247 | { |
| 248 | self.expect(TokenKind::Comma); |
| 249 | let arg = self.expect_number(); |
| 250 | args.push(self.extract_double(arg, None)); |
| 251 | } |
| 252 | self.expect(TokenKind::Close(Paren)); |
| 253 | } |
| 254 | |
| 255 | while let Some(&token) = self.tokens.peek() { |
| 256 | if !self.is_target_start(&token) { |
| 257 | break; |
| 258 | } |
| 259 | targets.push(self.parse_target()); |
| 260 | } |
| 261 | |
| 262 | let hi = targets |
| 263 | .last() |
| 264 | .map(|t| t.span.hi) |
| 265 | .unwrap_or(name_token.span.hi); |
| 266 | |
| 267 | Instruction { |
| 268 | span: Span { lo, hi }, |
| 269 | name, |
| 270 | tag, |
| 271 | args, |
| 272 | targets, |
| 273 | } |
| 274 | } |
| 275 | |
| 276 | fn is_target_start(&self, token: &Token) -> bool { |
| 277 | match token.kind { |
| 278 | TokenKind::Uint |
| 279 | | TokenKind::Rec |
| 280 | | TokenKind::Sweep |
| 281 | | TokenKind::Bang |
| 282 | | TokenKind::Star => true, |
| 283 | TokenKind::InstructionName => { |
| 284 | let text = self.extract_string(*token, None); |
| 285 | text.starts_with('X') || text.starts_with('Y') || text.starts_with('Z') // STARTS WITH PAULI |
| 286 | } |
| 287 | _ => false, |
| 288 | } |
| 289 | } |
| 290 | |
| 291 | fn parse_target(&mut self) -> Target { |
| 292 | let negated_token = self.tokens.next_if(|t| t.kind == TokenKind::Bang); |
| 293 | let negated = negated_token.is_some(); |
| 294 | let first_token = self.tokens.next().expect("target empty"); |
| 295 | let lo = negated_token.map_or(first_token.span.lo, |t| t.span.lo); |
| 296 | let span = Span { |
| 297 | lo, |
| 298 | hi: first_token.span.hi, |
| 299 | }; |
| 300 | |
| 301 | match first_token.kind { |
| 302 | TokenKind::Uint => Target { |
| 303 | span, |
| 304 | kind: TargetKind::Qubit { |
| 305 | negated, |
| 306 | value: self.extract_uint(first_token, None), |
| 307 | }, |
| 308 | }, |
| 309 | TokenKind::InstructionName => Target { |
| 310 | span, |
| 311 | kind: TargetKind::Pauli { |
| 312 | negated, |
| 313 | pauli: self |
| 314 | .extract_string( |
| 315 | first_token, |
| 316 | Some(Span { |
| 317 | lo: span.lo, |
| 318 | hi: span.lo + 1, |
| 319 | }), |
| 320 | ) |
| 321 | .parse::<Pauli>() |
| 322 | .unwrap(), |
| 323 | value: self.extract_uint( |
| 324 | first_token, |
| 325 | Some(Span { |
| 326 | lo: span.lo + 1, |
| 327 | hi: span.hi, |
| 328 | }), |
| 329 | ), |
| 330 | }, |
| 331 | }, // Already validated |
| 332 | TokenKind::Rec => Target { |
| 333 | span, |
| 334 | kind: TargetKind::MeasurementRecord { |
| 335 | value: self.extract_uint( |
| 336 | first_token, |
| 337 | Some(Span { |
| 338 | lo: span.lo + 5, |
| 339 | hi: span.hi - 1, |
| 340 | }), |
| 341 | ), // Strips 'rec[-' prefix and trailing ']' TODO validate it |
| 342 | }, |
| 343 | }, |
| 344 | TokenKind::Sweep => Target { |
| 345 | span, |
| 346 | kind: TargetKind::SweepBit { |
| 347 | value: self.extract_uint( |
| 348 | first_token, |
| 349 | Some(Span { |
| 350 | lo: span.lo + 6, |
| 351 | hi: span.hi - 1, |
| 352 | }), |
| 353 | ), |
| 354 | }, // Strips 'sweep[' prefix and trailing ']' TODO validate it |
| 355 | }, |
| 356 | TokenKind::Star => Target { |
| 357 | span, |
| 358 | kind: TargetKind::Combiner, |
| 359 | }, |
| 360 | _ => panic!("Unexpected target kind"), |
| 361 | } |
| 362 | } |
| 363 | |
| 364 | fn extract_uint(&mut self, token: Token, span: Option<Span>) -> u32 { |
| 365 | self.extract_string(token, span).parse().unwrap() |
| 366 | } |
| 367 | |
| 368 | fn extract_double(&mut self, token: Token, span: Option<Span>) -> f64 { |
| 369 | self.extract_string(token, span).parse().unwrap() |
| 370 | } |
| 371 | |
| 372 | fn extract_string(&self, token: Token, span: Option<Span>) -> String { |
| 373 | if let Some(span) = span { |
| 374 | self.input[span.lo as usize..span.hi as usize].to_string() |
| 375 | } else { |
| 376 | self.input[token.span.lo as usize..token.span.hi as usize].to_string() |
| 377 | } |
| 378 | } |
| 379 | } |