Jump To …

rewriter.coffee

The CoffeeScript language has a good deal of optional syntax, implicit syntax, and shorthand syntax. This can greatly complicate a grammar and bloat the resulting parse table. Instead of making the parser handle it all, we take a series of passes over the token stream, using this Rewriter to convert shorthand into the unambiguous long form, add implicit indentation and parentheses, balance incorrect nestings, and generally clean things up.

The Rewriter class is used by the Lexer, directly against its internal array of tokens.

class exports.Rewriter

Helpful snippet for debugging: console.log (t[0] + '/' + t[1] for t in @tokens).join ' '

Rewrite the token stream in multiple passes, one logical filter at a time. This could certainly be changed into a single pass through the stream, with a big ol' efficient switch, but it's much nicer to work with like this. The order of these passes matters -- indentation must be corrected before implicit parentheses can be wrapped around blocks of code.

  rewrite: (@tokens) ->
    @removeLeadingNewlines()
    @removeMidExpressionNewlines()
    @closeOpenCalls()
    @closeOpenIndexes()
    @addImplicitIndentation()
    @tagPostfixConditionals()
    @addImplicitBraces()
    @addImplicitParentheses()
    @ensureBalance BALANCED_PAIRS
    @rewriteClosingParens()
    @tokens

Rewrite the token stream, looking one token ahead and behind. Allow the return value of the block to tell us how many tokens to move forwards (or backwards) in the stream, to make sure we don't miss anything as tokens are inserted and removed, and the stream changes length under our feet.

  scanTokens: (block) ->
    {tokens} = this
    i = 0
    i += block.call this, token, i, tokens while token = tokens[i]
    true

  detectEnd: (i, condition, action) ->
    {tokens} = this
    levels = 0
    while token = tokens[i]
      return action.call this, token, i     if levels is 0 and condition.call this, token, i
      return action.call this, token, i - 1 if not token or levels < 0
      if token[0] in EXPRESSION_START
        levels += 1
      else if token[0] in EXPRESSION_END
        levels -= 1
      i += 1
    i - 1

Leading newlines would introduce an ambiguity in the grammar, so we dispatch them here.

  removeLeadingNewlines: ->
    break for [tag], i in @tokens when tag isnt 'TERMINATOR'
    @tokens.splice 0, i if i

Some blocks occur in the middle of expressions -- when we're expecting this, remove their trailing newlines.

  removeMidExpressionNewlines: ->
    @scanTokens (token, i, tokens) ->
      return 1 unless token[0] is 'TERMINATOR' and @tag(i + 1) in EXPRESSION_CLOSE
      tokens.splice i, 1
      0

The lexer has tagged the opening parenthesis of a method call. Match it with its paired close. We have the mis-nested outdent case included here for calls that close on the same line, just before their outdent.

  closeOpenCalls: ->
    condition = (token, i) ->
      token[0] in [')', 'CALL_END'] or
      token[0] is 'OUTDENT' and @tag(i - 1) is ')'
    action = (token, i) ->
      @tokens[if token[0] is 'OUTDENT' then i - 1 else i][0] = 'CALL_END'
    @scanTokens (token, i) ->
      @detectEnd i + 1, condition, action if token[0] is 'CALL_START'
      1

The lexer has tagged the opening parenthesis of an indexing operation call. Match it with its paired close.

  closeOpenIndexes: ->
    condition = (token, i) -> token[0] in [']', 'INDEX_END']
    action    = (token, i) -> token[0] = 'INDEX_END'
    @scanTokens (token, i) ->
      @detectEnd i + 1, condition, action if token[0] is 'INDEX_START'
      1

Object literals may be written with implicit braces, for simple cases. Insert the missing braces here, so that the parser doesn't have to.

  addImplicitBraces: ->
    stack       = []
    start       = null
    startIndent = 0
    condition = (token, i) ->
      [one, two, three] = @tokens[i + 1 .. i + 3]
      return false if 'HERECOMMENT' is one?[0]
      [tag] = token
      (tag in ['TERMINATOR', 'OUTDENT'] and
        not (two?[0] is ':' or one?[0] is '@' and three?[0] is ':')) or
        (tag is ',' and one and
          one[0] not in ['IDENTIFIER', 'NUMBER', 'STRING', '@', 'TERMINATOR', 'OUTDENT'])
    action = (token, i) ->
      tok = ['}', '}', token[2]]
      tok.generated = yes
      @tokens.splice i, 0, tok
    @scanTokens (token, i, tokens) ->
      if (tag = token[0]) in EXPRESSION_START
        stack.push [(if tag is 'INDENT' and @tag(i - 1) is '{' then '{' else tag), i]
        return 1
      if tag in EXPRESSION_END
        start = stack.pop()
        return 1
      return 1 unless tag is ':' and
        ((ago = @tag i - 2) is ':' or stack[stack.length - 1]?[0] isnt '{')
      stack.push ['{']
      idx =  if ago is '@' then i - 2 else i - 1
      idx -= 2 while @tag(idx - 2) is 'HERECOMMENT'
      value = new String('{')
      value.generated = yes
      tok = ['{', value, token[2]]
      tok.generated = yes
      tokens.splice idx, 0, tok
      @detectEnd i + 2, condition, action
      2

Methods may be optionally called without parentheses, for simple cases. Insert the implicit parentheses here, so that the parser doesn't have to deal with them.

  addImplicitParentheses: ->
    noCall = no
    action = (token, i) ->
      idx = if token[0] is 'OUTDENT' then i + 1 else i
      @tokens.splice idx, 0, ['CALL_END', ')', token[2]]
    @scanTokens (token, i, tokens) ->
      tag     = token[0]
      noCall  = yes if tag in ['CLASS', 'IF']
      [prev, current, next] = tokens[i - 1 .. i + 1]
      callObject = not noCall and tag is 'INDENT' and
                   next and next.generated and next[0] is '{' and
                   prev and prev[0] in IMPLICIT_FUNC
      seenSingle = no
      noCall     = no if tag in LINEBREAKS
      token.call = yes if prev and not prev.spaced and tag is '?'
      return 1 if token.fromThen
      return 1 unless callObject or
        prev?.spaced and (prev.call or prev[0] in IMPLICIT_FUNC) and
        (tag in IMPLICIT_CALL or not (token.spaced or token.newLine) and tag in IMPLICIT_UNSPACED_CALL)
      tokens.splice i, 0, ['CALL_START', '(', token[2]]
      @detectEnd i + 1, (token, i) ->
        [tag] = token
        return yes if not seenSingle and token.fromThen
        seenSingle = yes if tag in ['IF', 'ELSE', '->', '=>']
        return yes if tag in ['.', '?.', '::'] and @tag(i - 1) is 'OUTDENT'
        not token.generated and @tag(i - 1) isnt ',' and tag in IMPLICIT_END and
        (tag isnt 'INDENT' or
         (@tag(i - 2) isnt 'CLASS' and @tag(i - 1) not in IMPLICIT_BLOCK and
          not ((post = @tokens[i + 1]) and post.generated and post[0] is '{')))
      , action
      prev[0] = 'FUNC_EXIST' if prev[0] is '?'
      2

Because our grammar is LALR(1), it can't handle some single-line expressions that lack ending delimiters. The Rewriter adds the implicit blocks, so it doesn't need to. ')' can close a single-line block, but we need to make sure it's balanced.

  addImplicitIndentation: ->
    @scanTokens (token, i, tokens) ->
      [tag] = token
      if tag is 'TERMINATOR' and @tag(i + 1) is 'THEN'
        tokens.splice i, 1
        return 0
      if tag is 'ELSE' and @tag(i - 1) isnt 'OUTDENT'
        tokens.splice i, 0, @indentation(token)...
        return 2
      if tag is 'CATCH' and @tag(i + 2) in ['OUTDENT', 'TERMINATOR', 'FINALLY']
        tokens.splice i + 2, 0, @indentation(token)...
        return 4
      if tag in SINGLE_LINERS and @tag(i + 1) isnt 'INDENT' and
         not (tag is 'ELSE' and @tag(i + 1) is 'IF')
        starter = tag
        [indent, outdent] = @indentation token
        indent.fromThen   = true if starter is 'THEN'
        indent.generated  = outdent.generated = true
        tokens.splice i + 1, 0, indent
        condition = (token, i) ->
          token[1] isnt ';' and token[0] in SINGLE_CLOSERS and
          not (token[0] is 'ELSE' and starter not in ['IF', 'THEN'])
        action = (token, i) ->
          @tokens.splice (if @tag(i - 1) is ',' then i - 1 else i), 0, outdent
        @detectEnd i + 2, condition, action
        tokens.splice i, 1 if tag is 'THEN'
        return 1
      return 1

Tag postfix conditionals as such, so that we can parse them with a different precedence.

  tagPostfixConditionals: ->
    condition = (token, i) -> token[0] in ['TERMINATOR', 'INDENT']
    @scanTokens (token, i) ->
      return 1 unless token[0] is 'IF'
      original = token
      @detectEnd i + 1, condition, (token, i) ->
        original[0] = 'POST_' + original[0] if token[0] isnt 'INDENT'
      1

Ensure that all listed pairs of tokens are correctly balanced throughout the course of the token stream.

  ensureBalance: (pairs) ->
    levels   = {}
    openLine = {}
    for token in @tokens
      [tag] = token
      for [open, close] in pairs
        levels[open] |= 0
        if tag is open
          openLine[open] = token[2] if levels[open]++ is 0
        else if tag is close and --levels[open] < 0
          throw Error "too many #{token[1]} on line #{token[2] + 1}"
    for open, level of levels when level > 0
      throw Error "unclosed #{ open } on line #{openLine[open] + 1}"
    this

We'd like to support syntax like this:

el.click((event) ->
  el.hide())

In order to accomplish this, move outdents that follow closing parens inwards, safely. The steps to accomplish this are:

  1. Check that all paired tokens are balanced and in order.
  2. Rewrite the stream with a stack: if you see an EXPRESSION_START, add it to the stack. If you see an EXPRESSION_END, pop the stack and replace it with the inverse of what we've just popped.
  3. Keep track of "debt" for tokens that we manufacture, to make sure we end up balanced in the end.
  4. Be careful not to alter array or parentheses delimiters with overzealous rewriting.
  rewriteClosingParens: ->
    stack = []
    debt  = {}
    debt[key] = 0 for key of INVERSES
    @scanTokens (token, i, tokens) ->
      if (tag = token[0]) in EXPRESSION_START
        stack.push token
        return 1
      return 1 unless tag in EXPRESSION_END
      if debt[inv = INVERSES[tag]] > 0
        debt[inv] -= 1
        tokens.splice i, 1
        return 0
      match = stack.pop()
      mtag  = match[0]
      oppos = INVERSES[mtag]
      return 1 if tag is oppos
      debt[mtag] += 1
      val = [oppos, if mtag is 'INDENT' then match[1] else oppos]
      if @tag(i + 2) is mtag
        tokens.splice i + 3, 0, val
        stack.push match
      else
        tokens.splice i, 0, val
      1

Generate the indentation tokens, based on another token on the same line.

  indentation: (token) ->
    [['INDENT', 2, token[2]], ['OUTDENT', 2, token[2]]]

Look up a tag by token index.

  tag: (i) -> @tokens[i]?[0]

Constants

List of the token pairs that must be balanced.

BALANCED_PAIRS = [
  ['(', ')']
  ['[', ']']
  ['{', '}']
  ['INDENT', 'OUTDENT'],
  ['CALL_START', 'CALL_END']
  ['PARAM_START', 'PARAM_END']
  ['INDEX_START', 'INDEX_END']
]

The inverse mappings of BALANCED_PAIRS we're trying to fix up, so we can look things up from either end.

INVERSES = {}

The tokens that signal the start/end of a balanced pair.

EXPRESSION_START = []
EXPRESSION_END   = []

for [left, rite] in BALANCED_PAIRS
  EXPRESSION_START.push INVERSES[rite] = left
  EXPRESSION_END  .push INVERSES[left] = rite

Tokens that indicate the close of a clause of an expression.

EXPRESSION_CLOSE = ['CATCH', 'WHEN', 'ELSE', 'FINALLY'].concat EXPRESSION_END

Tokens that, if followed by an IMPLICIT_CALL, indicate a function invocation.

IMPLICIT_FUNC    = ['IDENTIFIER', 'SUPER', ')', 'CALL_END', ']', 'INDEX_END', '@', 'THIS']

If preceded by an IMPLICIT_FUNC, indicates a function invocation.

IMPLICIT_CALL    = [
  'IDENTIFIER', 'NUMBER', 'STRING', 'JS', 'REGEX', 'NEW', 'PARAM_START', 'CLASS'
  'IF', 'TRY', 'SWITCH', 'THIS', 'BOOL', 'UNARY', 'SUPER'
  '@', '->', '=>', '[', '(', '{', '--', '++'
]

IMPLICIT_UNSPACED_CALL = ['+', '-']

Tokens indicating that the implicit call must enclose a block of expressions.

IMPLICIT_BLOCK   = ['->', '=>', '{', '[', ',']

Tokens that always mark the end of an implicit call for single-liners.

IMPLICIT_END     = ['POST_IF', 'FOR', 'WHILE', 'UNTIL', 'WHEN', 'BY', 'LOOP', 'TERMINATOR', 'INDENT']

Single-line flavors of block expressions that have unclosed endings. The grammar can't disambiguate them, so we insert the implicit indentation.

SINGLE_LINERS    = ['ELSE', '->', '=>', 'TRY', 'FINALLY', 'THEN']
SINGLE_CLOSERS   = ['TERMINATOR', 'CATCH', 'FINALLY', 'ELSE', 'OUTDENT', 'LEADING_WHEN']

Tokens that end a line.

LINEBREAKS       = ['TERMINATOR', 'INDENT', 'OUTDENT']