正如事實證明越來越Unicode支持Flex中,除非Flex源本身增加了這將是一個痛苦。在那裏似乎有一些實驗性的東西,但從來沒有把它發佈到我能找到的版本中。
Ragel doc非常具有洞察力,並且內置了對Unicode的支持。我從那以後發現了this article,它給出了一個如何讓Ragel和C++發揮出色的例子。似乎是更好的選擇,所以這樣做。
希望這可以節省別人花費的時間來解決這個問題。
EDIT
「內置支持」 如上所述也許是誇張。獲得unicode支持一直比較容易,但它不僅僅是一種開箱即用的方式。 使用cmake我從派生的UCD 7文件生成狀態機。 在的CMakeLists.txt我做的:
#Ruby is required to generate a unicode Ragel machine
FIND_PACKAGE(Ruby REQUIRED)
MESSAGE("Found Ruby ${RUBY_VERSION}")
SET(UNICODE_MACHINE_PATH "${PROJECT_SOURCE_DIR}/src/unicode.rl")
if(NOT EXISTS ${UNICODE_MACHINE_PATH} OR gen_unicode)
MESSAGE("Attempting to generate unicode state machine")
EXECUTE_PROCESS(COMMAND ${RUBY_EXECUTABLE} ${PROJECT_SOURCE_DIR}/unicode2ragel.rb
OUTPUT_FILE ${UNICODE_MACHINE_PATH}
RESULT_VARIABLE RAGEL_UNICODE_GEN_RES)
if(${RAGEL_UNICODE_GEN_RES} EQUAL 0)
MESSAGE("Generaged Ragel Unicode state machine")
else()
MESSAGE(SEND_ERROR "Unable to generate unicode state machine")
endif()
endif()
然後在unicode2ragel.rb(附帶Ragel和UCD 7略有修改)
#!/usr/bin/env ruby
#
# This script uses the unicode spec to generate a Ragel state machine
# that recognizes unicode alphanumeric characters. It generates 5
# character classes: uupper, ulower, ualpha, udigit, and ualnum.
# Currently supported encodings are UTF-8 [default] and UCS-4.
#
# Usage: unicode2ragel.rb [options]
# -e, --encoding [ucs4 | utf8] Data encoding
# -h, --help Show this message
#
# This script was originally written as part of the Ferret search
# engine library.
#
# Author: Rakan El-Khalil <[email protected]>
require 'optparse'
require 'open-uri'
ENCODINGS = [ :utf8, :ucs4 ]
ALPHTYPES = { :utf8 => "unsigned char", :ucs4 => "unsigned int" }
CHART_URL = "http://www.unicode.org/Public/7.0.0/ucd/extracted/DerivedGeneralCategory.txt"#"http://www.unicode.org/Public/7.0.0/ucd/DerivedCoreProperties.txt"
###
# Display vars & default option
TOTAL_WIDTH = 80
RANGE_WIDTH = 23
@encoding = :utf8
###
# Option parsing
cli_opts = OptionParser.new do |opts|
opts.on("-e", "--encoding [ucs4 | utf8]", "Data encoding") do |o|
@encoding = o.downcase.to_sym
end
opts.on("-h", "--help", "Show this message") do
puts opts
exit
end
end
cli_opts.parse(ARGV)
unless ENCODINGS.member? @encoding
puts "Invalid encoding: #{@encoding}"
puts cli_opts
exit
end
##
# Downloads the document at url and yields every alpha line's hex
# range and description.
def each_alpha(url, property)
open(url) do |file|
file.each_line do |line|
next if line =~ /^#/;
next if line !~ /; #{property} #/;
range, description = line.split(/;/)
range.strip!
description.gsub!(/.*#/, '').strip!
if range =~ /\.\./
start, stop = range.split '..'
else start = stop = range
end
yield start.hex .. stop.hex, description
end
end
end
###
# Formats to hex at minimum width
def to_hex(n)
r = "%0X" % n
r = "0#{r}" unless (r.length % 2).zero?
r
end
###
# UCS4 is just a straight hex conversion of the unicode codepoint.
def to_ucs4(range)
rangestr = "0x" + to_hex(range.begin)
rangestr << "..0x" + to_hex(range.end) if range.begin != range.end
[ rangestr ]
end
##
# 0x00 - 0x7f -> 0zzzzzzz[7]
# 0x80 - 0x7ff -> 110yyyyy[5] 10zzzzzz[6]
# 0x800 - 0xffff -> 1110xxxx[4] 10yyyyyy[6] 10zzzzzz[6]
# 0x010000 - 0x10ffff -> 11110www[3] 10xxxxxx[6] 10yyyyyy[6] 10zzzzzz[6]
UTF8_BOUNDARIES = [0x7f, 0x7ff, 0xffff, 0x10ffff]
def to_utf8_enc(n)
r = 0
if n <= 0x7f
r = n
elsif n <= 0x7ff
y = 0xc0 | (n >> 6)
z = 0x80 | (n & 0x3f)
r = y << 8 | z
elsif n <= 0xffff
x = 0xe0 | (n >> 12)
y = 0x80 | (n >> 6) & 0x3f
z = 0x80 | n & 0x3f
r = x << 16 | y << 8 | z
elsif n <= 0x10ffff
w = 0xf0 | (n >> 18)
x = 0x80 | (n >> 12) & 0x3f
y = 0x80 | (n >> 6) & 0x3f
z = 0x80 | n & 0x3f
r = w << 24 | x << 16 | y << 8 | z
end
to_hex(r)
end
def from_utf8_enc(n)
n = n.hex
r = 0
if n <= 0x7f
r = n
elsif n <= 0xdfff
y = (n >> 8) & 0x1f
z = n & 0x3f
r = y << 6 | z
elsif n <= 0xefffff
x = (n >> 16) & 0x0f
y = (n >> 8) & 0x3f
z = n & 0x3f
r = x << 10 | y << 6 | z
elsif n <= 0xf7ffffff
w = (n >> 24) & 0x07
x = (n >> 16) & 0x3f
y = (n >> 8) & 0x3f
z = n & 0x3f
r = w << 18 | x << 12 | y << 6 | z
end
r
end
###
# Given a range, splits it up into ranges that can be continuously
# encoded into utf8. Eg: 0x00 .. 0xff => [0x00..0x7f, 0x80..0xff]
# This is not strictly needed since the current [5.1] unicode standard
# doesn't have ranges that straddle utf8 boundaries. This is included
# for completeness as there is no telling if that will ever change.
def utf8_ranges(range)
ranges = []
UTF8_BOUNDARIES.each do |max|
if range.begin <= max
return ranges << range if range.end <= max
ranges << range.begin .. max
range = (max + 1) .. range.end
end
end
ranges
end
def build_range(start, stop)
size = start.size/2
left = size - 1
return [""] if size < 1
a = start[0..1]
b = stop[0..1]
###
# Shared prefix
if a == b
return build_range(start[2..-1], stop[2..-1]).map do |elt|
"0x#{a} " + elt
end
end
###
# Unshared prefix, end of run
return ["0x#{a}..0x#{b} "] if left.zero?
###
# Unshared prefix, not end of run
# Range can be 0x123456..0x56789A
# Which is equivalent to:
# 0x123456 .. 0x12FFFF
# 0x130000 .. 0x55FFFF
# 0x560000 .. 0x56789A
ret = []
ret << build_range(start, a + "FF" * left)
###
# Only generate middle range if need be.
if a.hex+1 != b.hex
max = to_hex(b.hex - 1)
max = "FF" if b == "FF"
ret << "0x#{to_hex(a.hex+1)}..0x#{max} " + "0x00..0xFF " * left
end
###
# Don't generate last range if it is covered by first range
ret << build_range(b + "00" * left, stop) unless b == "FF"
ret.flatten!
end
def to_utf8(range)
utf8_ranges(range).map do |r|
build_range to_utf8_enc(r.begin), to_utf8_enc(r.end)
end.flatten!
end
##
# Perform a 3-way comparison of the number of codepoints advertised by
# the unicode spec for the given range, the originally parsed range,
# and the resulting utf8 encoded range.
def count_codepoints(code)
code.split(' ').inject(1) do |acc, elt|
if elt =~ /0x(.+)\.\.0x(.+)/
if @encoding == :utf8
acc * (from_utf8_enc($2) - from_utf8_enc($1) + 1)
else
acc * ($2.hex - $1.hex + 1)
end
else
acc
end
end
end
def is_valid?(range, desc, codes)
spec_count = 1
spec_count = $1.to_i if desc =~ /\[(\d+)\]/
range_count = range.end - range.begin + 1
sum = codes.inject(0) { |acc, elt| acc + count_codepoints(elt) }
sum == spec_count and sum == range_count
end
##
# Generate the state maching to stdout
def generate_machine(name, property)
pipe = " "
puts " #{name} = "
each_alpha(CHART_URL, property) do |range, desc|
codes = (@encoding == :ucs4) ? to_ucs4(range) : to_utf8(range)
raise "Invalid encoding of range #{range}: #{codes.inspect}" unless
is_valid? range, desc, codes
range_width = codes.map { |a| a.size }.max
range_width = RANGE_WIDTH if range_width < RANGE_WIDTH
desc_width = TOTAL_WIDTH - RANGE_WIDTH - 11
desc_width -= (range_width - RANGE_WIDTH) if range_width > RANGE_WIDTH
if desc.size > desc_width
desc = desc[0..desc_width - 4] + "..."
end
codes.each_with_index do |r, idx|
desc = "" unless idx.zero?
code = "%-#{range_width}s" % r
puts " #{pipe} #{code} ##{desc}"
pipe = "|"
end
end
puts " ;"
puts ""
end
puts <<EOF
# The following Ragel file was autogenerated from: #{CHART_URL}
#
# It defines ualpha, udigit, ualnum.
#
# To use this, make sure that your alphtype is set to #{ALPHTYPES[@encoding]},
# and that your input is in #{@encoding}.
%%{
machine WChar;
EOF
generate_machine(:uUppercaseLetter, "Lu")
generate_machine(:uLowercaseLetter, "Ll")
generate_machine(:uTitlecaseLetter, "Lt")
generate_machine(:uModifierLetter, "Lm")
generate_machine(:uOtherLetter, "Lo")
generate_machine(:uNonspacingMark, "Mn")
generate_machine(:uEnclosingMark, "Me")
generate_machine(:uSpacingMark, "Mc")
generate_machine(:uDecimalNumber, "Nd")
generate_machine(:uLetterNumber, "Nl")
generate_machine(:uOtherNumber, "No")
generate_machine(:uSpaceSeparator, "Zs")
generate_machine(:uLineSeparator, "Zl")
generate_machine(:uParagraphSeparator, "Zp")
generate_machine(:uFormat, "Cf")
generate_machine(:uPrivateUse, "Co")
generate_machine(:uSurrogate, "Cs")
generate_machine(:uDashPunctuation, "Pd")
generate_machine(:uOpenPunctuation, "Ps")
generate_machine(:uClosePunctuation, "Pe")
generate_machine(:uConnectorPunctuation, "Pc")
generate_machine(:uOtherPunctuation, "Po")
generate_machine(:uMathSymbol, "Sm")
generate_machine(:uCurrencySymbol, "Sc")
generate_machine(:uModifierSymbol, "Sk")
generate_machine(:uOtherSymbol, "So")
generate_machine(:uInitialPunctuation, "Pi")
generate_machine(:uFinalPunctuation, "Pf")
puts <<EOF
}%%
EOF
然後在你的ragel機文件可以包含unicode.rl並獲得各組的unicode訪問定義如uUppercaseLetter等等...
我不認爲有一種方法可以做到這一點以外,從列表編譯(長)的正則表達式所需的UTF-8編碼。例如,手動操作會很麻煩,但是用Python編寫它不會太難。但是,結果將與掃描儀生成時的Unicode數據庫相關聯,因此每次UCD更改時都需要重新生成掃描程序。 – rici 2015-03-08 18:53:21
嗯,這開始看起來像一個痛苦的任務。不要幻想從http://www.unicode.org/Public/UCD/latest/ucd/中的所有東西中生成一個大型文件,直到我嘗試過,但它聽起來非常低效。看看我能否得到我想要做的事[Ragel](http://www.colm。淨/文件/ ragel/ragel引導-6.9.pdf) – zcourts 2015-03-09 22:22:21