fix #2: add charwidth function

.gitignore

@@ -9,6 +9,8 @@
 *.dylib
 *.dSYM
 *.txt
+*.ttf
+*.sfd
 *.out
 bench/bench
 bench/icu

.travis.yml

@@ -4,6 +4,11 @@ compiler:
  - clang
 notifications:
     email: false
+before_install:
+  - sudo add-apt-repository ppa:staticfloat/julia-deps -y
+  - sudo add-apt-repository ppa:staticfloat/juliareleases -y
+  - sudo apt-get update -qq -y
+  - sudo apt-get install libpcre3-dev julia -y
 script:
     - make prefix=`pwd`/local install
     - make check

Makefile

@@ -7,6 +7,7 @@ PERL=perl
 MAKE=make
 AR=ar
 INSTALL=install
+JULIA=julia
 
 # compiler settings
 cflags = -O2 -std=c99 -pedantic -Wall -fpic -DUTF8PROC_EXPORTS $(CFLAGS)
@@ -41,16 +42,19 @@ all: c-library
 c-library: libutf8proc.a libutf8proc.$(SHLIB_EXT)
 
 clean:
-	rm -f utf8proc.o libutf8proc.a libutf8proc.$(SHLIB_VERS_EXT) libutf8proc.$(SHLIB_EXT) test/normtest test/graphemetest data/UnicodeData.txt data/DerivedCoreProperties.txt data/CompositionExclusions.txt data/CaseFolding.txt data/NormalizationTest.txt data/GraphemeBreakTest.txt
+	rm -f utf8proc.o libutf8proc.a libutf8proc.$(SHLIB_VERS_EXT) libutf8proc.$(SHLIB_EXT) test/normtest test/graphemetest data/UnicodeData.txt data/DerivedCoreProperties.txt data/CompositionExclusions.txt data/CaseFolding.txt data/NormalizationTest.txt data/GraphemeBreakTest.txt data/CharWidths.txt data/unifont*.ttf data/unifont*.sfd
 	$(MAKE) -C bench clean
 
 update: utf8proc_data.c.new
 	cp -f utf8proc_data.c.new utf8proc_data.c
 
 # real targets
 
-utf8proc_data.c.new: data/data_generator.rb data/UnicodeData.txt data/GraphemeBreakProperty.txt data/DerivedCoreProperties.txt data/CompositionExclusions.txt data/CaseFolding.txt
-	(cd data; $(RUBY) data_generator.rb < UnicodeData.txt) > utf8proc_data.c.new
+utf8proc_data.c.new: data/data_generator.rb data/UnicodeData.txt data/GraphemeBreakProperty.txt data/DerivedCoreProperties.txt data/CompositionExclusions.txt data/CaseFolding.txt data/CharWidths.txt
+	(cd data; $(RUBY) data_generator.rb < UnicodeData.txt) > $@
+
+data/CharWidths.txt: data/charwidths.jl libutf8proc.$(SHLIB_EXT)
+	(cd data; $(JULIA) charwidths.jl) > $@
 
 data/UnicodeData.txt:
 	$(CURL) -o $@ -O http://www.unicode.org/Public/UNIDATA/UnicodeData.txt
@@ -85,7 +89,7 @@ libutf8proc.$(MAJOR).dylib: utf8proc.o
 	$(cc) -dynamiclib -o $@ $^ -install_name $(libdir)/$@ -Wl,-compatibility_version -Wl,$(MAJOR) -Wl,-current_version -Wl,$(MAJOR).$(MINOR).$(PATCH)
 
 libutf8proc.dylib: libutf8proc.$(MAJOR).dylib
-	ln -s libutf8proc.$(MAJOR).dylib $@
+	ln -f -s libutf8proc.$(MAJOR).dylib $@
 
 install: libutf8proc.a libutf8proc.$(SHLIB_EXT) libutf8proc.$(SHLIB_VERS_EXT)
 	mkdir -m 755 -p $(includedir)

data/charwidths.jl

@@ -0,0 +1,151 @@
+# Following work by @jiahao, we compute character widths using a combination of
+#   * advance widths from GNU Unifont (advance width 512 = 1 en)
+#   * UAX 11: East Asian Width
+#   * a few exceptions as needed
+# Adapted from http://nbviewer.ipython.org/gist/jiahao/07e8b08bf6d8671e9734
+#
+# Requires Julia (obviously) and FontForge.
+
+#############################################################################
+# Widths from GNU Unifont
+
+universion="7.0.06"
+for fontfile in ["unifont-$universion", "unifont_upper-$universion"]
+    isfile("$fontfile.ttf") || download("http://unifoundry.com/pub/unifont-$universion/font-builds/$fontfile.ttf", "$fontfile.ttf")
+    isfile("$fontfile.sfd") || run(`fontforge -lang=ff -c "Open(\"$fontfile.ttf\");Save(\"$fontfile.sfd\");Quit(0);"`)
+end
+
+#Read sfdfile for character widths
+function parsesfd(filename::String, CharWidths::Dict{Int,Int}=Dict{Int,Int}())
+    state=:seekchar
+    lineno = 0
+    for line in readlines(open(filename))
+        lineno += 1
+        if state==:seekchar         #StartChar: nonmarkingreturn
+            if contains(line, "StartChar: ")
+                codepoint = nothing
+                width = nothing
+                state = :readdata
+            end
+        elseif state==:readdata #Encoding: 65538 -1 2, Width: 1024
+            contains(line, "Encoding:") && (codepoint = int(split(line)[3]))
+            contains(line, "Width:") && (width = int(split(line)[2]))
+            if codepoint!=nothing && width!=nothing && codepoint >= 0
+                CharWidths[codepoint]=width
+                state = :seekchar
+            end
+        end
+    end
+    CharWidths
+end
+CharWidths=parsesfd("unifont-$universion.sfd")
+CharWidths=parsesfd("unifont_upper-$universion.sfd", CharWidths)
+
+# convert from advance width (512 units to the en) to character width
+for (c,v) in CharWidths
+    CharWidths[c] = div(v, 512)
+end
+
+#############################################################################
+# Widths from UAX #11: East Asian Width
+
+isfile("EastAsianWidth.txt") || download("http://www.unicode.org/Public/UNIDATA/EastAsianWidth.txt", "EastAsianWidth.txt")
+for line in readlines(open("EastAsianWidth.txt"))
+    #Strip comments
+    line[1] == '#' && continue
+    precomment = split(line, '#')[1]
+    #Parse code point range and width code
+    tokens = split(precomment, ';')
+    length(tokens) >= 2 || continue
+    charrange = tokens[1]
+    width = strip(tokens[2])
+    #Parse code point range into Julia UnitRange
+    rangetokens = split(charrange, "..")
+    charstart = uint32("0x"*rangetokens[1])
+    charend = uint32("0x"*rangetokens[length(rangetokens)>1 ? 2 : 1])
+
+    #Assign widths
+    for c in charstart:charend
+        width=="N" && continue #Ignore neutral characters
+        CharWidths[c]=(width=="W" || width=="F") ? 2 : #Wide or full
+          (width=="Na"|| width=="H" || width=="A") ? 1 : #Narrow or half or ambiguous (default to narrow in non-East-Asian contexts, which we can assume to be the default)
+          error("Unknown East Asian width code: $width for code point: $c")
+    end
+end
+
+#############################################################################
+# A few exceptions to the above cases, found by manual comparison
+# to other wcwidth functions.
+
+# Use ../libutf8proc for category codes, rather than the one in Julia,
+# to minimize bootstrapping complexity when a new version of Unicode comes out.
+function catcode(c)
+    uint(c) > 0x10FFFF && return 0x0000 # see utf8proc_get_property docs
+    return unsafe_load(ccall((:utf8proc_get_property,"../libutf8proc"), Ptr{UInt16}, (Int32,), c))
+end
+
+
+# use Base.UTF8proc module to get category codes constants, since
+# we aren't goint to change these in utf8proc.
+import Base.UTF8proc
+
+# make sure format control character (category Cf) have width 0,
+# except for the Arabic characters 0x06xx (see unicode std 6.2, sec. 8.2)
+for c in keys(CharWidths)
+    if catcode(c)==UTF8proc.UTF8PROC_CATEGORY_CF &&
+       c ∉ [0x0601,0x0602,0x0603,0x06dd]
+        CharWidths[c]=0
+    end
+end
+
+#By definition, should have zero width (on the same line)
+#0x002028 '
' category: Zl name: LINE SEPARATOR/
+#0x002029 '
' category: Zp name: PARAGRAPH SEPARATOR/
+CharWidths[0x2028]=0
+CharWidths[0x2029]=0
+
+#By definition, should be narrow = width of 1 en space
+#0x00202f ' ' category: Zs name: NARROW NO-BREAK SPACE/
+CharWidths[0x202f]=1
+
+#By definition, should be wide = width of 1 em space
+#0x002001 ' ' category: Zs name: EM QUAD/
+#0x002003 ' ' category: Zs name: EM SPACE/
+CharWidths[0x2001]=2
+CharWidths[0x2003]=2
+
+#############################################################################
+# Non-printable control characters will be assigned a width of zero
+# (wcwidth returns -1 for such characters)
+
+isprintable(c::Union(Char,Integer)) = c <= 0x10ffff && is_valid_char(c) && isprintable_category(catcode(c))
+isprintable_category(category) =
+    !(   category==UTF8proc.UTF8PROC_CATEGORY_CN # Unassigned 
+      || category==UTF8proc.UTF8PROC_CATEGORY_CS # Surrogate
+      || category==UTF8proc.UTF8PROC_CATEGORY_CC # Control
+      || category==0 # Invalid
+     )
+
+# Question: should we just use Julia's isprint algorithm here?   It is different,
+# though it is also based on the character category.
+
+#############################################################################
+# Output (to a file or pipe) for processing by data_generator.rb
+# ... don't bother to output zero widths since that will be the default.
+
+firstc = 0x000000
+lastv = 0
+uhex(c) = uppercase(hex(c,4))
+for c in 0x0000:0x110000
+    v = isprintable(c) ? get(CharWidths, c, 0) : 0
+    if v != lastv || c == 0x110000
+        v < 4 || error("invalid charwidth $v for $c")
+        if firstc+1 < c
+            println(uhex(firstc), "..", uhex(c-1), "; ", lastv)
+        else
+            println(uhex(firstc), "; ", lastv)
+        end
+        firstc = c
+        lastv = v
+    end
+end

data/data_generator.rb

@@ -85,14 +85,23 @@
   end
 end
 
+$charwidth_list = File.read("CharWidths.txt")
+$charwidth = Hash.new(0)
+$charwidth_list.each_line do |entry|
+  if entry =~ /^([0-9A-F]+)\.\.([0-9A-F]+)\s*;\s*([0-9]+)/
+    $1.hex.upto($2.hex) { |e2| $charwidth[e2] = $3.to_i }
+  elsif entry =~ /^([0-9A-F]+)\s*;\s*([0-9]+)/
+    $charwidth[$1.hex] = $2.to_i
+  end
+end
+
 $exclusions = File.read("CompositionExclusions.txt")[/# \(1\) Script Specifics.*?# Total code points:/m]
 $exclusions = $exclusions.chomp.split("\n").collect { |e| e.hex }
 
 $excl_version = File.read("CompositionExclusions.txt")[/# \(2\) Post Composition Version precomposed characters.*?# Total code points:/m]
 $excl_version = $excl_version.chomp.split("\n").collect { |e| e.hex }
 
-$case_folding_string = File.open("CaseFolding.txt").read
-
+$case_folding_string = File.open("CaseFolding.txt", :encoding => 'utf-8').read
 $case_folding = {}
 $case_folding_string.chomp.split("\n").each do |line|
   next unless line =~ /([0-9A-F]+); [CFS]; ([0-9A-F ]+);/i
@@ -172,7 +181,8 @@ def c_entry(comb1_indicies, comb2_indicies)
     "#{$exclusions.include?(code) or $excl_version.include?(code)}, " <<
     "#{$ignorable.include?(code)}, " <<
     "#{%W[Zl Zp Cc Cf].include?(category) and not [0x200C, 0x200D].include?(category)}, " <<
-    "#{$grapheme_boundclass[code]}},\n"
+    "#{$grapheme_boundclass[code]}, " <<
+    "#{$charwidth[code]}},\n"
   end
 end
 
@@ -295,7 +305,7 @@ def c_entry(comb1_indicies, comb2_indicies)
 $stdout << "};\n\n"
 
 $stdout << "const utf8proc_property_t utf8proc_properties[] = {\n"
-$stdout << "  {0, 0, 0, 0, NULL, NULL, -1, -1, -1, -1, -1, false,false,false,false, UTF8PROC_BOUNDCLASS_OTHER},\n"
+$stdout << "  {0, 0, 0, 0, NULL, NULL, -1, -1, -1, -1, -1, false,false,false,false, UTF8PROC_BOUNDCLASS_OTHER, 0},\n"
 properties.each { |line|
   $stdout << line
 }

utf8proc.c

@@ -223,6 +223,12 @@ DLLEXPORT bool utf8proc_grapheme_break(int32_t c1, int32_t c2) {
                            utf8proc_get_property(c2)->boundclass);
 }
 
+/* return a character width analogous to wcwidth (except portable and
+   hopefully less buggy than most system wcwidth functions). */
+DLLEXPORT int utf8proc_charwidth(int32_t c) {
+     return utf8proc_get_property(c)->charwidth;
+}
+
 #define utf8proc_decompose_lump(replacement_uc) \
   return utf8proc_decompose_char((replacement_uc), dst, bufsize, \
   options & ~UTF8PROC_LUMP, last_boundclass)

utf8proc.h

@@ -181,6 +181,7 @@ typedef struct utf8proc_property_struct {
   unsigned ignorable:1;
   unsigned control_boundary:1;
   unsigned boundclass:4;
+  unsigned charwidth:2;
 } utf8proc_property_t;
 
 #define UTF8PROC_CATEGORY_CN  0
@@ -388,6 +389,9 @@ DLLEXPORT bool utf8proc_grapheme_break(int32_t c1, int32_t c2);
  * permitted between them (as defined by the extended grapheme clusters in UAX#29).
  */
 
+DLLEXPORT int utf8proc_charwidth(int32_t c);
+/* Given a codepoint c, return a character width analogous to wcwidth(c). */
+
 DLLEXPORT ssize_t utf8proc_map(
   const uint8_t *str, ssize_t strlen, uint8_t **dstptr, int options
 );