started patching parse.nim

private/parse.nim

@@ -5,70 +5,37 @@
 #    distribution, for details about the copyright.
 
 type
-  ScalarType = enum
+  FastParseLevelKind = enum
-    stFlow, stLiteral, stFolded
+    fplUnknown, fplSequence, fplMapKey, fplMapValue, fplSinglePairKey,
+    fplSinglePairValue, fplScalar, fplDocument
 
-  LexedDirective = enum
+  FastParseLevel = object
-    ldYaml, ldTag, ldUnknown
+    kind: FastParseLevelKind
-
+    indentation: int
-  YamlContext = enum
-    cBlock, cFlow
-
-  ChompType = enum
-    ctKeep, ctClip, ctStrip
 
   ParserContext = ref object of YamlStream
     p: YamlParser
+    lex: YamlLexer
     storedState: proc(s: YamlStream, e: var YamlStreamEvent): bool
-    scalarType: ScalarType
-    chomp: ChompType
     atSequenceItem: bool
-    recentWasMoreIndented: bool
     flowdepth: int
-    explicitFlowKey: bool
-    content, after: string
     ancestry: seq[FastParseLevel]
     level: FastParseLevel
-    tagUri: string
     tag: TagId
     anchor: AnchorId
     shorthands: Table[string, string]
     nextAnchorId: AnchorId
     newlines: int
-    indentation: int
 
   LevelEndResult = enum
     lerNothing, lerOne, lerAdditionalMapEnd
 
-const
-  space          = {' ', '\t'}
-  lineEnd        = {'\l', '\c', EndOfFile}
-  spaceOrLineEnd = {' ', '\t', '\l', '\c', EndOfFile}
-  digits         = {'0'..'9'}
-  flowIndicators = {'[', ']', '{', '}', ','}
-
-  UTF8NextLine           = toUTF8(0x85.Rune)
-  UTF8NonBreakingSpace   = toUTF8(0xA0.Rune)
-  UTF8LineSeparator      = toUTF8(0x2028.Rune)
-  UTF8ParagraphSeparator = toUTF8(0x2029.Rune)
-  UnknownIndentation = int.low
-
 proc newYamlParser*(tagLib: TagLibrary = initExtendedTagLibrary(),
                     callback: WarningCallback = nil): YamlParser =
   new(result)
   result.tagLib = tagLib
   result.callback = callback
 
-proc getLineNumber*(p: YamlParser): int = p.lexer.lineNumber
-
-proc getColNumber*(p: YamlParser): int = p.tokenstart + 1 # column is 1-based
-
-proc getLineContent*(p: YamlParser, marker: bool = true): string =
-  result = p.lexer.getCurrentLine(false)
-  if marker: result.add(repeat(' ', p.tokenstart) & "^\n")
-
-proc lexer(c: ParserContext): var BaseLexer {.inline.} = c.p.lexer
-
 template debug(message: string) {.dirty.} =
   when defined(yamlDebug):
     try: styledWriteLine(stdout, fgBlue, message)
@@ -77,31 +44,8 @@ template debug(message: string) {.dirty.} =
 proc generateError(c: ParserContext, message: string):
     ref YamlParserError {.raises: [].} =
   result = newException(YamlParserError, message)
-  result.line = c.lexer.lineNumber
+  (result.line, result.column) = c.lex.curStartPos
-  result.column = c.p.tokenstart + 1
+  result.lineContent = c.lex.getTokenLine()
-  result.lineContent = c.p.getLineContent(true)
-
-proc generateError(lx: BaseLexer, message: string):
-    ref YamlParserError {.raises: [].} =
-  result = newException(YamlParserError, message)
-  result.line = lx.lineNumber
-  result.column = lx.bufpos + 1
-  result.lineContent = lx.getCurrentLine(false) &
-      repeat(' ', lx.getColNumber(lx.bufpos)) & "^\n"
-
-template lexCR(lexer: BaseLexer) {.dirty.} =
-  try: lexer.bufpos = lexer.handleCR(lexer.bufpos)
-  except:
-    var e = generateError(lexer, "I/O Error: " & getCurrentExceptionMsg())
-    e.parent = getCurrentException()
-    raise e
-
-template lexLF(lexer: BaseLexer) {.dirty.} =
-  try: lexer.bufpos = lexer.handleLF(lexer.bufpos)
-  except:
-    var e = generateError(lexer, "I/O Error: " & getCurrentExceptionMsg())
-    e.parent = getCurrentException()
-    raise e
 
 proc callCallback(c: ParserContext, msg: string) {.raises: [YamlParserError].} =
   try:
@@ -114,12 +58,6 @@ proc callCallback(c: ParserContext, msg: string) {.raises: [YamlParserError].} =
     e.parent = getCurrentException()
     raise e
 
-proc addMultiple(s: var string, c: char, num: int) {.raises: [], inline.} =
-  for i in 1..num:
-    s.add(c)
-
-proc reset(buffer: var string) {.raises: [], inline.} = buffer.setLen(0)
-
 proc initLevel(k: FastParseLevelKind): FastParseLevel {.raises: [], inline.} =
   FastParseLevel(kind: k, indentation: UnknownIndentation)
 
@@ -130,18 +68,12 @@ proc emptyScalar(c: ParserContext): YamlStreamEvent {.raises: [], inline.} =
 
 proc currentScalar(c: ParserContext): YamlStreamEvent {.raises: [], inline.} =
   result = YamlStreamEvent(kind: yamlScalar, scalarTag: c.tag,
-                           scalarAnchor: c.anchor, scalarContent: c.content)
+                           scalarAnchor: c.anchor)
+  shallowCopy(result.scalarContent, c.lex.buf)
+  c.lex.buf = newStringOfCap(256)
   c.tag = yTagQuestionMark
   c.anchor = yAnchorNone
 
-proc handleLineEnd(c: ParserContext, incNewlines: static[bool]): bool =
-  case c.lexer.buf[c.lexer.bufpos]
-  of '\l': c.lexer.lexLF()
-  of '\c': c.lexer.lexCR()
-  of EndOfFile: return true
-  else: discard
-  when incNewlines: c.newlines.inc()
-
 proc objectStart(c: ParserContext, k: static[YamlStreamEventKind],
                  single: bool = false): YamlStreamEvent {.raises: [].} =
   yAssert(c.level.kind == fplUnknown)
@@ -181,607 +113,58 @@ proc initDocValues(c: ParserContext) {.raises: [].} =
   c.anchor = yAnchorNone
   c.ancestry.add(FastParseLevel(kind: fplDocument, indentation: -1))
 
-proc startToken(c: ParserContext) {.raises: [], inline.} =
-  c.p.tokenstart = c.lexer.getColNumber(c.lexer.bufpos)
-
-proc anchorName(c: ParserContext) {.raises: [].} =
-  debug("lex: anchorName")
-  while true:
-    c.lexer.bufpos.inc()
-    let ch = c.lexer.buf[c.lexer.bufpos]
-    case ch
-    of spaceOrLineEnd, '[', ']', '{', '}', ',': break
-    else: c.content.add(ch)
-
 proc handleAnchor(c: ParserContext) {.raises: [YamlParserError].} =
-  c.startToken()
   if c.level.kind != fplUnknown: raise c.generateError("Unexpected token")
   if c.anchor != yAnchorNone:
     raise c.generateError("Only one anchor is allowed per node")
-  c.content.reset()
-  c.anchorName()
   c.anchor = c.nextAnchorId
-  c.p.anchors[c.content] = c.anchor
+  c.p.anchors[c.lex.buf] = c.anchor
   c.nextAnchorId = AnchorId(int(c.nextAnchorId) + 1)
-
+  c.lex.buf.setLen(0)
-proc finishLine(lexer: var BaseLexer) {.raises: [], inline.} =
-  debug("lex: finishLine")
-  while lexer.buf[lexer.bufpos] notin lineEnd:
-    lexer.bufpos.inc()
-
-proc skipWhitespace(lexer: var BaseLexer) {.raises: [], inline.} =
-  debug("lex: skipWhitespace")
-  while lexer.buf[lexer.bufpos] in space: lexer.bufpos.inc()
-
-# TODO: {.raises: [].}
-proc skipWhitespaceCommentsAndNewlines(lexer: var BaseLexer) {.inline.} =
-  debug("lex: skipWhitespaceCommentsAndNewlines")
-  if lexer.buf[lexer.bufpos] != '#':
-    while true:
-      case lexer.buf[lexer.bufpos]
-      of space: lexer.bufpos.inc()
-      of '\l': lexer.lexLF()
-      of '\c': lexer.lexCR()
-      of '#': # also skip comments
-        lexer.bufpos.inc()
-        while lexer.buf[lexer.bufpos] notin lineEnd:
-          lexer.bufpos.inc()
-      else: break
-
-proc skipIndentation(lexer: var BaseLexer) {.raises: [], inline.} =
-  debug("lex: skipIndentation")
-  while lexer.buf[lexer.bufpos] == ' ': lexer.bufpos.inc()
-
-proc directiveName(lexer: var BaseLexer, directive: var LexedDirective)
-    {.raises: [].} =
-  debug("lex: directiveName")
-  directive = ldUnknown
-  lexer.bufpos.inc()
-  if lexer.buf[lexer.bufpos] == 'Y':
-    lexer.bufpos.inc()
-    if lexer.buf[lexer.bufpos] == 'A':
-      lexer.bufpos.inc()
-      if lexer.buf[lexer.bufpos] == 'M':
-        lexer.bufpos.inc()
-        if lexer.buf[lexer.bufpos] == 'L':
-          lexer.bufpos.inc()
-          if lexer.buf[lexer.bufpos] in spaceOrLineEnd:
-            directive = ldYaml
-  elif lexer.buf[lexer.bufpos] == 'T':
-    lexer.bufpos.inc()
-    if lexer.buf[lexer.bufpos] == 'A':
-      lexer.bufpos.inc()
-      if lexer.buf[lexer.bufpos] == 'G':
-        lexer.bufpos.inc()
-        if lexer.buf[lexer.bufpos] in spaceOrLineEnd:
-          directive = ldTag
-  while lexer.buf[lexer.bufpos] notin spaceOrLineEnd:
-    lexer.bufpos.inc()
-
-proc yamlVersion(lexer: var BaseLexer, o: var string)
-    {.raises: [YamlParserError], inline.} =
-  debug("lex: yamlVersion")
-  while lexer.buf[lexer.bufpos] in space: lexer.bufpos.inc()
-  var c = lexer.buf[lexer.bufpos]
-  if c notin digits: raise lexer.generateError("Invalid YAML version number")
-  o.add(c)
-  lexer.bufpos.inc()
-  c = lexer.buf[lexer.bufpos]
-  while c in digits:
-    lexer.bufpos.inc()
-    o.add(c)
-    c = lexer.buf[lexer.bufpos]
-  if lexer.buf[lexer.bufpos] != '.':
-    raise lexer.generateError("Invalid YAML version number")
-  o.add('.')
-  lexer.bufpos.inc()
-  c = lexer.buf[lexer.bufpos]
-  if c notin digits: raise lexer.generateError("Invalid YAML version number")
-  o.add(c)
-  lexer.bufpos.inc()
-  c = lexer.buf[lexer.bufpos]
-  while c in digits:
-    o.add(c)
-    lexer.bufpos.inc()
-    c = lexer.buf[lexer.bufpos]
-  if lexer.buf[lexer.bufpos] notin spaceOrLineEnd:
-    raise lexer.generateError("Invalid YAML version number")
-
-proc lineEnding(c: ParserContext) {.raises: [YamlParserError], inline.} =
-  debug("lex: lineEnding")
-  if c.lexer.buf[c.lexer.bufpos] notin lineEnd:
-    while c.lexer.buf[c.lexer.bufpos] in space: c.lexer.bufpos.inc()
-    if c.lexer.buf[c.lexer.bufpos] in lineEnd: discard
-    elif c.lexer.buf[c.lexer.bufpos] == '#':
-      while c.lexer.buf[c.lexer.bufpos] notin lineEnd: c.lexer.bufpos.inc()
-    else:
-      c.startToken()
-      raise c.generateError("Unexpected token (expected comment or line end)")
-
-proc tagShorthand(lexer: var BaseLexer, shorthand: var string) {.inline.} =
-  debug("lex: tagShorthand")
-  while lexer.buf[lexer.bufpos] in space: lexer.bufpos.inc()
-  yAssert lexer.buf[lexer.bufpos] == '!'
-  shorthand.add('!')
-  lexer.bufpos.inc()
-  var ch = lexer.buf[lexer.bufpos]
-  if ch in spaceOrLineEnd: discard
-  else:
-    while ch != '!':
-      case ch
-      of 'a' .. 'z', 'A' .. 'Z', '0' .. '9', '-':
-        shorthand.add(ch)
-        lexer.bufpos.inc()
-        ch = lexer.buf[lexer.bufpos]
-      else: raise lexer.generateError("Illegal character in tag shorthand")
-    shorthand.add(ch)
-    lexer.bufpos.inc()
-  if lexer.buf[lexer.bufpos] notin spaceOrLineEnd:
-    raise lexer.generateError("Missing space after tag shorthand")
-
-proc tagUriMapping(lexer: var BaseLexer, uri: var string)
-    {.raises: [YamlParserError].} =
-  debug("lex: tagUriMapping")
-  while lexer.buf[lexer.bufpos] in space:
-    lexer.bufpos.inc()
-  var ch = lexer.buf[lexer.bufpos]
-  if ch == '!':
-    uri.add(ch)
-    lexer.bufpos.inc()
-    ch = lexer.buf[lexer.bufpos]
-  while ch notin spaceOrLineEnd:
-    case ch
-    of 'a' .. 'z', 'A' .. 'Z', '0' .. '9', '#', ';', '/', '?', ':', '@', '&',
-       '-', '=', '+', '$', ',', '_', '.', '~', '*', '\'', '(', ')':
-      uri.add(ch)
-      lexer.bufpos.inc()
-      ch = lexer.buf[lexer.bufpos]
-    else: raise lexer.generateError("Invalid tag uri")
-
-proc directivesEndMarker(lexer: var BaseLexer, success: var bool)
-    {.raises: [].} =
-  debug("lex: directivesEndMarker")
-  success = true
-  for i in 0..2:
-    if lexer.buf[lexer.bufpos + i] != '-':
-      success = false
-      break
-  if success: success = lexer.buf[lexer.bufpos + 3] in spaceOrLineEnd
-
-proc documentEndMarker(lexer: var BaseLexer, success: var bool) {.raises: [].} =
-  debug("lex: documentEndMarker")
-  success = true
-  for i in 0..2:
-    if lexer.buf[lexer.bufpos + i] != '.':
-      success = false
-      break
-  if success: success = lexer.buf[lexer.bufpos + 3] in spaceOrLineEnd
-
-proc unicodeSequence(lexer: var BaseLexer, length: int):
-      string {.raises: [YamlParserError].} =
-  debug("lex: unicodeSequence")
-  var unicodeChar = 0.int
-  for i in countup(0, length - 1):
-    lexer.bufpos.inc()
-    let
-      digitPosition = length - i - 1
-      ch = lexer.buf[lexer.bufpos]
-    case ch
-    of EndOFFile, '\l', '\c':
-      raise lexer.generateError("Unfinished unicode escape sequence")
-    of '0' .. '9':
-      unicodeChar = unicodechar or (int(ch) - 0x30) shl (digitPosition * 4)
-    of 'A' .. 'F':
-      unicodeChar = unicodechar or (int(ch) - 0x37) shl (digitPosition * 4)
-    of 'a' .. 'f':
-      unicodeChar = unicodechar or (int(ch) - 0x57) shl (digitPosition * 4)
-    else:
-      raise lexer.generateError(
-          "Invalid character in unicode escape sequence")
-  return toUTF8(Rune(unicodeChar))
-
-proc byteSequence(lexer: var BaseLexer): char {.raises: [YamlParserError].} =
-  debug("lex: byteSequence")
-  var charCode = 0.int8
-  for i in 0 .. 1:
-    lexer.bufpos.inc()
-    let
-      digitPosition = int8(1 - i)
-      ch = lexer.buf[lexer.bufpos]
-    case ch
-    of EndOfFile, '\l', 'r':
-      raise lexer.generateError("Unfinished octet escape sequence")
-    of '0' .. '9':
-      charCode = charCode or (int8(ch) - 0x30.int8) shl (digitPosition * 4)
-    of 'A' .. 'F':
-      charCode = charCode or (int8(ch) - 0x37.int8) shl (digitPosition * 4)
-    of 'a' .. 'f':
-      charCode = charCode or (int8(ch) - 0x57.int8) shl (digitPosition * 4)
-    else:
-      raise lexer.generateError("Invalid character in octet escape sequence")
-  return char(charCode)
-
-# TODO: {.raises: [].}
-proc processQuotedWhitespace(c: ParserContext, newlines: var int) =
-  c.after.reset()
-  block outer:
-    while true:
-      case c.lexer.buf[c.lexer.bufpos]
-      of ' ', '\t': c.after.add(c.lexer.buf[c.lexer.bufpos])
-      of '\l':
-        c.lexer.bufpos = c.lexer.handleLF(c.lexer.bufpos)
-        break
-      of '\c':
-        c.lexer.bufpos = c.lexer.handleLF(c.lexer.bufpos)
-        break
-      else:
-        c.content.add(c.after)
-        break outer
-      c.lexer.bufpos.inc()
-    while true:
-      case c.lexer.buf[c.lexer.bufpos]
-      of ' ', '\t': discard
-      of '\l':
-        c.lexer.lexLF()
-        newlines.inc()
-        continue
-      of '\c':
-        c.lexer.lexCR()
-        newlines.inc()
-        continue
-      else:
-        if newlines == 0: discard
-        elif newlines == 1: c.content.add(' ')
-        else: c.content.addMultiple('\l', newlines - 1)
-        break
-      c.lexer.bufpos.inc()
-
-# TODO: {.raises: [YamlParserError].}
-proc doubleQuotedScalar(c: ParserContext) =
-  debug("lex: doubleQuotedScalar")
-  c.lexer.bufpos.inc()
-  while true:
-    var ch = c.lexer.buf[c.lexer.bufpos]
-    case ch
-    of EndOfFile:
-      raise c.lexer.generateError("Unfinished double quoted string")
-    of '\\':
-      c.lexer.bufpos.inc()
-      case c.lexer.buf[c.lexer.bufpos]
-      of EndOfFile:
-        raise c.lexer.generateError("Unfinished escape sequence")
-      of '0':       c.content.add('\0')
-      of 'a':       c.content.add('\x07')
-      of 'b':       c.content.add('\x08')
-      of '\t', 't': c.content.add('\t')
-      of 'n':       c.content.add('\l')
-      of 'v':       c.content.add('\v')
-      of 'f':       c.content.add('\f')
-      of 'r':       c.content.add('\c')
-      of 'e':       c.content.add('\e')
-      of ' ':       c.content.add(' ')
-      of '"':       c.content.add('"')
-      of '/':       c.content.add('/')
-      of '\\':      c.content.add('\\')
-      of 'N':       c.content.add(UTF8NextLine)
-      of '_':       c.content.add(UTF8NonBreakingSpace)
-      of 'L':       c.content.add(UTF8LineSeparator)
-      of 'P':       c.content.add(UTF8ParagraphSeparator)
-      of 'x':       c.content.add(c.lexer.unicodeSequence(2))
-      of 'u':       c.content.add(c.lexer.unicodeSequence(4))
-      of 'U':       c.content.add(c.lexer.unicodeSequence(8))
-      of '\l', '\c':
-        var newlines = 0
-        c.processQuotedWhitespace(newlines)
-        continue
-      else: raise c.lexer.generateError("Illegal character in escape sequence")
-    of '"':
-      c.lexer.bufpos.inc()
-      break
-    of '\l', '\c', '\t', ' ':
-      var newlines = 1
-      c.processQuotedWhitespace(newlines)
-      continue
-    else: c.content.add(ch)
-    c.lexer.bufpos.inc()
-
-# TODO: {.raises: [].}
-proc singleQuotedScalar(c: ParserContext) =
-  debug("lex: singleQuotedScalar")
-  c.lexer.bufpos.inc()
-  while true:
-    case c.lexer.buf[c.lexer.bufpos]
-    of '\'':
-      c.lexer.bufpos.inc()
-      if c.lexer.buf[c.lexer.bufpos] == '\'': c.content.add('\'')
-      else: break
-    of EndOfFile: raise c.lexer.generateError("Unfinished single quoted string")
-    of '\l', '\c', '\t', ' ':
-      var newlines = 1
-      c.processQuotedWhitespace(newlines)
-      continue
-    else: c.content.add(c.lexer.buf[c.lexer.bufpos])
-    c.lexer.bufpos.inc()
-
-proc isPlainSafe(lexer: BaseLexer, index: int, context: YamlContext): bool
-    {.raises: [].} =
-  case lexer.buf[lexer.bufpos + 1]
-  of spaceOrLineEnd: result = false
-  of flowIndicators: result = context == cBlock
-  else: result = true
-
-# tried this for performance optimization, but it didn't optimize any
-# performance. keeping it around for future reference.
-#const
-#  plainCharOut   = {'!', '\"', '$'..'9',  ';'..'\xFF'}
-#  plainCharIn    = {'!', '\"', '$'..'+', '-'..'9', ';'..'Z', '\\', '^'..'z',
-#                    '|', '~'..'\xFF'}
-#template isPlainChar(c: char, context: YamlContext): bool =
-#  when context == cBlock: c in plainCharOut
-#  else: c in plainCharIn
-
-proc plainScalar(c: ParserContext, context: static[YamlContext])
-    {.raises: [].} =
-  debug("lex: plainScalar")
-  c.content.add(c.lexer.buf[c.lexer.bufpos])
-  block outer:
-    while true:
-      c.lexer.bufpos.inc()
-      let ch = c.lexer.buf[c.lexer.bufpos]
-      case ch
-      of ' ', '\t':
-        c.after.setLen(1)
-        c.after[0] = ch
-        while true:
-          c.lexer.bufpos.inc()
-          let ch2 = c.lexer.buf[c.lexer.bufpos]
-          case ch2
-          of ' ', '\t': c.after.add(ch2)
-          of lineEnd: break outer
-          of ':':
-            if c.lexer.isPlainSafe(c.lexer.bufpos + 1, context):
-              c.content.add(c.after & ':')
-              break
-            else: break outer
-          of '#': break outer
-          of flowIndicators:
-            if context == cBlock:
-              c.content.add(c.after)
-              c.content.add(ch2)
-              break
-            else: break outer
-          else:
-            c.content.add(c.after)
-            c.content.add(ch2)
-            break
-      of flowIndicators:
-        when context == cFlow: break
-        else: c.content.add(ch)
-      of lineEnd: break
-      of ':':
-        if c.lexer.isPlainSafe(c.lexer.bufpos + 1, context): c.content.add(':')
-        else: break outer
-      else: c.content.add(ch)
-  debug("lex: \"" & c.content & '\"')
 
 proc continueMultilineScalar(c: ParserContext) {.raises: [].} =
-  c.content.add(if c.newlines == 1: " " else: repeat('\l', c.newlines - 1))
+  c.lex.buf.add(if c.newlines == 1: " " else: repeat('\l', c.newlines - 1))
-  c.startToken()
+  c.newlines = 0
-  c.plainScalar(cBlock)
 
 template startScalar(t: ScalarType) {.dirty.} =
   c.newlines = 0
   c.level.kind = fplScalar
   c.scalarType = t
 
-proc blockScalarHeader(c: ParserContext): bool =
-  debug("lex: blockScalarHeader")
-  c.chomp = ctClip
-  c.level.indentation = UnknownIndentation
-  if c.tag == yTagQuestionMark: c.tag = yTagExclamationMark
-  let t = if c.lexer.buf[c.lexer.bufpos] == '|': stLiteral else: stFolded
-  while true:
-    c.lexer.bufpos.inc()
-    case c.lexer.buf[c.lexer.bufpos]
-    of '+':
-      if c.chomp != ctClip:
-        raise c.lexer.generateError("Only one chomping indicator is allowed")
-      c.chomp = ctKeep
-    of '-':
-      if c.chomp != ctClip:
-        raise c.lexer.generateError("Only one chomping indicator is allowed")
-      c.chomp = ctStrip
-    of '1'..'9':
-      if c.level.indentation != UnknownIndentation:
-        raise c.lexer.generateError("Only one p.indentation indicator is allowed")
-      c.level.indentation = c.ancestry[c.ancestry.high].indentation +
-          ord(c.lexer.buf[c.lexer.bufpos]) - ord('\x30')
-    of spaceOrLineEnd: break
-    else:
-      raise c.lexer.generateError(
-          "Illegal character in block scalar header: '" &
-          c.lexer.buf[c.lexer.bufpos] & "'")
-  c.recentWasMoreIndented = false
-  c.lineEnding()
-  result = c.handleLineEnd(true)
-  if not result:
-    startScalar(t)
-    c.content.reset()
-
-proc blockScalarLine(c: ParserContext):
-    bool {.raises: [YamlParserError].} =
-  debug("lex: blockScalarLine")
-  result = false
-  if c.level.indentation == UnknownIndentation:
-    if c.lexer.buf[c.lexer.bufpos] in lineEnd:
-      return c.handleLineEnd(true)
-    else:
-      c.level.indentation = c.indentation
-      c.content.addMultiple('\l', c.newlines)
-  elif c.indentation > c.level.indentation or
-      c.lexer.buf[c.lexer.bufpos] == '\t':
-    c.content.addMultiple('\l', c.newlines)
-    c.recentWasMoreIndented = true
-    c.content.addMultiple(' ', c.indentation - c.level.indentation)
-  elif c.scalarType == stFolded:
-    if c.recentWasMoreIndented:
-      c.recentWasMoreIndented = false
-      c.newlines.inc()
-    if c.newlines == 0: discard
-    elif c.newlines == 1: c.content.add(' ')
-    else: c.content.addMultiple('\l', c.newlines - 1)
-  else: c.content.addMultiple('\l', c.newlines)
-  c.newlines = 0
-  while c.lexer.buf[c.lexer.bufpos] notin lineEnd:
-    c.content.add(c.lexer.buf[c.lexer.bufpos])
-    c.lexer.bufpos.inc()
-  result = c.handleLineEnd(true)
-
-proc tagHandle(c: ParserContext, shorthandEnd: var int)
-    {.raises: [YamlParserError].} =
-  debug("lex: tagHandle")
-  shorthandEnd = 0
-  c.content.add(c.lexer.buf[c.lexer.bufpos])
-  var i = 0
-  while true:
-    c.lexer.bufpos.inc()
-    i.inc()
-    let ch = c.lexer.buf[c.lexer.bufpos]
-    case ch
-    of spaceOrLineEnd:
-      if shorthandEnd == -1:
-        raise c.lexer.generateError("Unclosed verbatim tag")
-      break
-    of '!':
-      if shorthandEnd == -1 and i == 2:
-        c.content.add(ch)
-        continue
-      elif shorthandEnd != 0:
-        raise c.lexer.generateError("Illegal character in tag suffix")
-      shorthandEnd = i
-      c.content.add(ch)
-    of 'a' .. 'z', 'A' .. 'Z', '0' .. '9', '#', ';', '/', '?', ':', '@', '&',
-       '-', '=', '+', '$', '_', '.', '~', '*', '\'', '(', ')':
-      c.content.add(ch)
-    of ',':
-      if shortHandEnd > 0: break # ',' after shorthand is flow indicator
-      c.content.add(ch)
-    of '<':
-      if i == 1:
-        shorthandEnd = -1
-        c.content.reset()
-      else: raise c.lexer.generateError("Illegal character in tag handle")
-    of '>':
-      if shorthandEnd == -1:
-        c.lexer.bufpos.inc()
-        if c.lexer.buf[c.lexer.bufpos] notin spaceOrLineEnd:
-          raise c.lexer.generateError("Missing space after verbatim tag handle")
-        break
-      else: raise c.lexer.generateError("Illegal character in tag handle")
-    of '%':
-      if shorthandEnd != 0: c.content.add(c.lexer.byteSequence())
-      else: raise c.lexer.generateError("Illegal character in tag handle")
-    else: raise c.lexer.generateError("Illegal character in tag handle")
-
 proc handleTagHandle(c: ParserContext) {.raises: [YamlParserError].} =
-  c.startToken()
   if c.level.kind != fplUnknown: raise c.generateError("Unexpected tag handle")
   if c.tag != yTagQuestionMark:
     raise c.generateError("Only one tag handle is allowed per node")
-  c.content.reset()
+  if c.lex.cur == ltTagHandle:
-  var
+    var tagUri = ""
-    shorthandEnd: int
-  c.tagHandle(shorthandEnd)
-  if shorthandEnd != -1:
     try:
-      c.tagUri.reset()
+      tagUri.add(c.shorthands[c.lex.buf[0..c.lex.shorthandEnd]])
-      c.tagUri.add(c.shorthands[c.content[0..shorthandEnd]])
+      tagUri.add(c.lex.buf[c.lex.shorthandEnd + 1 .. ^1])
-      c.tagUri.add(c.content[shorthandEnd + 1 .. ^1])
     except KeyError:
       raise c.generateError(
-          "Undefined tag shorthand: " & c.content[0..shorthandEnd])
+          "Undefined tag shorthand: " & c.lex.buf[0..c.lex.shorthandEnd])
-    try: c.tag = c.p.tagLib.tags[c.tagUri]
+    try: c.tag = c.p.tagLib.tags[tagUri]
-    except KeyError: c.tag = c.p.tagLib.registerUri(c.tagUri)
+    except KeyError: c.tag = c.p.tagLib.registerUri(tagUri)
   else:
-    try: c.tag = c.p.tagLib.tags[c.content]
+    try: c.tag = c.p.tagLib.tags[c.lex.buf]
-    except KeyError: c.tag = c.p.tagLib.registerUri(c.content)
+    except KeyError: c.tag = c.p.tagLib.registerUri(c.lex.buf)
-
-proc consumeLineIfEmpty(c: ParserContext, newlines: var int): bool =
-  result = true
-  while true:
-    c.lexer.bufpos.inc()
-    case c.lexer.buf[c.lexer.bufpos]
-    of ' ', '\t': discard
-    of '\l':
-      c.lexer.lexLF()
-      break
-    of '\c':
-      c.lexer.lexCR()
-      break
-    of '#', EndOfFile:
-      c.lineEnding()
-      discard c.handleLineEnd(true)
-      break
-    else:
-      result = false
-      break
 
 proc handlePossibleMapStart(c: ParserContext, e: var YamlStreamEvent,
     flow: bool = false, single: bool = false): bool =
   result = false
   if c.level.indentation == UnknownIndentation:
-    var flowDepth = 0
+    if c.lex.isImplicitKeyStart():
-    var pos = c.lexer.bufpos
+      e = c.objectStart(yamlStartMap, single)
-    var recentJsonStyle = false
+      result = true
-    while pos < c.lexer.bufpos + 1024:
-      case c.lexer.buf[pos]
-      of ':':
-        if flowDepth == 0 and (c.lexer.buf[pos + 1] in spaceOrLineEnd or
-            recentJsonStyle):
-          e = c.objectStart(yamlStartMap, single)
-          result = true
-          break
-      of lineEnd: break
-      of '[', '{': flowDepth.inc()
-      of '}', ']':
-        flowDepth.inc(-1)
-        if flowDepth < 0: break
-      of '?', ',':
-        if flowDepth == 0: break
-      of '#':
-        if c.lexer.buf[pos - 1] in space: break
-      of '"':
-        pos.inc()
-        while c.lexer.buf[pos] notin {'"', EndOfFile, '\l', '\c'}:
-          if c.lexer.buf[pos] == '\\': pos.inc()
-          pos.inc()
-        if c.lexer.buf[pos] != '"': break
-      of '\'':
-        pos.inc()
-        while c.lexer.buf[pos] notin {'\'', '\l', '\c', EndOfFile}:
-          pos.inc()
-      of '&', '*', '!':
-        if pos == c.lexer.bufpos or c.lexer.buf[c.lexer.bufpos] in space:
-          pos.inc()
-          while c.lexer.buf[pos] notin spaceOrLineEnd:
-            pos.inc()
-          continue
-      else: discard
-      if flow and c.lexer.buf[pos] notin space:
-        recentJsonStyle = c.lexer.buf[pos] in {']', '}', '\'', '"'}
-      pos.inc()
-    if c.level.indentation == UnknownIndentation:
-      c.level.indentation = c.indentation
 
 proc handleMapKeyIndicator(c: ParserContext, e: var YamlStreamEvent): bool =
   result = false
-  c.startToken()
   case c.level.kind
   of fplUnknown:
     e = c.objectStart(yamlStartMap)
     result = true
   of fplMapValue:
-    if c.level.indentation != c.indentation:
+    if c.level.indentation != c.lex.indentation:
       raise c.generateError("Invalid p.indentation of map key indicator")
     e = scalarEvent("", yTagQuestionMark, yAnchorNone)
     result = true
@@ -789,7 +172,7 @@ proc handleMapKeyIndicator(c: ParserContext, e: var YamlStreamEvent): bool =
     c.ancestry.add(c.level)
     c.level = initLevel(fplUnknown)
   of fplMapKey:
-    if c.level.indentation != c.indentation:
+    if c.level.indentation != c.lex.indentation:
       raise c.generateError("Invalid p.indentation of map key indicator")
     c.ancestry.add(c.level)
     c.level = initLevel(fplUnknown)
@@ -800,25 +183,26 @@ proc handleMapKeyIndicator(c: ParserContext, e: var YamlStreamEvent): bool =
         "Unexpected map key indicator (expected multiline scalar end)")
   of fplSinglePairKey, fplSinglePairValue, fplDocument:
     internalError("Unexpected level kind: " & $c.level.kind)
-  c.lexer.skipWhitespace()
+  # TODO: why was this there?
-  c.indentation = c.lexer.getColNumber(c.lexer.bufpos)
+  # c.lexer.skipWhitespace()
+  # c.indentation = c.lexer.getColNumber(c.lexer.bufpos)
 
 proc handleBlockSequenceIndicator(c: ParserContext, e: var YamlStreamEvent):
     bool =
   result = false
-  c.startToken()
   case c.level.kind
   of fplUnknown:
     e = c.objectStart(yamlStartSeq)
     result = true
   of fplSequence:
-    if c.level.indentation != c.indentation:
+    if c.level.indentation != c.lex.indentation:
       raise c.generateError("Invalid p.indentation of block sequence indicator")
     c.ancestry.add(c.level)
     c.level = initLevel(fplUnknown)
   else: raise c.generateError("Illegal sequence item in map")
-  c.lexer.skipWhitespace()
+  # TODO: why was this there?
-  c.indentation = c.lexer.getColNumber(c.lexer.bufpos)
+  # c.lexer.skipWhitespace()
+  # c.indentation = c.lexer.getColNumber(c.lexer.bufpos)
 
 proc handleBlockItemStart(c: ParserContext, e: var YamlStreamEvent): bool =
   result = false
@@ -846,42 +230,10 @@ proc handleFlowItemStart(c: ParserContext, e: var YamlStreamEvent): bool =
     result = c.handlePossibleMapStart(e, true, true)
 
 proc handleFlowPlainScalar(c: ParserContext, e: var YamlStreamEvent) =
-  c.content.reset()
+  while c.lex.cur in {ltScalarPart, ltEmptyLine}:
-  c.startToken()
+    c.lex.newlines.inc()
-  c.plainScalar(cFlow)
+    c.lex.next()
-  if c.lexer.buf[c.lexer.bufpos] in {'{', '}', '[', ']', ',', ':', '#'}:
+  c.lex.newlines = 0
-    discard
-  else:
-    c.newlines = 0
-    while true:
-      case c.lexer.buf[c.lexer.bufpos]
-      of ':':
-        if c.lexer.isPlainSafe(c.lexer.bufpos + 1, cFlow):
-          if c.newlines == 1:
-            c.content.add(' ')
-            c.newlines = 0
-          elif c.newlines > 1:
-            c.content.addMultiple(' ', c.newlines - 1)
-            c.newlines = 0
-          c.plainScalar(cFlow)
-        break
-      of '#', EndOfFile: break
-      of '\l':
-        c.lexer.bufpos = c.lexer.handleLF(c.lexer.bufpos)
-        c.newlines.inc()
-      of '\c':
-        c.lexer.bufpos = c.lexer.handleCR(c.lexer.bufpos)
-        c.newlines.inc()
-      of flowIndicators: break
-      of ' ', '\t': c.lexer.skipWhitespace()
-      else:
-        if c.newlines == 1:
-          c.content.add(' ')
-          c.newlines = 0
-        elif c.newlines > 1:
-          c.content.addMultiple(' ', c.newlines - 1)
-          c.newlines = 0
-        c.plainScalar(cFlow)
   e = c.currentScalar()
 
 # --- macros for defining parser states ---
@@ -965,7 +317,7 @@ parserStates(initial, blockObjectStart, blockAfterPlainScalar, blockAfterObject,
              leaveFlowSinglePairMap)
 
 proc closeEverything(c: ParserContext) =
-  c.indentation = -1
+  c.lex.indentation = -1
   c.nextImpl = stateCloseMoreIndentedLevels
   c.atSequenceItem = false
 
@@ -1068,59 +420,42 @@ proc leaveFlowLevel(c: ParserContext, e: var YamlStreamEvent): bool =
   c.nextImpl = stateObjectEnd
 
 parserState initial:
-  case c.lexer.buf[c.lexer.bufpos]
+  c.lex.next()
-  of '%':
+  case c.lex.cur
-    var ld: LexedDirective
+  of ltYamlDirective:
-    c.startToken()
+    c.lex.next()
-    c.lexer.directiveName(ld)
+    assert c.lex.cur == ltYamlVersion
-    case ld
+    if c.lex.buf != "1.2":
-    of ldYaml:
+      c.callCallback("Version is not 1.2, but " & c.lex.buf)
-      var version = ""
+  of ltTagDirective:
-      c.startToken()
+    c.lex.next()
-      c.lexer.yamlVersion(version)
+    assert c.lex.cur == ltTagShorthand
-      if version != "1.2":
+    var tagShorthand: string
-        c.callCallback("Version is not 1.2, but " & version)
+    shallowCopy(tagShorthand, c.lex.buf)
-      c.lineEnding()
+    c.lex.buf = ""
-      discard c.handleLineEnd(true)
+    c.lex.next()
-    of ldTag:
+    assert c.lex.cur == ltTagUri
-      var shorthand = ""
+    c.shorthands[tagShorthand] = c.lex.buf
-      c.tagUri.reset()
+    c.lex.buf.setLen(0)
-      c.startToken()
+  of ltUnknownDirective:
-      c.lexer.tagShorthand(shorthand)
+    c.callCallback("Unknown directive: " & c.lex.buf)
-      c.lexer.tagUriMapping(c.tagUri)
+    c.lex.buf.setLen(0)
-      c.shorthands[shorthand] = c.tagUri
+    c.lex.next()
-      c.lineEnding()
+    assert c.lex.cur == ltUnknownDirectiveParams
-      discard c.handleLineEnd(true)
+  of ltIndentation:
-    of ldUnknown:
-      c.callCallback("Unknown directive")
-      c.lexer.finishLine()
-      discard c.handleLineEnd(true)
-  of ' ', '\t':
-    if not c.consumeLineIfEmpty(c.newlines):
-      c.indentation = c.lexer.getColNumber(c.lexer.bufpos)
-      e = startDocEvent()
-      result = true
-      state = blockObjectStart
-  of '\l': c.lexer.lexLF()
-  of '\c': c.lexer.lexCR()
-  of EndOfFile: c.isFinished = true
-  of '#':
-    c.lineEnding()
-    discard c.handleLineEnd(true)
-  of '-':
-    var success: bool
-    c.startToken()
-    c.lexer.directivesEndMarker(success)
-    if success: c.lexer.bufpos.inc(3)
     e = startDocEvent()
     result = true
     state = blockObjectStart
-  else:
+  of ltStreamEnd: c.isFinished = true
+  of ltDirectivesEnd:
     e = startDocEvent()
     result = true
     state = blockObjectStart
+  else: internalError("Unexpected lexer token: " & $c.lex.cur)
 
 parserState blockObjectStart:
+  c.next()
+
   c.lexer.skipIndentation()
   c.indentation = c.lexer.getColNumber(c.lexer.bufpos)
   if c.indentation == 0:
@@ -1153,8 +488,6 @@ parserState blockObjectStart:
         stored = afterDocument
       return false
     else:
-      c.atSequenceItem = c.lexer.buf[c.lexer.bufpos] == '-' and
-          not c.lexer.isPlainSafe(c.lexer.bufpos + 1, cBlock)
       state = closeMoreIndentedLevels
       stored = blockObjectStart
       return false
@@ -1808,15 +1141,19 @@ parserState flowAfterObject:
 
 # --- parser initialization ---
 
-proc parse*(p: YamlParser, s: Stream): YamlStream =
+proc init(c: ParserContext, p: YamlParser) =
-  result = new(ParserContext)
-  let c = ParserContext(result)
-  c.content = ""
-  c.after = ""
-  c.tagUri = ""
-  c.ancestry = newSeq[FastParseLevel]()
   c.p = p
-  try: p.lexer.open(s)
+  c.ancestry = newSeq[FastParseLevel]()
+  c.initDocValues()
+  c.flowdepth = 0
+  c.isFinished = false
+  c.peeked = false
+  c.nextImpl = stateInitial
+
+proc parse*(p: YamlParser, s: Stream): YamlStream =
+  let c = new(ParserContext)
+  c.init(p)
+  try: c.lex = newYamlLexer(s)
   except:
     let e = newException(YamlParserError,
         "Error while opening stream: " & getCurrentExceptionMsg())
@@ -1825,9 +1162,10 @@ proc parse*(p: YamlParser, s: Stream): YamlStream =
     e.column = 1
     e.lineContent = ""
     raise e
-  c.initDocValues()
+  result = c
-  c.atSequenceItem = false
+
-  c.flowdepth = 0
+proc parse*(p: YamlParser, str: string): YamlStream =
-  result.isFinished = false
+  let c = new(ParserContext)
-  result.peeked = false
+  c.init(p)
-  result.nextImpl = stateInitial
+  c.lex = newYamlLexer(str)
+  result = c

yaml.nim

@@ -17,7 +17,7 @@
 ## this enhances interoperability with other languages.
 
 import streams, unicode, lexbase, tables, strutils, json, hashes, queues,
-       macros, typetraits, parseutils
+       macros, typetraits, parseutils, private/lex
 export streams, tables, json
 
 when defined(yamlDebug): import terminal
@@ -143,14 +143,6 @@ type
     ##   ``1.2``.
     ## - If there is an unknown directive encountered.
 
-  FastParseLevelKind = enum
-    fplUnknown, fplSequence, fplMapKey, fplMapValue, fplSinglePairKey,
-    fplSinglePairValue, fplScalar, fplDocument
-
-  FastParseLevel = object
-    kind: FastParseLevelKind
-    indentation: int
-
   YamlParser* = ref object
     ## A parser object. Retains its ``TagLibrary`` across calls to
     ## `parse <#parse,YamlParser,Stream>`_. Can be used
@@ -160,8 +152,6 @@ type
     tagLib: TagLibrary
     callback: WarningCallback
     anchors: Table[string, AnchorId]
-    lexer: BaseLexer
-    tokenstart: int
 
   PresentationStyle* = enum
     ## Different styles for YAML character stream output.