import moo from 'moo';

// preserve precompiled lexer across runs
const lexer = moo.states({
  main: {
    comment: /#.*$/,
    blockString: { match: /:\s*"""/, push: 'blockString' }, // treat these differently from descriptions
    description: { match: '"""', push: 'description' },
    string: /"(?:\\"|[^"])*"/,
    variable: /\$[A-z\d]+/,
    id: /[A-z\d]+!?/,
    ws: { match: /[\s\t]+/, lineBreaks: true }, // whitespace to be removed
    any: /./,
  },
  description: {
    descriptionEnd: { match: '"""', pop: 1 }, // exit description
    ws: { match: /[\s\t]+/, lineBreaks: true },
    description: /./,
  },
  blockString: {
    esc: { match: '"""', pop: 1 }, // exit block string
    stringSpace: { match: /[\s\t]+/, lineBreaks: true },
    any: /./,
  },
});

export default function minify(query: string) {
  lexer.reset(query);

  // filter out comments & descriptions to make ws removal easier
  let withoutComments = '';
  for (const { type, value } of lexer) {
    if (
      type !== 'comment' &&
      type !== 'description' &&
      type !== 'descriptionEnd'
    ) {
      withoutComments += value;
    }
  }
  lexer.reset(withoutComments);

  // conditionally strip out ws where possible
  const tokens = Array.from(lexer);
  return tokens
    .map(({ type, value }, i) => {
      if (type === 'blockString') {
        // trim whitespace within multi-line arguments
        return value.replace(/\s*/g, '');
      }
      // handle whitespace surrounding IDs and variables
      if (type === 'ws') {
        const prevType = tokens[i - 1]?.type;
        const nextType = tokens[i + 1]?.type;

        return (prevType === 'id' && nextType === 'id') ||
          (prevType === 'variable' && nextType === 'id') ||
          (prevType === 'id' && nextType === 'variable')
          ? ' ' // preserve space between these token combinations
          : ''; // otherwise remove
      }
      return value;
    })
    .join('')
    .trim();
}