break up twtstr somewhat

Avoid computing e.g. charwidth data for http which does not need it
at all.

20 files changed, 402 insertions, 382 deletions

diff --git a/Makefile b/Makefile
index 2238170c..023f66db 100644
--- a/Makefile
+++ b/Makefile
@@ -82,12 +82,14 @@ $(OUTDIR_CGI_BIN)/data: adapter/protocol/data.nim src/utils/twtstr.nim
 	$(NIMC) $(FLAGS) -o:"$(OUTDIR_CGI_BIN)/data" adapter/protocol/data.nim
 
 $(OUTDIR_CGI_BIN)/file: adapter/protocol/file.nim adapter/protocol/dirlist.nim \
-		src/utils/twtstr.nim src/loader/connecterror.nim
+		src/utils/twtstr.nim src/utils/strwidth.nim src/data/charwidth.nim \
+		res/map/EastAsianWidth.txt src/loader/connecterror.nim
 	$(NIMC) $(FLAGS) -o:"$(OUTDIR_CGI_BIN)/file" adapter/protocol/file.nim
 
 $(OUTDIR_CGI_BIN)/ftp: adapter/protocol/ftp.nim src/bindings/curl.nim \
 		adapter/protocol/dirlist.nim src/utils/twtstr.nim src/types/url.nim \
-		src/types/opt.nim src/loader/connecterror.nim
+		src/utils/twtstr.nim src/utils/strwidth.nim src/data/charwidth.nim \
+		res/map/EastAsianWidth.txt src/types/opt.nim src/loader/connecterror.nim
 	$(NIMC) $(FLAGS) -d:curlLibName:$(CURLLIBNAME) \
 		-o:"$(OUTDIR_CGI_BIN)/ftp" adapter/protocol/ftp.nim
 
diff --git a/adapter/protocol/dirlist.nim b/adapter/protocol/dirlist.nim
index 2328cd55..33dcc6ab 100644
--- a/adapter/protocol/dirlist.nim
+++ b/adapter/protocol/dirlist.nim
@@ -1,5 +1,6 @@
 import algorithm
 
+import utils/strwidth
 import utils/twtstr
 
 type DirlistItemType = enum
diff --git a/src/display/lineedit.nim b/src/display/lineedit.nim
index 6dccdded..5edb31d0 100644
--- a/src/display/lineedit.nim
+++ b/src/display/lineedit.nim
@@ -7,6 +7,7 @@ import display/winattrs
 import js/javascript
 import types/cell
 import types/opt
+import utils/strwidth
 import utils/twtstr
 
 import chakasu/charset
diff --git a/src/display/term.nim b/src/display/term.nim
index 3a3a7773..05d54f98 100644
--- a/src/display/term.nim
+++ b/src/display/term.nim
@@ -14,6 +14,7 @@ import display/winattrs
 import types/cell
 import types/color
 import types/opt
+import utils/strwidth
 import utils/twtstr
 
 import chakasu/charset
diff --git a/src/html/dom.nim b/src/html/dom.nim
index 70dc7a84..aa6a3d0e 100644
--- a/src/html/dom.nim
+++ b/src/html/dom.nim
@@ -35,6 +35,7 @@ import types/referer
 import types/url
 import types/vector
 import utils/mimeguess
+import utils/strwidth
 import utils/twtstr
 
 import chakasu/charset
diff --git a/src/layout/engine.nim b/src/layout/engine.nim
index 3219f06d..637285a6 100644
--- a/src/layout/engine.nim
+++ b/src/layout/engine.nim
@@ -8,6 +8,7 @@ import css/values
 import display/winattrs
 import layout/box
 import layout/layoutunit
+import utils/strwidth
 import utils/twtstr
 
 type
diff --git a/src/local/container.nim b/src/local/container.nim
index 355549cd..09ec5a37 100644
--- a/src/local/container.nim
+++ b/src/local/container.nim
@@ -24,6 +24,7 @@ import types/color
 import types/cookie
 import types/url
 import utils/mimeguess
+import utils/strwidth
 import utils/twtstr
 
 import chakasu/charset
diff --git a/src/local/pager.nim b/src/local/pager.nim
index 6977a4eb..a6a4d77b 100644
--- a/src/local/pager.nim
+++ b/src/local/pager.nim
@@ -39,6 +39,7 @@ import types/cookie
 import types/opt
 import types/urimethodmap
 import types/url
+import utils/strwidth
 import utils/twtstr
 
 import chakasu/charset
diff --git a/src/local/select.nim b/src/local/select.nim
index ed543638..95c89117 100644
--- a/src/local/select.nim
+++ b/src/local/select.nim
@@ -3,6 +3,7 @@ import unicode
 import js/regex
 import server/buffer
 import types/cell
+import utils/strwidth
 import utils/twtstr
 
 type
diff --git a/src/main.nim b/src/main.nim
index b3742904..0d04bf3f 100644
--- a/src/main.nim
+++ b/src/main.nim
@@ -16,6 +16,7 @@ import config/config
 import io/serversocket
 import local/client
 import types/opt
+import utils/strwidth
 import utils/twtstr
 
 import chakasu/charset
diff --git a/src/render/renderdocument.nim b/src/render/renderdocument.nim
index 61eb0d72..f1404c6d 100644
--- a/src/render/renderdocument.nim
+++ b/src/render/renderdocument.nim
@@ -9,7 +9,7 @@ import layout/engine
 import layout/layoutunit
 import types/cell
 import types/color
-import utils/twtstr
+import utils/strwidth
 
 func toFormat(computed: CSSComputedValues): Format =
   if computed == nil:
diff --git a/src/render/rendertext.nim b/src/render/rendertext.nim
index 012e01e7..39380ee0 100644
--- a/src/render/rendertext.nim
+++ b/src/render/rendertext.nim
@@ -2,6 +2,7 @@ import streams
 import unicode
 
 import types/cell
+import utils/strwidth
 import utils/twtstr
 
 import chakasu/charset
diff --git a/src/server/buffer.nim b/src/server/buffer.nim
index 77404c7c..4138f1e3 100644
--- a/src/server/buffer.nim
+++ b/src/server/buffer.nim
@@ -50,6 +50,7 @@ import types/formdata
 import types/opt
 import types/referer
 import types/url
+import utils/strwidth
 import utils/twtstr
 import xhr/formdata as formdata_impl
 
diff --git a/src/server/forkserver.nim b/src/server/forkserver.nim
index 5563817b..7da7d3f6 100644
--- a/src/server/forkserver.nim
+++ b/src/server/forkserver.nim
@@ -18,7 +18,7 @@ import types/buffersource
 import types/cookie
 import types/urimethodmap
 import types/url
-import utils/twtstr
+import utils/strwidth
 
 type
   ForkCommand* = enum
diff --git a/src/types/cell.nim b/src/types/cell.nim
index 3a8967ee..a0ba3459 100644
--- a/src/types/cell.nim
+++ b/src/types/cell.nim
@@ -3,6 +3,7 @@ import tables
 
 import css/stylednode
 import types/color
+import utils/strwidth
 import utils/twtstr
 
 type
diff --git a/src/types/url.nim b/src/types/url.nim
index e115d54c..d5fbed9e 100644
--- a/src/types/url.nim
+++ b/src/types/url.nim
@@ -1,12 +1,16 @@
 # See https://url.spec.whatwg.org/#url-parsing.
-import strutils
-import tables
-import options
-import unicode
-import math
-
+import std/algorithm
+import std/math
+import std/options
+import std/strutils
+import std/tables
+import std/unicode
+
+import bindings/libunicode
+import data/idna
 import js/error
 import js/javascript
+import lib/punycode
 import types/blob
 import utils/twtstr
 
@@ -240,6 +244,106 @@ func endsInNumber(input: string): bool =
     return true
   return false
 
+type u32pair {.packed.} = object
+  a: uint32
+  b: uint32
+
+func cmpRange(x: u32pair, y: uint32): int =
+  if x.a < y:
+    return -1
+  elif x.b > y:
+    return 1
+  return 0
+
+func processIdna(str: string, checkhyphens, checkbidi, checkjoiners,
+    transitionalprocessing: bool): Option[string] =
+  var mapped: seq[Rune]
+  for r in str.runes():
+    let status = getIdnaTableStatus(r)
+    case status
+    of IDNA_DISALLOWED: return none(string) #error
+    of IDNA_IGNORED: discard
+    of IDNA_MAPPED: mapped &= getIdnaMapped(r).toRunes()
+    of IDNA_DEVIATION:
+      if transitionalprocessing:
+        mapped &= getDeviationMapped(r).toRunes()
+      else:
+        mapped &= r
+    of IDNA_VALID: mapped &= r
+  if mapped.len == 0: return
+  mapped.mnormalize()
+  var cr: CharRange
+  {.cast(noSideEffect).}:
+    cr_init(addr cr, nil, passRealloc)
+    let r = unicode_general_category(addr cr, "Mark")
+    assert r == 0
+  var labels: seq[string]
+  for label in ($mapped).split('.'):
+    if label.startsWith("xn--"):
+      try:
+        let s = punycode.decode(label.substr("xn--".len))
+        let x0 = s.toRunes()
+        let x1 = normalize(x0)
+        if x0 != x1:
+          return none(string) #error
+        if checkhyphens:
+          if s.len >= 4 and s[2] == '-' and s[3] == '-':
+            return none(string) #error
+          if s.len > 0 and s[0] == '-' and s[^1] == '-':
+            return none(string) #error
+        if x0.len > 0:
+          let cps = cast[ptr UncheckedArray[u32pair]](cr.points)
+          let c = cast[uint32](x0[0])
+          if binarySearch(toOpenArray(cps, 0, cr.len div 2 - 1), c, cmpRange) != -1:
+            return none(string) #error
+        for r in x0:
+          if r == Rune('.'):
+            return none(string) #error
+          let status = getIdnaTableStatus(r)
+          case status
+          of IDNA_DISALLOWED, IDNA_IGNORED, IDNA_MAPPED:
+            return none(string) #error
+          of IDNA_DEVIATION:
+            if transitionalprocessing:
+              return none(string) #error
+          of IDNA_VALID: discard
+          #TODO check joiners
+          #TODO check bidi
+        labels.add(s)
+      except PunyError:
+        return none(string) #error
+    else:
+      labels.add(label)
+  cr_free(addr cr)
+  return some(labels.join('.'))
+
+func unicodeToAscii(s: string, checkhyphens, checkbidi, checkjoiners,
+    transitionalprocessing, verifydnslength: bool): Option[string] =
+  let processed = s.processIdna(checkhyphens, checkbidi, checkjoiners,
+    transitionalprocessing)
+  if processed.isnone:
+    return none(string) #error
+  var labels: seq[string]
+  var all = 0
+  for label in processed.get.split('.'):
+    if not label.isAscii():
+      try:
+        let converted = "xn--" & punycode.encode(label)
+        labels.add(converted)
+      except PunyError:
+        return none(string) #error
+    else:
+      labels.add(label)
+    if verifydnslength:
+      let rl = labels[^1].runeLen()
+      if rl notin 1..63:
+        return none(string)
+      all += rl
+  if verifydnslength:
+    if all notin 1..253:
+      return none(string) #error
+  return some(labels.join('.'))
+
 func domainToAscii*(domain: string, bestrict = false): Option[string] =
   var needsprocessing = false
   for s in domain.split('.'):
diff --git a/src/utils/charcategory.nim b/src/utils/charcategory.nim
new file mode 100644
index 00000000..84a63a7e
--- /dev/null
+++ b/src/utils/charcategory.nim
@@ -0,0 +1,16 @@
+import std/unicode
+
+const C0Controls* = {chr(0x00)..chr(0x1F)}
+const Controls* = (C0Controls + {chr(0x7F)})
+const Ascii* = {chr(0x00)..chr(0x7F)}
+const AsciiUpperAlpha* = {'A'..'Z'}
+const AsciiLowerAlpha* = {'a'..'z'}
+const AsciiAlpha* = (AsciiUpperAlpha + AsciiLowerAlpha)
+const NonAscii* = {char(0x80)..char(0xFF)}
+const AsciiDigit* = {'0'..'9'}
+const AsciiAlphaNumeric* = AsciiAlpha + AsciiDigit
+const AsciiHexDigit* = (AsciiDigit + {'a'..'f', 'A'..'F'})
+const AsciiWhitespace* = {' ', '\n', '\r', '\t', '\f'}
+
+func isDigitAscii*(r: Rune): bool =
+  return uint32(r) < 256 and char(r) in AsciiDigit
diff --git a/src/utils/map.nim b/src/utils/map.nim
index 526b9154..2a3515a1 100644
--- a/src/utils/map.nim
+++ b/src/utils/map.nim
@@ -1,15 +1,15 @@
 import algorithm
 
-func searchInMap*[U, T](a: openarray[(U, T)], u: U): int =
+func searchInMap*[U, T](a: openArray[(U, T)], u: U): int =
   when not (typeof(u) is U):
     if c > cast[typeof(c)](high(U)):
       return -1
   binarySearch(a, u, proc(x: (U, T), y: U): int = cmp(x[0], y))
 
-func isInMap*[U, T](a: openarray[(U, T)], u: U): bool =
+func isInMap*[U, T](a: openArray[(U, T)], u: U): bool =
   a.searchInMap(u) != -1
 
-func isInRange*[U](a: openarray[(U, U)], u: U): bool =
+func isInRange*[U](a: openArray[(U, U)], u: U): bool =
   let res = binarySearch(a, u, proc(x: (U, U), y: U): int =
     if x[0] < y:
       -1
diff --git a/src/utils/strwidth.nim b/src/utils/strwidth.nim
new file mode 100644
index 00000000..eb4a99d1
--- /dev/null
+++ b/src/utils/strwidth.nim
@@ -0,0 +1,246 @@
+import std/strutils
+import std/unicode
+
+import data/charwidth
+import js/error
+import types/opt
+import utils/charcategory
+import utils/map
+
+# Combining chars from https://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c
+#
+# The following two functions define the column width of an ISO 10646
+# character as follows:
+#
+#   - The null character (U+0000) has a column width of 0.
+#
+#   - Other C0/C1 control characters and DEL will lead to a return value of 0
+#
+#   - Non-spacing and enclosing combining characters (general category code Mn
+#     or Me in the Unicode database) have a column width of 0.
+#
+#   - SOFT HYPHEN (U+00AD) has a column width of 1.
+#
+#   - Other format characters (general category code Cf in the Unicode
+#     database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
+#
+#   - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF) have a
+#     column width of 0.
+#
+#   - Spacing characters in the East Asian Wide (W) or East Asian Full-width
+#     (F) category as defined in Unicode Technical Report #11 have a column
+#     width of 2.
+#
+#   - All remaining characters (including all printable ISO 8859-1 and WGL4
+#     characters, Unicode control characters, etc.) have a column width of 1.
+#
+
+# sorted list of non-overlapping intervals of non-spacing characters generated
+# by "uniset +cat=Me +cat=Mn +cat=Cf -00AD +1160-11FF +200B c"
+const Combining = [
+  (0x0300u32, 0x036Fu32), (0x0483u32, 0x0486u32), (0x0488u32, 0x0489u32),
+  (0x0591u32, 0x05BDu32), (0x05BFu32, 0x05BFu32), (0x05C1u32, 0x05C2u32),
+  (0x05C4u32, 0x05C5u32), (0x05C7u32, 0x05C7u32), (0x0600u32, 0x0603u32),
+  (0x0610u32, 0x0615u32), (0x064Bu32, 0x065Eu32), (0x0670u32, 0x0670u32),
+  (0x06D6u32, 0x06E4u32), (0x06E7u32, 0x06E8u32), (0x06EAu32, 0x06EDu32),
+  (0x070Fu32, 0x070Fu32), (0x0711u32, 0x0711u32), (0x0730u32, 0x074Au32),
+  (0x07A6u32, 0x07B0u32), (0x07EBu32, 0x07F3u32), (0x0901u32, 0x0902u32),
+  (0x093Cu32, 0x093Cu32), (0x0941u32, 0x0948u32), (0x094Du32, 0x094Du32),
+  (0x0951u32, 0x0954u32), (0x0962u32, 0x0963u32), (0x0981u32, 0x0981u32),
+  (0x09BCu32, 0x09BCu32), (0x09C1u32, 0x09C4u32), (0x09CDu32, 0x09CDu32),
+  (0x09E2u32, 0x09E3u32), (0x0A01u32, 0x0A02u32), (0x0A3Cu32, 0x0A3Cu32),
+  (0x0A41u32, 0x0A42u32), (0x0A47u32, 0x0A48u32), (0x0A4Bu32, 0x0A4Du32),
+  (0x0A70u32, 0x0A71u32), (0x0A81u32, 0x0A82u32), (0x0ABCu32, 0x0ABCu32),
+  (0x0AC1u32, 0x0AC5u32), (0x0AC7u32, 0x0AC8u32), (0x0ACDu32, 0x0ACDu32),
+  (0x0AE2u32, 0x0AE3u32), (0x0B01u32, 0x0B01u32), (0x0B3Cu32, 0x0B3Cu32),
+  (0x0B3Fu32, 0x0B3Fu32), (0x0B41u32, 0x0B43u32), (0x0B4Du32, 0x0B4Du32),
+  (0x0B56u32, 0x0B56u32), (0x0B82u32, 0x0B82u32), (0x0BC0u32, 0x0BC0u32),
+  (0x0BCDu32, 0x0BCDu32), (0x0C3Eu32, 0x0C40u32), (0x0C46u32, 0x0C48u32),
+  (0x0C4Au32, 0x0C4Du32), (0x0C55u32, 0x0C56u32), (0x0CBCu32, 0x0CBCu32),
+  (0x0CBFu32, 0x0CBFu32), (0x0CC6u32, 0x0CC6u32), (0x0CCCu32, 0x0CCDu32),
+  (0x0CE2u32, 0x0CE3u32), (0x0D41u32, 0x0D43u32), (0x0D4Du32, 0x0D4Du32),
+  (0x0DCAu32, 0x0DCAu32), (0x0DD2u32, 0x0DD4u32), (0x0DD6u32, 0x0DD6u32),
+  (0x0E31u32, 0x0E31u32), (0x0E34u32, 0x0E3Au32), (0x0E47u32, 0x0E4Eu32),
+  (0x0EB1u32, 0x0EB1u32), (0x0EB4u32, 0x0EB9u32), (0x0EBBu32, 0x0EBCu32),
+  (0x0EC8u32, 0x0ECDu32), (0x0F18u32, 0x0F19u32), (0x0F35u32, 0x0F35u32),
+  (0x0F37u32, 0x0F37u32), (0x0F39u32, 0x0F39u32), (0x0F71u32, 0x0F7Eu32),
+  (0x0F80u32, 0x0F84u32), (0x0F86u32, 0x0F87u32), (0x0F90u32, 0x0F97u32),
+  (0x0F99u32, 0x0FBCu32), (0x0FC6u32, 0x0FC6u32), (0x102Du32, 0x1030u32),
+  (0x1032u32, 0x1032u32), (0x1036u32, 0x1037u32), (0x1039u32, 0x1039u32),
+  (0x1058u32, 0x1059u32), (0x1160u32, 0x11FFu32), (0x135Fu32, 0x135Fu32),
+  (0x1712u32, 0x1714u32), (0x1732u32, 0x1734u32), (0x1752u32, 0x1753u32),
+  (0x1772u32, 0x1773u32), (0x17B4u32, 0x17B5u32), (0x17B7u32, 0x17BDu32),
+  (0x17C6u32, 0x17C6u32), (0x17C9u32, 0x17D3u32), (0x17DDu32, 0x17DDu32),
+  (0x180Bu32, 0x180Du32), (0x18A9u32, 0x18A9u32), (0x1920u32, 0x1922u32),
+  (0x1927u32, 0x1928u32), (0x1932u32, 0x1932u32), (0x1939u32, 0x193Bu32),
+  (0x1A17u32, 0x1A18u32), (0x1B00u32, 0x1B03u32), (0x1B34u32, 0x1B34u32),
+  (0x1B36u32, 0x1B3Au32), (0x1B3Cu32, 0x1B3Cu32), (0x1B42u32, 0x1B42u32),
+  (0x1B6Bu32, 0x1B73u32), (0x1DC0u32, 0x1DCAu32), (0x1DFEu32, 0x1DFFu32),
+  (0x200Bu32, 0x200Fu32), (0x202Au32, 0x202Eu32), (0x2060u32, 0x2063u32),
+  (0x206Au32, 0x206Fu32), (0x20D0u32, 0x20EFu32), (0x302Au32, 0x302Fu32),
+  (0x3099u32, 0x309Au32), (0xA806u32, 0xA806u32), (0xA80Bu32, 0xA80Bu32),
+  (0xA825u32, 0xA826u32), (0xFB1Eu32, 0xFB1Eu32), (0xFE00u32, 0xFE0Fu32),
+  (0xFE20u32, 0xFE23u32), (0xFEFFu32, 0xFEFFu32), (0xFFF9u32, 0xFFFBu32),
+  (0x10A01u32, 0x10A03u32), (0x10A05u32, 0x10A06u32), (0x10A0Cu32, 0x10A0Fu32),
+  (0x10A38u32, 0x10A3Au32), (0x10A3Fu32, 0x10A3Fu32), (0x1D167u32, 0x1D169u32),
+  (0x1D173u32, 0x1D182u32), (0x1D185u32, 0x1D18Bu32), (0x1D1AAu32, 0x1D1ADu32),
+  (0x1D242u32, 0x1D244u32), (0xE0001u32, 0xE0001u32), (0xE0020u32, 0xE007Fu32),
+  (0xE0100u32, 0xE01EFu32)
+]
+
+func isDoubleWidthHigh(r: Rune): bool =
+  return DoubleWidthRanges.isInRange(uint32(r))
+
+func isDoubleWidthAmbiguousHigh(r: Rune): bool =
+  # binary search in table of non-spacing characters
+  if DoubleWidthAmbiguousRanges.isInRange(uint32(r)):
+    return true
+  return r.isDoubleWidthHigh()
+
+func isCombining(r: Rune): bool =
+  return DoubleWidthAmbiguousRanges.isInRange(uint32(r))
+
+# Lookup tables for characters on the BMP. This "only" takes up 8k of space
+# per table, as opposed to the 135k that storing all characters would require.
+# The downside is obviously that we need a binary search fallback for non-bmp.
+# We do not store a lookup table of ambiguous ranges, either.
+
+static:
+  # following won't work on 16-bit
+  doAssert sizeof(int) >= sizeof(Rune)
+
+type PropertyTable = array[0x10000 div (sizeof(int) * 8), int]
+
+type RangeMap = openarray[(uint32, uint32)]
+
+func makePropertyTable(ranges: RangeMap, skip: RangeMap = @[]): PropertyTable =
+  var ucs: uint32 = 0
+  var j = 0
+  var k = 0
+  while ucs <= 0xFFFF:
+    if k > ranges.len:
+      break
+    if ranges[k][0] > ucs:
+      ucs = ranges[k][0]
+      continue
+    if ranges[k][1] < ucs:
+      inc k
+      continue
+    if j != skip.len and ucs == skip[j][0]:
+      ucs = skip[j][1] + 1
+      continue
+    let i = ucs div (sizeof(int) * 8)
+    let m = ucs mod (sizeof(int) * 8)
+    result[i] = result[i] or (1 shl m)
+    inc ucs
+
+const DoubleWidthTable = (func(): PropertyTable =
+  var ptab = makePropertyTable(DoubleWidthRanges, Combining)
+  # Control chars return a width of 2, and are displayed as ^{letter}.
+  for c in Controls:
+    let u = int(c)
+    let i = u div (sizeof(int) * 8)
+    let m = u mod (sizeof(int) * 8)
+    ptab[i] = ptab[i] or (1 shl m)
+  return ptab
+)()
+
+const CombiningTable = makePropertyTable(Combining)
+
+# One of the few global variables in the code. Honestly, it should not exist.
+var is_cjk_ambiguous = false
+proc set_cjk_ambiguous*(b: bool) =
+  is_cjk_ambiguous = b
+
+{.push boundChecks:off.}
+func contains(props: PropertyTable, r: Rune): bool =
+  let u = int(r)
+  let i = u div (sizeof(int) * 8)
+  let m = u mod (sizeof(int) * 8)
+  return (props[i] and (1 shl m)) != 0
+{.pop.}
+
+# Warning: this shouldn't be called without normalization.
+# We could make this function more efficient in edge cases, but it's already
+# too complex for my taste.
+func width*(r: Rune): int =
+  {.cast(noSideEffect).}:
+    let u = uint32(r)
+    if u <= 0xFFFF:
+      if r in CombiningTable:
+        return 0
+      if not is_cjk_ambiguous:
+        if r in DoubleWidthTable:
+          return 2
+      else:
+        if r in DoubleWidthTable or DoubleWidthAmbiguousRanges.isInRange(u):
+          return 2
+    else:
+      if r.isCombining():
+        return 0
+      if not is_cjk_ambiguous:
+        if r.isDoubleWidthHigh():
+          return 2
+      else:
+        if r.isDoubleWidthAmbiguousHigh():
+          return 2
+    return 1
+
+# Width, but also works with tabs.
+# Needs the column width of the text so far.
+func twidth*(r: Rune, w: int): int =
+  if r != Rune('\t'):
+    return r.width()
+  return ((w div 8) + 1) * 8 - w
+
+func width*(s: string): int =
+  for r in s.runes():
+    result += r.twidth(result)
+
+func width*(s: string, start, len: int): int =
+  var i = start
+  var m = len
+  if m > s.len: m = s.len
+  while i < m:
+    var r: Rune
+    fastRuneAt(s, i, r)
+    result += r.twidth(result)
+
+func notwidth*(s: string): int =
+  for r in s.runes:
+    result += r.width()
+
+func twidth*(s: string, w: int): int =
+  var i = w
+  for r in s.runes():
+    i += r.twidth(w)
+  return i - w
+
+func padToWidth*(str: string, size: int, schar = '$'): string =
+  if str.width() < size:
+    return str & ' '.repeat(size - str.width())
+  else:
+    let size = size - 1
+    result = newStringOfCap(str.len)
+    var w = 0
+    var i = 0
+    while i < str.len:
+      var r: Rune
+      fastRuneAt(str, i, r)
+      if w + r.width <= size:
+        result &= r
+        w += r.width
+    result &= schar
+
+func breaksWord*(r: Rune): bool =
+  return not (r.isDigitAscii() or r.width() == 0 or r.isAlpha())
+
+type BoundaryFunction* = proc(x: Rune): JSResult[bool]
+
+proc breaksWord*(r: Rune, check: Opt[BoundaryFunction]): bool =
+  if check.isSome:
+    let f = check.get()
+    let v = f(r)
+    if v.isSome: #TODO report error?
+      return v.get()
+  return r.breaksWord()
diff --git a/src/utils/twtstr.nim b/src/utils/twtstr.nim
index c5078568..b235651f 100644
--- a/src/utils/twtstr.nim
+++ b/src/utils/twtstr.nim
@@ -8,28 +8,15 @@ import tables
 import unicode
 
 import bindings/libunicode
-import data/charwidth
-import data/idna
-import js/error
-import utils/map
 import types/opt
-import lib/punycode
+import utils/charcategory
+import utils/map
+
+export charcategory
 
 when defined(posix):
   import posix
 
-const C0Controls* = {chr(0x00)..chr(0x1F)}
-const Controls* = (C0Controls + {chr(0x7F)})
-const Ascii* = {chr(0x00)..chr(0x7F)}
-const AsciiUpperAlpha* = {'A'..'Z'}
-const AsciiLowerAlpha* = {'a'..'z'}
-const AsciiAlpha* = (AsciiUpperAlpha + AsciiLowerAlpha)
-const NonAscii* = (AllChars - Ascii)
-const AsciiDigit* = {'0'..'9'}
-const AsciiAlphaNumeric* = AsciiAlpha + AsciiDigit
-const AsciiHexDigit* = (AsciiDigit + {'a'..'f', 'A'..'F'})
-const AsciiWhitespace* = {' ', '\n', '\r', '\t', '\f'}
-
 func onlyWhitespace*(s: string): bool =
   for c in s:
     if c notin AsciiWhitespace:
@@ -191,9 +178,6 @@ func endsWithIgnoreCase*(s1, s2: string): bool =
       return false
   return true
 
-func isDigitAscii*(r: Rune): bool =
-  return int(r) < 256 and isDigit(char(r))
-
 func stripAndCollapse*(s: string): string =
   var i = 0
   while i < s.len and s[i] in AsciiWhitespace:
@@ -588,15 +572,16 @@ func join*(ss: openarray[string], sep: char): string =
     result &= sep
     result &= ss[i]
 
-proc passRealloc(opaque: pointer, p: pointer, size: csize_t): pointer {.cdecl.} =
+proc passRealloc*(opaque: pointer, p: pointer, size: csize_t): pointer
+    {.cdecl.} =
   return realloc(p, size)
 
 proc mnormalize*(rs: var seq[Rune], form = UNICODE_NFC) = {.cast(noSideEffect).}:
   if rs.len == 0: return
   var outbuf: ptr uint32
-  let out_len = unicode_normalize(addr outbuf,
-                                  cast[ptr uint32](unsafeAddr rs[0]),
-                                  cint(rs.len), form, nil, passRealloc)
+  let p = cast[ptr uint32](unsafeAddr rs[0])
+  let out_len = unicode_normalize(addr outbuf, p, cint(rs.len), form, nil,
+    passRealloc)
   if out_len < 0:
     raise newException(Defect, "Unicode normalization failed")
   if out_len == 0:
@@ -627,105 +612,6 @@ func normalize*(rs: seq[Rune], form = UNICODE_NFC): seq[Rune] = {.cast(noSideEff
   copyMem(addr result[0], outbuf, out_len * sizeof(uint32))
   dealloc(outbuf)
 
-type u32pair {.packed.} = object
-  a: uint32
-  b: uint32
-
-func cmpRange(x: u32pair, y: uint32): int =
-  if x.a < y:
-    return -1
-  elif x.b > y:
-    return 1
-  return 0
-
-func processIdna(str: string, checkhyphens, checkbidi, checkjoiners, transitionalprocessing: bool): Option[string] =
-  var mapped: seq[Rune]
-  for r in str.runes():
-    let status = getIdnaTableStatus(r)
-    case status
-    of IDNA_DISALLOWED: return none(string) #error
-    of IDNA_IGNORED: discard
-    of IDNA_MAPPED: mapped &= getIdnaMapped(r).toRunes()
-    of IDNA_DEVIATION:
-      if transitionalprocessing:
-        mapped &= getDeviationMapped(r).toRunes()
-      else:
-        mapped &= r
-    of IDNA_VALID: mapped &= r
-  if mapped.len == 0: return
-  mapped.mnormalize()
-  var cr: CharRange
-  {.cast(noSideEffect).}:
-    cr_init(addr cr, nil, passRealloc)
-    let r = unicode_general_category(addr cr, "Mark")
-    assert r == 0
-  var labels: seq[string]
-  for label in ($mapped).split('.'):
-    if label.startsWith("xn--"):
-      try:
-        let s = punycode.decode(label.substr("xn--".len))
-        let x0 = s.toRunes()
-        let x1 = normalize(x0)
-        if x0 != x1:
-          return none(string) #error
-        if checkhyphens:
-          if s.len >= 4 and s[2] == '-' and s[3] == '-':
-            return none(string) #error
-          if s.len > 0 and s[0] == '-' and s[^1] == '-':
-            return none(string) #error
-        if x0.len > 0:
-          let cps = cast[ptr UncheckedArray[u32pair]](cr.points)
-          let c = cast[uint32](x0[0])
-          if binarySearch(toOpenArray(cps, 0, cr.len div 2 - 1), c, cmpRange) != -1:
-            return none(string) #error
-        for r in x0:
-          if r == Rune('.'):
-            return none(string) #error
-          let status = getIdnaTableStatus(r)
-          case status
-          of IDNA_DISALLOWED, IDNA_IGNORED, IDNA_MAPPED:
-            return none(string) #error
-          of IDNA_DEVIATION:
-            if transitionalprocessing:
-              return none(string) #error
-          of IDNA_VALID: discard
-          #TODO check joiners
-          #TODO check bidi
-        labels.add(s)
-      except PunyError:
-        return none(string) #error
-    else:
-      labels.add(label)
-  cr_free(addr cr)
-  return some(labels.join('.'))
-
-func unicodeToAscii*(s: string, checkhyphens, checkbidi, checkjoiners, transitionalprocessing, verifydnslength: bool): Option[string] =
-  let processed = s.processIdna(checkhyphens, checkbidi, checkjoiners,
-                                transitionalprocessing)
-  if processed.isnone:
-    return none(string) #error
-  var labels: seq[string]
-  var all = 0
-  for label in processed.get.split('.'):
-    if not label.isAscii():
-      try:
-        let converted = "xn--" & punycode.encode(label)
-        labels.add(converted)
-      except PunyError:
-        return none(string) #error
-    else:
-      labels.add(label)
-    if verifydnslength:
-      let rl = labels[^1].runeLen()
-      if rl notin 1..63:
-        return none(string)
-      all += rl
-  if verifydnslength:
-    if all notin 1..253:
-      return none(string) #error
-  return some(labels.join('.'))
-
-
 # https://www.w3.org/TR/xml/#NT-Name
 const NameStartCharRanges = [
   (0xC0, 0xD6),
@@ -812,252 +698,6 @@ proc expandPath*(path: string): string =
         return $p.pw_dir / path.substr(usr.len)
     return path
 
-# Combining chars from https://www.cl.cam.ac.uk/~mgk25/ucs/wcwidth.c
-#
-# The following two functions define the column width of an ISO 10646
-# character as follows:
-#
-#   - The null character (U+0000) has a column width of 0.
-#
-#   - Other C0/C1 control characters and DEL will lead to a return value of 0
-#
-#   - Non-spacing and enclosing combining characters (general category code Mn
-#     or Me in the Unicode database) have a column width of 0.
-#
-#   - SOFT HYPHEN (U+00AD) has a column width of 1.
-#
-#   - Other format characters (general category code Cf in the Unicode
-#     database) and ZERO WIDTH SPACE (U+200B) have a column width of 0.
-#
-#   - Hangul Jamo medial vowels and final consonants (U+1160-U+11FF) have a
-#     column width of 0.
-#
-#   - Spacing characters in the East Asian Wide (W) or East Asian Full-width
-#     (F) category as defined in Unicode Technical Report #11 have a column
-#     width of 2.
-#
-#   - All remaining characters (including all printable ISO 8859-1 and WGL4
-#     characters, Unicode control characters, etc.) have a column width of 1.
-#
-
-# sorted list of non-overlapping intervals of non-spacing characters generated
-# by "uniset +cat=Me +cat=Mn +cat=Cf -00AD +1160-11FF +200B c"
-const Combining = [
-  (0x0300u32, 0x036Fu32), (0x0483u32, 0x0486u32), (0x0488u32, 0x0489u32),
-  (0x0591u32, 0x05BDu32), (0x05BFu32, 0x05BFu32), (0x05C1u32, 0x05C2u32),
-  (0x05C4u32, 0x05C5u32), (0x05C7u32, 0x05C7u32), (0x0600u32, 0x0603u32),
-  (0x0610u32, 0x0615u32), (0x064Bu32, 0x065Eu32), (0x0670u32, 0x0670u32),
-  (0x06D6u32, 0x06E4u32), (0x06E7u32, 0x06E8u32), (0x06EAu32, 0x06EDu32),
-  (0x070Fu32, 0x070Fu32), (0x0711u32, 0x0711u32), (0x0730u32, 0x074Au32),
-  (0x07A6u32, 0x07B0u32), (0x07EBu32, 0x07F3u32), (0x0901u32, 0x0902u32),
-  (0x093Cu32, 0x093Cu32), (0x0941u32, 0x0948u32), (0x094Du32, 0x094Du32),
-  (0x0951u32, 0x0954u32), (0x0962u32, 0x0963u32), (0x0981u32, 0x0981u32),
-  (0x09BCu32, 0x09BCu32), (0x09C1u32, 0x09C4u32), (0x09CDu32, 0x09CDu32),
-  (0x09E2u32, 0x09E3u32), (0x0A01u32, 0x0A02u32), (0x0A3Cu32, 0x0A3Cu32),
-  (0x0A41u32, 0x0A42u32), (0x0A47u32, 0x0A48u32), (0x0A4Bu32, 0x0A4Du32),
-  (0x0A70u32, 0x0A71u32), (0x0A81u32, 0x0A82u32), (0x0ABCu32, 0x0ABCu32),
-  (0x0AC1u32, 0x0AC5u32), (0x0AC7u32, 0x0AC8u32), (0x0ACDu32, 0x0ACDu32),
-  (0x0AE2u32, 0x0AE3u32), (0x0B01u32, 0x0B01u32), (0x0B3Cu32, 0x0B3Cu32),
-  (0x0B3Fu32, 0x0B3Fu32), (0x0B41u32, 0x0B43u32), (0x0B4Du32, 0x0B4Du32),
-  (0x0B56u32, 0x0B56u32), (0x0B82u32, 0x0B82u32), (0x0BC0u32, 0x0BC0u32),
-  (0x0BCDu32, 0x0BCDu32), (0x0C3Eu32, 0x0C40u32), (0x0C46u32, 0x0C48u32),
-  (0x0C4Au32, 0x0C4Du32), (0x0C55u32, 0x0C56u32), (0x0CBCu32, 0x0CBCu32),
-  (0x0CBFu32, 0x0CBFu32), (0x0CC6u32, 0x0CC6u32), (0x0CCCu32, 0x0CCDu32),
-  (0x0CE2u32, 0x0CE3u32), (0x0D41u32, 0x0D43u32), (0x0D4Du32, 0x0D4Du32),
-  (0x0DCAu32, 0x0DCAu32), (0x0DD2u32, 0x0DD4u32), (0x0DD6u32, 0x0DD6u32),
-  (0x0E31u32, 0x0E31u32), (0x0E34u32, 0x0E3Au32), (0x0E47u32, 0x0E4Eu32),
-  (0x0EB1u32, 0x0EB1u32), (0x0EB4u32, 0x0EB9u32), (0x0EBBu32, 0x0EBCu32),
-  (0x0EC8u32, 0x0ECDu32), (0x0F18u32, 0x0F19u32), (0x0F35u32, 0x0F35u32),
-  (0x0F37u32, 0x0F37u32), (0x0F39u32, 0x0F39u32), (0x0F71u32, 0x0F7Eu32),
-  (0x0F80u32, 0x0F84u32), (0x0F86u32, 0x0F87u32), (0x0F90u32, 0x0F97u32),
-  (0x0F99u32, 0x0FBCu32), (0x0FC6u32, 0x0FC6u32), (0x102Du32, 0x1030u32),
-  (0x1032u32, 0x1032u32), (0x1036u32, 0x1037u32), (0x1039u32, 0x1039u32),
-  (0x1058u32, 0x1059u32), (0x1160u32, 0x11FFu32), (0x135Fu32, 0x135Fu32),
-  (0x1712u32, 0x1714u32), (0x1732u32, 0x1734u32), (0x1752u32, 0x1753u32),
-  (0x1772u32, 0x1773u32), (0x17B4u32, 0x17B5u32), (0x17B7u32, 0x17BDu32),
-  (0x17C6u32, 0x17C6u32), (0x17C9u32, 0x17D3u32), (0x17DDu32, 0x17DDu32),
-  (0x180Bu32, 0x180Du32), (0x18A9u32, 0x18A9u32), (0x1920u32, 0x1922u32),
-  (0x1927u32, 0x1928u32), (0x1932u32, 0x1932u32), (0x1939u32, 0x193Bu32),
-  (0x1A17u32, 0x1A18u32), (0x1B00u32, 0x1B03u32), (0x1B34u32, 0x1B34u32),
-  (0x1B36u32, 0x1B3Au32), (0x1B3Cu32, 0x1B3Cu32), (0x1B42u32, 0x1B42u32),
-  (0x1B6Bu32, 0x1B73u32), (0x1DC0u32, 0x1DCAu32), (0x1DFEu32, 0x1DFFu32),
-  (0x200Bu32, 0x200Fu32), (0x202Au32, 0x202Eu32), (0x2060u32, 0x2063u32),
-  (0x206Au32, 0x206Fu32), (0x20D0u32, 0x20EFu32), (0x302Au32, 0x302Fu32),
-  (0x3099u32, 0x309Au32), (0xA806u32, 0xA806u32), (0xA80Bu32, 0xA80Bu32),
-  (0xA825u32, 0xA826u32), (0xFB1Eu32, 0xFB1Eu32), (0xFE00u32, 0xFE0Fu32),
-  (0xFE20u32, 0xFE23u32), (0xFEFFu32, 0xFEFFu32), (0xFFF9u32, 0xFFFBu32),
-  (0x10A01u32, 0x10A03u32), (0x10A05u32, 0x10A06u32), (0x10A0Cu32, 0x10A0Fu32),
-  (0x10A38u32, 0x10A3Au32), (0x10A3Fu32, 0x10A3Fu32), (0x1D167u32, 0x1D169u32),
-  (0x1D173u32, 0x1D182u32), (0x1D185u32, 0x1D18Bu32), (0x1D1AAu32, 0x1D1ADu32),
-  (0x1D242u32, 0x1D244u32), (0xE0001u32, 0xE0001u32), (0xE0020u32, 0xE007Fu32),
-  (0xE0100u32, 0xE01EFu32)
-]
-
-func cmp(range: (uint32, uint32), r: Rune): int =
-  if range[1] < cast[uint32](r):
-    return -1
-  elif range[0] > cast[uint32](r):
-    return 1
-  return 0
-
-func isDoubleWidthHigh(r: Rune): bool =
-  return binarySearch(DoubleWidthRanges, r, twtstr.cmp) != -1
-
-func isDoubleWidthAmbiguousHigh(r: Rune): bool =
-  # binary search in table of non-spacing characters
-  if binarySearch(DoubleWidthAmbiguousRanges, r, twtstr.cmp) != -1:
-    return true
-  return r.isDoubleWidthHigh()
-
-func isCombining(r: Rune): bool =
-  return binarySearch(Combining, r, twtstr.cmp) != -1
-
-# Lookup tables for characters on the BMP. This "only" takes up 8k of space
-# per table, as opposed to the 135k that storing all characters would require.
-# The downside is obviously that we need a binary search fallback for non-bmp.
-# We do not store a lookup table of ambiguous ranges, either.
-
-static:
-  # following won't work on 16-bit
-  doAssert sizeof(int) >= sizeof(Rune)
-
-type PropertyTable = array[0x10000 div (sizeof(int) * 8), int]
-
-type RangeMap = openarray[(uint32, uint32)]
-
-func makePropertyTable(ranges: RangeMap, skip: RangeMap = @[]): PropertyTable =
-  var ucs: uint32 = 0
-  var j = 0
-  var k = 0
-  while ucs <= 0xFFFF:
-    if k > ranges.len:
-      break
-    if ranges[k][0] > ucs:
-      ucs = ranges[k][0]
-      continue
-    if ranges[k][1] < ucs:
-      inc k
-      continue
-    if j != skip.len and ucs == skip[j][0]:
-      ucs = skip[j][1] + 1
-      continue
-    let i = ucs div (sizeof(int) * 8)
-    let m = ucs mod (sizeof(int) * 8)
-    result[i] = result[i] or (1 shl m)
-    inc ucs
-
-const DoubleWidthTable = (func(): PropertyTable =
-  var ptab = makePropertyTable(DoubleWidthRanges, Combining)
-  # Control chars return a width of 2, and are displayed as ^{letter}.
-  for c in Controls:
-    let u = int(c)
-    let i = u div (sizeof(int) * 8)
-    let m = u mod (sizeof(int) * 8)
-    ptab[i] = ptab[i] or (1 shl m)
-  return ptab
-)()
-
-const CombiningTable = makePropertyTable(Combining)
-
-# One of the few global variables in the code. Honestly, it should not exist.
-var is_cjk_ambiguous = false
-proc set_cjk_ambiguous*(b: bool) =
-  is_cjk_ambiguous = b
-
-{.push boundChecks:off.}
-func contains(props: PropertyTable, r: Rune): bool =
-  let u = int(r)
-  let i = u div (sizeof(int) * 8)
-  let m = u mod (sizeof(int) * 8)
-  return (props[i] and (1 shl m)) != 0
-{.pop.}
-
-# Warning: this shouldn't be called without normalization.
-# We could make this function more efficient in edge cases, but it's already
-# too complex for my taste.
-func width*(r: Rune): int =
-  {.cast(noSideEffect).}:
-    if cast[uint32](r) <= 0xFFFF:
-      if r in CombiningTable:
-        return 0
-      if not is_cjk_ambiguous:
-        if r in DoubleWidthTable:
-          return 2
-      else:
-        if r in DoubleWidthTable or
-            binarySearch(DoubleWidthAmbiguousRanges, r, twtstr.cmp) != -1:
-          return 2
-    else:
-      if r.isCombining():
-        return 0
-      if not is_cjk_ambiguous:
-        if r.isDoubleWidthHigh():
-          return 2
-      else:
-        if r.isDoubleWidthAmbiguousHigh():
-          return 2
-    return 1
-
-# Width, but also works with tabs.
-# Needs the column width of the text so far.
-func twidth*(r: Rune, w: int): int =
-  if r != Rune('\t'):
-    return r.width()
-  return ((w div 8) + 1) * 8 - w
-
-func width*(s: string): int =
-  for r in s.runes():
-    result += r.twidth(result)
-
-func width*(s: string, start, len: int): int =
-  var i = start
-  var m = len
-  if m > s.len: m = s.len
-  while i < m:
-    var r: Rune
-    fastRuneAt(s, i, r)
-    result += r.twidth(result)
-
-func notwidth*(s: string): int =
-  for r in s.runes:
-    result += r.width()
-
-func twidth*(s: string, w: int): int =
-  var i = w
-  for r in s.runes():
-    i += r.twidth(w)
-  return i - w
-
-func breaksWord*(r: Rune): bool =
-  return not (r.isDigitAscii() or r.width() == 0 or r.isAlpha())
-
-type BoundaryFunction* = proc(x: Rune): JSResult[bool]
-
-proc breaksWord*(r: Rune, check: Opt[BoundaryFunction]): bool =
-  if check.isSome:
-    let f = check.get()
-    let v = f(r)
-    if v.isSome: #TODO report error?
-      return v.get()
-  return r.breaksWord()
-
-func padToWidth*(str: string, size: int, schar = '$'): string =
-  if str.width() < size:
-    return str & ' '.repeat(size - str.width())
-  else:
-    let size = size - 1
-    result = newStringOfCap(str.len)
-
-    var w = 0
-    var i = 0
-    while i < str.len:
-      var r: Rune
-      fastRuneAt(str, i, r)
-      if w + r.width <= size:
-        result &= r
-        w += r.width
-    result &= schar
-
 func deleteChars*(s: string, todel: set[char]): string =
   var i = 0
   block earlyret: