#
#            Nim's Runtime Library
#        (c) Copyright 2015 Nim Contributors
#
#    See the file "copying.txt", included in this
#    distribution, for details about the copyright.
#

when defined(js):
  {.error: "This library needs to be compiled with a c-like backend, and depends on PCRE; See jsre for JS backend.".}

## What is NRE?
## ============
##
## A regular expression library for Nim using PCRE to do the hard work.
##
## For documentation on how to write patterns, there exists `the official PCRE
## pattern documentation
## <https://www.pcre.org/original/doc/html/pcrepattern.html>`_. You can also
## search the internet for a wide variety of third-party documentation and
## tools.
##
## **Note**: If you love ``sequtils.toSeq`` we have bad news for you. This
## library doesn't work with it due to documented compiler limitations. As
## a workaround, use this:
##
## .. code-block:: nim
##
##    import nre except toSeq
##
##
## Licencing
## ---------
##
## PCRE has `some additional terms`_ that you must agree to in order to use
## this module.
##
## .. _`some additional terms`: http://pcre.sourceforge.net/license.txt
##
runnableExamples:
  let vowels = re"[aeoui]"

  let expectedResults = [
    1 .. 1,
    2 .. 2,
    4 .. 4,
    6 .. 6,
    7 .. 7,
  ]
  var i = 0
  for match in "moigagoo".findIter(vowels):
    doAssert match.matchBounds == expectedResults[i]
    inc i

  let firstVowel = "foo".find(vowels)
  let hasVowel = firstVowel.isSome()
  if hasVowel:
    let matchBounds = firstVowel.get().captureBounds[-1]
    doAssert matchBounds.a == 1

  ## as with module `re`, unless specified otherwise, `start` parameter in each
  ## proc indicates where the scan starts, but outputs are relative to the start
  ## of the input string, not to `start`:
  doAssert find("uxabc", re"(?<=x|y)ab", start = 1).get.captures[-1] == "ab"
  doAssert find("uxabc", re"ab", start = 3).isNone

from pcre import nil
import nre/private/util
import tables
from strutils import `%`
from math import ceil
import options
from unicode import runeLenAt

export options

type
  Regex* = ref object
    ## Represents the pattern that things are matched against, constructed with
    ## ``re(string)``. Examples: ``re"foo"``, ``re(r"(*ANYCRLF)(?x)foo #
    ## comment".``
    ##
    ## ``pattern: string``
    ##     the string that was used to create the pattern. For details on how
    ##     to write a pattern, please see `the official PCRE pattern
    ##     documentation.
    ##     <https://www.pcre.org/original/doc/html/pcrepattern.html>`_
    ##
    ## ``captureCount: int``
    ##     the number of captures that the pattern has.
    ##
    ## ``captureNameId: Table[string, int]``
    ##     a table from the capture names to their numeric id.
    ##
    ##
    ## Options
    ## .......
    ##
    ## The following options may appear anywhere in the pattern, and they affect
    ## the rest of it.
    ##
    ## -  ``(?i)`` - case insensitive
    ## -  ``(?m)`` - multi-line: ``^`` and ``$`` match the beginning and end of
    ##    lines, not of the subject string
    ## -  ``(?s)`` - ``.`` also matches newline (*dotall*)
    ## -  ``(?U)`` - expressions are not greedy by default. ``?`` can be added
    ##    to a qualifier to make it greedy
    ## -  ``(?x)`` - whitespace and comments (``#``) are ignored (*extended*)
    ## -  ``(?X)`` - character escapes without special meaning (``\w`` vs.
    ##    ``\a``) are errors (*extra*)
    ##
    ## One or a combination of these options may appear only at the beginning
    ## of the pattern:
    ##
    ## -  ``(*UTF8)`` - treat both the pattern and subject as UTF-8
    ## -  ``(*UCP)`` - Unicode character properties; ``\w`` matches ``я``
    ## -  ``(*U)`` - a combination of the two options above
    ## -  ``(*FIRSTLINE*)`` - fails if there is not a match on the first line
    ## -  ``(*NO_AUTO_CAPTURE)`` - turn off auto-capture for groups;
    ##    ``(?<name>...)`` can be used to capture
    ## -  ``(*CR)`` - newlines are separated by ``\r``
    ## -  ``(*LF)`` - newlines are separated by ``\n`` (UNIX default)
    ## -  ``(*CRLF)`` - newlines are separated by ``\r\n`` (Windows default)
    ## -  ``(*ANYCRLF)`` - newlines are separated by any of the above
    ## -  ``(*ANY)`` - newlines are separated by any of the above and Unicode
    ##    newlines:
    ##
    ##     single characters VT (vertical tab, U+000B), FF (form feed, U+000C),
    ##     NEL (next line, U+0085), LS (line separator, U+2028), and PS
    ##     (paragraph separator, U+2029). For the 8-bit library, the last two
    ##     are recognized only in UTF-8 mode.
    ##     —  man pcre
    ##
    ## -  ``(*JAVASCRIPT_COMPAT)`` - JavaScript compatibility
    ## -  ``(*NO_STUDY)`` - turn off studying; study is enabled by default
    ##
    ## For more details on the leading option groups, see the `Option
    ## Setting <http://man7.org/linux/man-pages/man3/pcresyntax.3.html#OPTION_SETTING>`_
    ## and the `Newline
    ## Convention <http://man7.org/linux/man-pages/man3/pcresyntax.3.html#NEWLINE_CONVENTION>`_
    ## sections of the `PCRE syntax
    ## manual <http://man7.org/linux/man-pages/man3/pcresyntax.3.html>`_.
    ##
    ## Some of these options are not part of PCRE and are converted by nre
    ## into PCRE flags. These include ``NEVER_UTF``, ``ANCHORED``,
    ## ``DOLLAR_ENDONLY``, ``FIRSTLINE``, ``NO_AUTO_CAPTURE``,
    ## ``JAVASCRIPT_COMPAT``, ``U``, ``NO_STUDY``. In other PCRE wrappers, you
    ## will need to pass these as separate flags to PCRE.
    pattern*: string  ## not nil
    pcreObj: ptr pcre.Pcre  ## not nil
    pcreExtra: ptr pcre.ExtraData  ## nil

    captureNameToId: Table[string, int]

  RegexMatch* = object
    ## Usually seen as Option[RegexMatch], it represents the result of an
    ## execution. On failure, it is none, on success, it is some.
    ##
    ## ``pattern: Regex``
    ##     the pattern that is being matched
    ##
    ## ``str: string``
    ##     the string that was matched against
    ##
    ## ``captures[]: string``
    ##     the string value of whatever was captured at that id. If the value
    ##     is invalid, then behavior is undefined. If the id is ``-1``, then
    ##     the whole match is returned. If the given capture was not matched,
    ##     ``nil`` is returned.
    ##
    ##     -  ``"abc".match(re"(\w)").get.captures[0] == "a"``
    ##     -  ``"abc".match(re"(?<letter>\w)").get.captures["letter"] == "a"``
    ##     -  ``"abc".match(re"(\w)\w").get.captures[-1] == "ab"``
    ##
    ## ``captureBounds[]: HSlice[int, int]``
    ##     gets the bounds of the given capture according to the same rules as
    ##     the above. If the capture is not filled, then ``None`` is returned.
    ##     The bounds are both inclusive.
    ##
    ##     -  ``"abc".match(re"(\w)").get.captureBounds[0] == 0 .. 0``
    ##     -  ``0 in "abc".match(re"(\w)").get.captureBounds == true``
    ##     -  ``"abc".match(re"").get.captureBounds[-1] == 0 .. -1``
    ##     -  ``"abc".match(re"abc").get.captureBounds[-1] == 0 .. 2``
    ##
    ## ``match: string``
    ##     the full text of the match.
    ##
    ## ``matchBounds: HSlice[int, int]``
    ##     the bounds of the match, as in ``captureBounds[]``
    ##
    ## ``(captureBounds|captures).toTable``
    ##     returns a table with each named capture as a key.
    ##
    ## ``(captureBounds|captures).toSeq``
    ##     returns all the captures by their number.
    ##
    ## ``$: string``
    ##     same as ``match``
    pattern*: Regex  ## The regex doing the matching.
                     ## Not nil.
    str*: string  ## The string that was matched against.
                  ## Not nil.
    pcreMatchBounds: seq[HSlice[cint, cint]] ## First item is the bounds of the match
                                            ## Other items are the captures
                                            ## `a` is inclusive start, `b` is exclusive end

  Captures* = distinct RegexMatch
  CaptureBounds* = distinct RegexMatch

  RegexError* = ref object of CatchableError

  RegexInternalError* = ref object of RegexError
    ## Internal error in the module, this probably means that there is a bug

  InvalidUnicodeError* = ref object of RegexError
    ## Thrown when matching fails due to invalid unicode in strings
    pos*: int  ## the location of the invalid unicode in bytes

  SyntaxError* = ref object of RegexError
    ## Thrown when there is a syntax error in the
    ## regular expression string passed in
    pos*: int  ## the location of the syntax error in bytes
    pattern*: string  ## the pattern that caused the problem

  StudyError* = ref object of RegexError
    ## Thrown when studying the regular expression fails
    ## for whatever reason. The message contains the error
    ## code.

runnableExamples:
  # This MUST be kept in sync with the examples in RegexMatch
  doAssert "abc".match(re"(\w)").get.captures[0] == "a"
  doAssert "abc".match(re"(?<letter>\w)").get.captures["letter"] == "a"
  doAssert "abc".match(re"(\w)\w").get.captures[-1] == "ab"

  doAssert "abc".match(re"(\w)").get.captureBounds[0] == 0 .. 0
  doAssert 0 in "abc".match(re"(\w)").get.captureBounds == true
  doAssert "abc".match(re"").get.captureBounds[-1] == 0 .. -1
  doAssert "abc".match(re"abc").get.captureBounds[-1] == 0 .. 2


proc destroyRegex(pattern: Regex) =
  pcre.free_substring(cast[cstring](pattern.pcreObj))
  if pattern.pcreExtra != nil:
    pcre.free_study(pattern.pcreExtra)

proc getinfo[T](pattern: Regex, opt: cint): T =
  let retcode = pcre.fullinfo(pattern.pcreObj, pattern.pcreExtra, opt, addr result)

  if retcode < 0:
    # XXX Error message that doesn't expose implementation details
    raise newException(FieldDefect, "Invalid getinfo for $1, errno $2" % [$opt, $retcode])

proc getNameToNumberTable(pattern: Regex): Table[string, int] =
  let entryCount = getinfo[cint](pattern, pcre.INFO_NAMECOUNT)
  let entrySize = getinfo[cint](pattern, pcre.INFO_NAMEENTRYSIZE)
  let table = cast[ptr UncheckedArray[uint8]](
                getinfo[int](pattern, pcre.INFO_NAMETABLE))

  result = initTable[string, int]()

  for i in 0 ..< entryCount:
    let pos = i * entrySize
    let num = (int(table[pos]) shl 8) or int(table[pos + 1]) - 1
    var name = ""

    var idx = 2
    while table[pos + idx] != 0:
      name.add(char(table[pos + idx]))
      idx += 1

    result[name] = num

proc initRegex(pattern: string, flags: int, study = true): Regex =
  new(result, destroyRegex)
  result.pattern = pattern

  var errorMsg: cstring
  var errOffset: cint

  result.pcreObj = pcre.compile(cstring(pattern),
                                # better hope int is at least 4 bytes..
                                cint(flags), addr errorMsg,
                                addr errOffset, nil)
  if result.pcreObj == nil:
    # failed to compile
    raise SyntaxError(msg: $errorMsg, pos: errOffset, pattern: pattern)

  if study:
    var options: cint = 0
    var hasJit: cint
    if pcre.config(pcre.CONFIG_JIT, addr hasJit) == 0:
      if hasJit == 1'i32:
        options = pcre.STUDY_JIT_COMPILE
    result.pcreExtra = pcre.study(result.pcreObj, options, addr errorMsg)
    if errorMsg != nil:
      raise StudyError(msg: $errorMsg)

  result.captureNameToId = result.getNameToNumberTable()

proc captureCount*(pattern: Regex): int =
  return getinfo[cint](pattern, pcre.INFO_CAPTURECOUNT)

proc captureNameId*(pattern: Regex): Table[string, int] =
  return pattern.captureNameToId

proc matchesCrLf(pattern: Regex): bool =
  let flags = uint32(getinfo[culong](pattern, pcre.INFO_OPTIONS))
  let newlineFlags = flags and (pcre.NEWLINE_CRLF or
                                pcre.NEWLINE_ANY or
                                pcre.NEWLINE_ANYCRLF)
  if newLineFlags > 0u32:
    return true

  # get flags from build config
  var confFlags: cint
  if pcre.config(pcre.CONFIG_NEWLINE, addr confFlags) != 0:
    assert(false, "CONFIG_NEWLINE apparently got screwed up")

  case confFlags
  of 13: return false
  of 10: return false
  of (13 shl 8) or 10: return true
  of -2: return true
  of -1: return true
  else: return false


func captureBounds*(pattern: RegexMatch): CaptureBounds = return CaptureBounds(pattern)

func captures*(pattern: RegexMatch): Captures = return Captures(pattern)

func contains*(pattern: CaptureBounds, i: int): bool =
  let pattern = RegexMatch(pattern)
  pattern.pcreMatchBounds[i + 1].a != -1

func contains*(pattern: Captures, i: int): bool =
  i in CaptureBounds(pattern)

func `[]`*(pattern: CaptureBounds, i: int): HSlice[int, int] =
  let pattern = RegexMatch(pattern)
  if not (i in pattern.captureBounds):
    raise newException(IndexDefect, "Group '" & $i & "' was not captured")

  let bounds = pattern.pcreMatchBounds[i + 1]
  int(bounds.a)..int(bounds.b-1)

func `[]`*(pattern: Captures, i: int): string =
  let pattern = RegexMatch(pattern)
  let bounds = pattern.captureBounds[i]

  pattern.str.substr(bounds.a, bounds.b)

func match*(pattern: RegexMatch): string =
  return pattern.captures[-1]

func matchBounds*(pattern: RegexMatch): HSlice[int, int] =
  return pattern.captureBounds[-1]

func contains*(pattern: CaptureBounds, name: string): bool =
  let pattern = RegexMatch(pattern)
  let nameToId = pattern.pattern.captureNameToId
  if not (name in nameToId):
      return false
  nameToId[name] in pattern.captureBounds

func contains*(pattern: Captures, name: string): bool =
  name in CaptureBounds(pattern)

func checkNamedCaptured(pattern: RegexMatch, name: string) =
  if not (name in pattern.captureBounds):
    raise newException(KeyError, "Group '" & name & "' was not captured")

func `[]`*(pattern: CaptureBounds, name: string): HSlice[int, int] =
  let pattern = RegexMatch(pattern)
  checkNamedCaptured(pattern, name)
  {.noSideEffect.}:
    result = pattern.captureBounds[pattern.pattern.captureNameToId[name]]

func `[]`*(pattern: Captures, name: string): string =
  let pattern = RegexMatch(pattern)
  checkNamedCaptured(pattern, name)
  {.noSideEffect.}:
    result = pattern.captures[pattern.pattern.captureNameToId[name]]

template toTableImpl() {.dirty.} =
  for key in RegexMatch(pattern).pattern.captureNameId.keys:
    if key in pattern:
        result[key] = pattern[key]

func toTable*(pattern: Captures): Table[string, string] =
  result = initTable[string, string]()
  toTableImpl()

func toTable*(pattern: CaptureBounds): Table[string, HSlice[int, int]] =
  result = initTable[string, HSlice[int, int]]()
  toTableImpl()

template itemsImpl() {.dirty.} =
  for i in 0 ..< RegexMatch(pattern).pattern.captureCount:
    # done in this roundabout way to avoid multiple yields (potential code
    # bloat)
    let nextYieldVal = if i in pattern:
      some(pattern[i])
    else:
      default

    yield nextYieldVal

iterator items*(pattern: CaptureBounds,
                default = none(HSlice[int, int])): Option[HSlice[int, int]] =
  itemsImpl()

iterator items*(pattern: Captures,
                default: Option[string] = none(string)): Option[string] =
  itemsImpl()

proc toSeq*(pattern: CaptureBounds,
            default = none(HSlice[int, int])): seq[Option[HSlice[int, int]]] =
  result = @[]
  for it in pattern.items(default): result.add it

proc toSeq*(pattern: Captures,
            default: Option[string] = none(string)): seq[Option[string]] =
  result = @[]
  for it in pattern.items(default): result.add it

proc `$`*(pattern: RegexMatch): string =
  return pattern.captures[-1]

proc `==`*(a, b: Regex): bool =
  if not a.isNil and not b.isNil:
    return a.pattern == b.pattern and
           a.pcreObj == b.pcreObj and
           a.pcreExtra == b.pcreExtra
  else:
    return system.`==`(a, b)

proc `==`*(a, b: RegexMatch): bool =
  return a.pattern == b.pattern and
         a.str == b.str

const PcreOptions = {
  "NEVER_UTF": pcre.NEVER_UTF,
  "ANCHORED": pcre.ANCHORED,
  "DOLLAR_ENDONLY": pcre.DOLLAR_ENDONLY,
  "FIRSTLINE": pcre.FIRSTLINE,
  "NO_AUTO_CAPTURE": pcre.NO_AUTO_CAPTURE,
  "JAVASCRIPT_COMPAT": pcre.JAVASCRIPT_COMPAT,
  "U": pcre.UTF8 or pcre.UCP
}.toTable

# Options that are supported inside regular expressions themselves
const SkipOptions = [
  "LIMIT_MATCH=", "LIMIT_RECURSION=", "NO_AUTO_POSSESS", "NO_START_OPT",
  "UTF8", "UTF16", "UTF32", "UTF", "UCP",
  "CR", "LF", "CRLF", "ANYCRLF", "ANY", "BSR_ANYCRLF", "BSR_UNICODE"
]

proc extractOptions(pattern: string): tuple[pattern: string, flags: int, study: bool] =
  result = ("", 0, true)

  var optionStart = 0
  var equals = false
  for i, c in pattern:
    if optionStart == i:
      if c != '(':
        break
      optionStart = i

    elif optionStart == i-1:
      if c != '*':
        break

    elif c == ')':
      let name = pattern[optionStart+2 .. i-1]
      if equals or name in SkipOptions:
        result.pattern.add pattern[optionStart .. i]
      elif PcreOptions.hasKey name:
        result.flags = result.flags or PcreOptions[name]
      elif name == "NO_STUDY":
        result.study = false
      else:
        break
      optionStart = i+1
      equals = false

    elif not equals:
      if c == '=':
        equals = true
        if pattern[optionStart+2 .. i] notin SkipOptions:
          break
      elif c notin {'A'..'Z', '0'..'9', '_'}:
        break

  result.pattern.add pattern[optionStart .. pattern.high]

proc re*(pattern: string): Regex =
  let (pattern, flags, study) = extractOptions(pattern)
  initRegex(pattern, flags, study)

proc matchImpl(str: string, pattern: Regex, start, endpos: int, flags: int): Option[RegexMatch] =
  var myResult = RegexMatch(pattern: pattern, str: str)
  # See PCRE man pages.
  # 2x capture count to make room for start-end pairs
  # 1x capture count as slack space for PCRE
  let vecsize = (pattern.captureCount() + 1) * 3
  # div 2 because each element is 2 cints long
  # plus 1 because we need the ceiling, not the floor
  myResult.pcreMatchBounds = newSeq[HSlice[cint, cint]]((vecsize + 1) div 2)
  myResult.pcreMatchBounds.setLen(vecsize div 3)

  let strlen = if endpos == int.high: str.len else: endpos+1
  doAssert(strlen <= str.len)  # don't want buffer overflows

  let execRet = pcre.exec(pattern.pcreObj,
                          pattern.pcreExtra,
                          cstring(str),
                          cint(strlen),
                          cint(start),
                          cint(flags),
                          cast[ptr cint](addr myResult.pcreMatchBounds[0]),
                          cint(vecsize))
  if execRet >= 0:
    return some(myResult)

  case execRet:
    of pcre.ERROR_NOMATCH:
      return none(RegexMatch)
    of pcre.ERROR_NULL:
      raise newException(AccessViolationDefect, "Expected non-null parameters")
    of pcre.ERROR_BADOPTION:
      raise RegexInternalError(msg: "Unknown pattern flag. Either a bug or " &
        "outdated PCRE.")
    of pcre.ERROR_BADUTF8, pcre.ERROR_SHORTUTF8, pcre.ERROR_BADUTF8_OFFSET:
      raise InvalidUnicodeError(msg: "Invalid unicode byte sequence",
        pos: myResult.pcreMatchBounds[0].a)
    else:
      raise RegexInternalError(msg: "Unknown internal error: " & $execRet)

proc match*(str: string, pattern: Regex, start = 0, endpos = int.high): Option[RegexMatch] =
  ## Like ` ``find(...)`` <#proc-find>`_, but anchored to the start of the
  ## string.
  ##
  runnableExamples:
    doAssert "foo".match(re"f").isSome
    doAssert "foo".match(re"o").isNone

  return str.matchImpl(pattern, start, endpos, pcre.ANCHORED)

iterator findIter*(str: string, pattern: Regex, start = 0, endpos = int.high): RegexMatch =
  ## Works the same as ` ``find(...)`` <#proc-find>`_, but finds every
  ## non-overlapping match. ``"2222".find(re"22")`` is ``"22", "22"``, not
  ## ``"22", "22", "22"``.
  ##
  ## Arguments are the same as ` ``find(...)`` <#proc-find>`_
  ##
  ## Variants:
  ##
  ## -  ``proc findAll(...)`` returns a ``seq[string]``
  # see pcredemo for explanation
  let matchesCrLf = pattern.matchesCrLf()
  let unicode = uint32(getinfo[culong](pattern, pcre.INFO_OPTIONS) and
    pcre.UTF8) > 0u32
  let strlen = if endpos == int.high: str.len else: endpos+1
  var offset = start
  var match: Option[RegexMatch]
  var neverMatched = true

  while true:
    var flags = 0
    if match.isSome and
       match.get.matchBounds.a > match.get.matchBounds.b:
      # 0-len match
      flags = pcre.NOTEMPTY_ATSTART
    match = str.matchImpl(pattern, offset, endpos, flags)

    if match.isNone:
      # either the end of the input or the string
      # cannot be split here - we also need to bail
      # if we've never matched and we've already tried to...
      if offset >= strlen or neverMatched:
        break

      if matchesCrLf and offset < (str.len - 1) and
         str[offset] == '\r' and str[offset + 1] == '\L':
        # if PCRE treats CrLf as newline, skip both at the same time
        offset += 2
      elif unicode:
        # XXX what about invalid unicode?
        offset += str.runeLenAt(offset)
        assert(offset <= strlen)
      else:
        offset += 1
    else:
      neverMatched = false
      offset = match.get.matchBounds.b + 1

      yield match.get

proc find*(str: string, pattern: Regex, start = 0, endpos = int.high): Option[RegexMatch] =
  ## Finds the given pattern in the string between the end and start
  ## positions.
  ##
  ## ``start``
  ##     The start point at which to start matching. ``|abc`` is ``0``;
  ##     ``a|bc`` is ``1``
  ##
  ## ``endpos``
  ##     The maximum index for a match; ``int.high`` means the end of the
  ##     string, otherwise it’s an inclusive upper bound.
  return str.matchImpl(pattern, start, endpos, 0)

proc findAll*(str: string, pattern: Regex, start = 0, endpos = int.high): seq[string] =
  result = @[]
  for match in str.findIter(pattern, start, endpos):
    result.add(match.match)

proc contains*(str: string, pattern: Regex, start = 0, endpos = int.high): bool =
  ## Determine if the string contains the given pattern between the end and
  ## start positions:
  ## This function is equivalent to ``isSome(str.find(pattern, start, endpos))``.
  ##
  runnableExamples:
    doAssert "abc".contains(re"bc")
    doAssert not "abc".contains(re"cd")
    doAssert not "abc".contains(re"a", start = 1)

  return isSome(str.find(pattern, start, endpos))

proc split*(str: string, pattern: Regex, maxSplit = -1, start = 0): seq[string] =
  ## Splits the string with the given regex. This works according to the
  ## rules that Perl and Javascript use.
  ##
  ## ``start`` behaves the same as in ` ``find(...)`` <#proc-find>`_.
  ##
  runnableExamples:
    # -  If the match is zero-width, then the string is still split:
    doAssert "123".split(re"") == @["1", "2", "3"]

    # -  If the pattern has a capture in it, it is added after the string
    #    split:
    doAssert "12".split(re"(\d)") == @["", "1", "", "2", ""]

    # -  If ``maxsplit != -1``, then the string will only be split
    #    ``maxsplit - 1`` times. This means that there will be ``maxsplit``
    #    strings in the output seq.
    doAssert "1.2.3".split(re"\.", maxsplit = 2) == @["1", "2.3"]

  result = @[]
  var lastIdx = start
  var splits = 0
  var bounds = 0 .. -1
  var never_ran = true

  for match in str.findIter(pattern, start = start):
    never_ran = false

    # bounds are inclusive:
    #
    # 0123456
    #  ^^^
    # (1, 3)
    bounds = match.matchBounds

    # "12".split("") would be @["", "1", "2"], but
    # if we skip an empty first match, it's the correct
    # @["1", "2"]
    if bounds.a <= bounds.b or bounds.a > start:
      result.add(str.substr(lastIdx, bounds.a - 1))
      splits += 1

    lastIdx = bounds.b + 1

    for cap in match.captures:
      # if there are captures, include them in the result
      if cap.isSome:
        result.add(cap.get)

    if splits == maxSplit - 1:
      break

  # "12".split("\b") would be @["1", "2", ""], but
  # if we skip an empty last match, it's the correct
  # @["1", "2"]
  # If matches were never found, then the input string is the result
  if bounds.a <= bounds.b or bounds.b < str.high or never_ran:
    # last match: Each match takes the previous substring,
    # but "1 2".split(/ /) needs to return @["1", "2"].
    # This handles "2"
    result.add(str.substr(bounds.b + 1, str.high))

template replaceImpl(str: string, pattern: Regex,
                     replacement: untyped) {.dirty.} =
  # XXX seems very similar to split, maybe I can reduce code duplication
  # somehow?
  result = ""
  var lastIdx = 0
  for match {.inject.} in str.findIter(pattern):
    let bounds = match.matchBounds
    result.add(str.substr(lastIdx, bounds.a - 1))
    let nextVal = replacement
    result.add(nextVal)

    lastIdx = bounds.b + 1

  result.add(str.substr(lastIdx, str.len - 1))
  return result

proc replace*(str: string, pattern: Regex,
              subproc: proc (match: RegexMatch): string): string =
  ## Replaces each match of Regex in the string with ``subproc``, which should
  ## never be or return ``nil``.
  ##
  ## If ``subproc`` is a ``proc (RegexMatch): string``, then it is executed with
  ## each match and the return value is the replacement value.
  ##
  ## If ``subproc`` is a ``proc (string): string``, then it is executed with the
  ## full text of the match and and the return value is the replacement
  ## value.
  ##
  ## If ``subproc`` is a string, the syntax is as follows:
  ##
  ## -  ``$$`` - literal ``$``
  ## -  ``$123`` - capture number ``123``
  ## -  ``$foo`` - named capture ``foo``
  ## -  ``${foo}`` - same as above
  ## -  ``$1$#`` - first and second captures
  ## -  ``$#`` - first capture
  ## -  ``$0`` - full match
  ##
  ## If a given capture is missing, ``IndexDefect`` thrown for un-named captures
  ## and ``KeyError`` for named captures.
  replaceImpl(str, pattern, subproc(match))

proc replace*(str: string, pattern: Regex,
              subproc: proc (match: string): string): string =
  replaceImpl(str, pattern, subproc(match.match))

proc replace*(str: string, pattern: Regex, sub: string): string =
  # - 1 because the string numbers are 0-indexed
  replaceImpl(str, pattern,
    formatStr(sub, match.captures[name], match.captures[id - 1]))

let SpecialCharMatcher = re"([\\+*?[^\]$(){}=!<>|:-])"
proc escapeRe*(str: string): string =
  ## Escapes the string so it doesn’t match any special characters.
  ## Incompatible with the Extra flag (``X``).
  str.replace(SpecialCharMatcher, "\\$1")