BPOS-TMP-REPO/packages/sfeel/src/parser.ts

/**
 * S-FEEL 子集的语法分析 —— 递归下降,纯函数(编码规范 §8.3)。
 *
 * 优先级(低 → 高):or < and < 比较/in < 加减 < 乘除 < 一元 < 基元
 */
import type { Node, UnaryTest } from './ast.ts';
import { SfeelSyntaxError, tokenize } from './lexer.ts';
import type { Token } from './lexer.ts';

/** 多词函数名 —— 子集封顶,只认这些(OS 03:无外部调用) */
const MULTIWORD = [
  ['starts', 'with'],
  ['ends', 'with'],
  ['string', 'length'],
  ['date', 'and', 'time'],
  ['days', 'and', 'time', 'duration'],
  ['upper', 'case'],
  ['lower', 'case'],
] as const;

const KEYWORDS = new Set(['and', 'or', 'not', 'true', 'false', 'null', 'in']);

class Parser {
  private index = 0;

  constructor(private readonly tokens: readonly Token[]) {}

  private peek(offset = 0): Token {
    return this.tokens[Math.min(this.index + offset, this.tokens.length - 1)]!;
  }

  private next(): Token {
    const token = this.peek();
    this.index += 1;
    return token;
  }

  private eat(text: string): boolean {
    if (this.peek().text === text && this.peek().kind !== 'string') {
      this.index += 1;
      return true;
    }
    return false;
  }

  private expect(text: string): Token {
    if (!this.eat(text)) throw new SfeelSyntaxError(`期望 ${text},实际 ${this.peek().text || 'EOF'}`, this.peek().pos);
    return this.tokens[this.index - 1]!;
  }

  atEnd(): boolean {
    return this.peek().kind === 'eof';
  }

  parseExpression(): Node {
    return this.parseOr();
  }

  private parseOr(): Node {
    let left = this.parseAnd();
    while (this.peek().kind === 'name' && this.peek().text === 'or') {
      this.next();
      left = { kind: 'binary', op: 'or', left, right: this.parseAnd() };
    }
    return left;
  }

  private parseAnd(): Node {
    let left = this.parseComparison();
    while (this.peek().kind === 'name' && this.peek().text === 'and') {
      this.next();
      left = { kind: 'binary', op: 'and', left, right: this.parseComparison() };
    }
    return left;
  }

  private parseComparison(): Node {
    const left = this.parseAdditive();
    const token = this.peek();

    if (token.kind === 'name' && token.text === 'in') {
      this.next();
      return { kind: 'in', value: left, candidates: this.parseAdditive() };
    }

    const ops = ['<=', '>=', '!=', '==', '<', '>', '='] as const;
    for (const op of ops) {
      if (token.kind === 'punct' && token.text === op) {
        this.next();
        const normalized = op === '==' ? '=' : op;
        return { kind: 'binary', op: normalized, left, right: this.parseAdditive() };
      }
    }
    return left;
  }

  private parseAdditive(): Node {
    let left = this.parseMultiplicative();
    for (;;) {
      const token = this.peek();
      if (token.kind === 'punct' && (token.text === '+' || token.text === '-')) {
        this.next();
        left = { kind: 'binary', op: token.text, left, right: this.parseMultiplicative() };
        continue;
      }
      return left;
    }
  }

  private parseMultiplicative(): Node {
    let left = this.parseUnary();
    for (;;) {
      const token = this.peek();
      if (token.kind === 'punct' && (token.text === '*' || token.text === '/')) {
        this.next();
        left = { kind: 'binary', op: token.text, left, right: this.parseUnary() };
        continue;
      }
      return left;
    }
  }

  private parseUnary(): Node {
    const token = this.peek();
    if (token.kind === 'punct' && token.text === '-') {
      this.next();
      return { kind: 'unary', op: '-', operand: this.parseUnary() };
    }
    if (token.kind === 'name' && token.text === 'not') {
      this.next();
      this.expect('(');
      const operand = this.parseExpression();
      this.expect(')');
      return { kind: 'unary', op: 'not', operand };
    }
    return this.parsePrimary();
  }

  /** 尝试合并多词函数名;不匹配则返回单词 */
  private readName(first: Token): string {
    for (const words of MULTIWORD) {
      if (words[0] !== first.text) continue;
      const matches = words.every((word, offset) => (offset === 0 ? true : this.peek(offset - 1).text === word));
      if (matches) {
        for (let k = 1; k < words.length; k += 1) this.next();
        return words.join(' ');
      }
    }
    return first.text;
  }

  private parsePrimary(): Node {
    const token = this.next();

    if (token.kind === 'number') return { kind: 'literal', value: token.value as number };
    if (token.kind === 'string') return { kind: 'literal', value: token.value as string };

    // `(` 兼两职:分组与开区间起点 —— 先解析内部,再看是不是 `..`
    if (token.kind === 'punct' && token.text === '(') {
      const inner = this.parseExpression();
      if (this.peek().kind === 'punct' && this.peek().text === '..') {
        this.next();
        return this.finishRange(true, inner);
      }
      this.expect(')');
      return inner;
    }

    // 列表或区间:[a, b] / [1..10]
    if (token.kind === 'punct' && token.text === '[') {
      return this.parseListOrRange(false);
    }

    if (token.kind === 'name') {
      if (token.text === 'true') return { kind: 'literal', value: true };
      if (token.text === 'false') return { kind: 'literal', value: false };
      if (token.text === 'null') return { kind: 'literal', value: null };

      const name = this.readName(token);

      if (this.peek().kind === 'punct' && this.peek().text === '(') {
        this.next();
        const args: Node[] = [];
        if (!(this.peek().kind === 'punct' && this.peek().text === ')')) {
          args.push(this.parseExpression());
          while (this.eat(',')) args.push(this.parseExpression());
        }
        this.expect(')');
        return { kind: 'call', name, args };
      }

      if (KEYWORDS.has(name)) throw new SfeelSyntaxError(`关键字 ${name} 不能作为路径`, token.pos);

      const segments = [name];
      while (this.peek().kind === 'punct' && this.peek().text === '.') {
        this.next();
        const part = this.next();
        if (part.kind !== 'name') throw new SfeelSyntaxError('路径分段必须是名字', part.pos);
        segments.push(part.text);
      }
      return { kind: 'path', segments };
    }

    throw new SfeelSyntaxError(`意外的记号 ${token.text || 'EOF'}`, token.pos);
  }

  private parseListOrRange(startOpen: boolean): Node {
    // 空列表
    if (this.peek().kind === 'punct' && this.peek().text === ']') {
      this.next();
      return { kind: 'list', items: [] };
    }
    const first = this.parseExpression();
    if (this.peek().kind === 'punct' && this.peek().text === '..') {
      this.next();
      return this.finishRange(startOpen, first);
    }
    const items = [first];
    while (this.eat(',')) items.push(this.parseExpression());
    this.expect(']');
    return { kind: 'list', items };
  }

  /** 区间收尾:`..` 已吃掉,读上界与闭合符 */
  private finishRange(startOpen: boolean, start: Node): Node {
    const end = this.parseExpression();
    const closing = this.next();
    if (closing.kind !== 'punct' || (closing.text !== ']' && closing.text !== ')')) {
      throw new SfeelSyntaxError('区间必须以 ] 或 ) 收尾', closing.pos);
    }
    return { kind: 'range', startOpen, endOpen: closing.text === ')', start, end };
  }

  /** 单元格谓词:`-` / `<10` / `[1..5]` / `"A","B"` */
  parseUnaryTest(): UnaryTest {
    if (this.peek().kind === 'punct' && this.peek().text === '-' && this.peek(1).kind === 'eof') {
      this.next();
      return { kind: 'any' };
    }

    const tests: UnaryTest[] = [this.parseSingleUnaryTest()];
    while (this.eat(',')) tests.push(this.parseSingleUnaryTest());
    return tests.length === 1 ? tests[0]! : { kind: 'disjunction', tests };
  }

  private parseSingleUnaryTest(): UnaryTest {
    const token = this.peek();
    const ops = ['<=', '>=', '!=', '<', '>', '='] as const;
    for (const op of ops) {
      if (token.kind === 'punct' && token.text === op) {
        this.next();
        return { kind: 'compare', op, operand: this.parseAdditive() };
      }
    }
    const node = this.parseExpression();
    if (node.kind === 'range') return { kind: 'range', node };
    return { kind: 'compare', op: '=', operand: node };
  }
}

export function parse(source: string): Node {
  const parser = new Parser(tokenize(source));
  const node = parser.parseExpression();
  if (!parser.atEnd()) throw new SfeelSyntaxError('表达式尾部有多余记号', 0);
  return node;
}

export function parseUnaryTest(source: string): UnaryTest {
  const trimmed = source.trim();
  if (trimmed === '' || trimmed === '-') return { kind: 'any' };
  const parser = new Parser(tokenize(trimmed));
  const test = parser.parseUnaryTest();
  if (!parser.atEnd()) throw new SfeelSyntaxError('单元格谓词尾部有多余记号', 0);
  return test;
}