Matching quoted strings in Ruby
An exercise in string processing and regexp matching, inspired by Parsing Quoted Strings in Ruby and Stupid Ruby Quoting Tricks.
#!/usr/local/bin/ruby -w # some input examples str = 'foo "bar baz" qux' str = 'foo "bar baz " "bar baz" " bar baz" "bar "klr mre" " " \' "abc" \' baz " qux' str = '" \' \' " \n " " \' \' "" foo \'ttt sss\' "bar "qqq zzz" baz" "added term" qux " \' \' " yyy xxx' str = '"""frickin \'#{bar}\'"""' str = '"" "frickin chicken " #{bar}""""' str = '"""frickin "#{bar}""""' str = '"a "b c" "d "e" f g" """h""""' # cf. http://snippets.dzone.com/posts/show/4852 # escaped quotes str = '\"' str = "\\\"" str = '\\\'' str = "\\'" # special cases str = '"G","H I"' str = '"G","H I""G","H I"' str = '"abc""def"' str = '"""a""b"' str = '"abc""def""abc""def""abc""def"' str = '"a"\'\'"b"' str = "\"abc'vv'tt\"'klt'" str = "abc,def,\"efg,hij\",klm,nop,\"qrstuv\",wxyz" str = "abc,def,\"efg,hij\",klm, 'nop, \"qrstuv\",wxyz,mmm '" str = "abc,def,\"efg,hij\",klm, \"nop, 'qrstuv',wxyz,mmm \"" puts puts "input string: #{str}" puts "str.inspect : #{str.inspect}" puts num_of_chars1 = str.count('a-zA-Z_0-9', "^\000ds") error_code = 0 # in case of a parsing error Shellwords will be used instead of regex1 & regex2 str2 = str.clone # encode escaped quotes str = str.gsub(/\\"|\\'/) { |m| m =~ /^\\"$/ ? "\000d\000" : "\000s\000" } dq_count = str.count('"') sq_count = str.count("'") if dq_count % 2 != 0 && sq_count % 2 != 0 raise ArgumentError, "\e[1modd number of single & double quotes\e[m in: #{str}\nsq_count: #{sq_count}\ndq_count: #{dq_count}\n" elsif dq_count % 2 != 0 raise ArgumentError, "\e[1modd number of double quotes\e[m in: #{str}\ndq_count: #{dq_count}\n" elsif sq_count % 2 != 0 raise ArgumentError, "\e[1modd number of single quotes\e[m in: #{str}\nsq_count: #{sq_count}\n" end # regex1 separates substrings that contain quotes from substrings that do not contain quotes regex1 = %r{[^"']+|["'].*?["'](?!.*["'])}m # example #"abc 'quote1' pjk 'quote2' xyz".scan(regex1) { |m| puts m } regex2 = %r{ # experimental: special cases \s*["'][^"']+["'][[:punct:]]["'][^"']+["']| # special case: xxx "ab c","def g" yyy \s*["'][^"']+["']{2,}[^"']+["']| # special case: xxx "abc""def" yyy \s"[^"]+"| # special case: xxx "abc 'def' ghi" \s'[^']+'| # special case: xxx 'abc "def" ghi' \s*["']\S+["']| # special case: "abc'vv'tt"'klt' \s'\s| # xxx ' yyy \s"\s| # xxx " yyy \s''\s| # xxx '' yyy \s""\s| # xxx "" yyy \s'\s+'\s| # xxx ' ' yyy \s"\s+"\s| # xxx " " yyy \s"\s[^"]+\s"\s| # xxx " abc " yyy \s'\s[^']+\s'\s| # xxx ' abc ' yyy \s["']["']+(?=[^"'\s])| # :qoblock: xxx "'""'abc yyy [^"'\s]["']["']+(?=\s)| # :qcblock: xxx abc"'""' yyy \s""+| # :dqoblock: xxx """abc yyy \s''+| # :sqoblock: xxx '''abc yyy [^"]""+| # :dqcblock: xxx abc"" yyy [^']''+| # :sqcblock: xxx abc'' yyy \s["'](?=[^"'\s])| # :dqo or :sqo: xxx "abc yyy or xxx 'abc yyy [^"'\s]["'](?=\s)| # :dqc or :sqc: xxx abc" yyy or xxx abc' yyy [^"']+[^"'\s](?=\s) # no quotes at all }mx =begin There are different kinds of quotes matched by regex2 below. They include: - :sqo (single quote open) - :sqc (single quote close) - :sqoblock (single quote open block) - :sqcblock (single quote close block) - :dqo (double quote open) - :dqc (double quote close) - :dqoblock (double quote open block) - :dqcblock (double quote close block) - :qoblock (quote open block) - :qcblock (quote close block) =end ret = [] str.scan(regex1) do |s| if s !~ /\A["']/ #puts "s1: #{s}" #puts "s1.inspect: #{s.inspect}" s.split(/\s+/m).each { |t| ret << t unless t.empty? } else #puts "s2: #{s}" #puts "s2.inspect: #{s.inspect}" open_quotes = 0 close_quotes = 0 ar = [] # add spaces to simplify regex2 matching s = "\x20" << s << "\x20" s.gsub!(/\x20/, "\x20\x20") s.scan(regex2) do |m| # get the index of the quote # + 1 for leading space or non-space # $` is the prematch string index = $`.length + 1 post_match = $' #puts #puts "index: #{index}" #puts "m: #{m.inspect}" #puts "m.length: #{m.length}" #puts "open_quotes: #{open_quotes}\nclose_quotes: #{close_quotes}" #puts "ret: #{ret.inspect}" #puts "ar: #{ar.inspect}" #puts if m =~ /\A\s''\s\z/ next unless open_quotes == 0 && close_quotes == 0 ret << '' next elsif m =~ /\A\s""\s\z/ next unless open_quotes == 0 && close_quotes == 0 ret << "" next # example: xxx "ab c","def g" yyy elsif open_quotes.zero? && close_quotes.zero? && m =~ /\A\s*["'][^"']+["'][[:punct:]]["'][^"']+["']\z/ && m.count('"') % 2 == 0 && m.count("'") % 2 == 0 m = m.gsub(/\x20\x20/, "\x20") # cf. http://henrik.nyh.se/2008/03/flickr-style-tag-splitting-in-ruby m = m.split(/"(.+?)"|\s+/).reject {|s| s.empty? } #m = m.split(/"(.+?)"|'(.+?)'|\s+/).reject {|s| s.empty? } #m = m.split(/"(.+?)"|'(.+?)'|([[:punct:]])|\s+/).reject {|s| s.empty? } ret.concat(m) next # example: xxx "abc""def" yyy elsif open_quotes.zero? && close_quotes.zero? && m =~ /\A\s*["'][^"']+["']{2,}[^"']+["']\z/ && m.count('"') % 2 == 0 && m.count("'") % 2 == 0 m = m.gsub(/\x20\x20/, "\x20") m = m.split(/"(.+?)"|\s+/).reject {|s| s.empty? } #m = m.split(/"(.+?)"|'(.+?)'|\s+/).reject {|s| s.empty? } ret.concat(m) next elsif open_quotes.zero? && close_quotes.zero? && m =~ /\A\s"[^"]+"\z/ && m.count('"') % 2 == 0 && m.count("'") % 2 == 0 ret.concat(m.split(/"(.+?)"|\s+/).reject {|sm| sm.empty? }) next elsif open_quotes.zero? && close_quotes.zero? && m =~ /\A\s'[^']+'\z/ && m.count('"') % 2 == 0 && m.count("'") % 2 == 0 ret.concat(m.split(/'(.+?)'|\s+/).reject {|sm| sm.empty? }) next elsif open_quotes.zero? && close_quotes.zero? && m =~ /\A\s*["']\S+["']\z/ && m.count('"') % 2 == 0 && m.count("'") % 2 == 0 ret.concat(m.split(/"(.+?)"|\s+/).reject {|s| s.empty? }) next elsif m =~ /\A\s"\s[^"]+\s"\s\z/ next unless open_quotes == 0 && close_quotes == 0 ret << m.gsub(/\x20\x20/, "\x20").strip[1..-2] next elsif m =~ /\A\s'\s[^']+\s"\s\z/ next unless open_quotes == 0 && close_quotes == 0 ret << m.gsub(/\x20\x20/, "\x20").strip[1..-2] next elsif m =~ /\A\s'\s+'\s\z/ next unless open_quotes == 0 && close_quotes == 0 ret << m.gsub(/\x20\x20/, "\x20").strip[1..-2] next elsif m =~ /\A\s"\s+"\s\z/ next unless open_quotes == 0 && close_quotes == 0 ret << m.gsub(/\x20\x20/, "\x20").strip[1..-2] next elsif m =~ /\A\s""+\z/ l = m.strip.length ar << [:dqoblock, index, l] old_open_quotes = open_quotes open_quotes += l if close_quotes == 0 && old_open_quotes == 0 && open_quotes % 2 == 0 && post_match !~ /"/ ret << m[2..-2] open_quotes = 0 ar.pop next end elsif m =~ /\A\s''+\z/ l = m.strip.length ar << [:sqoblock, index, l] old_open_quotes = open_quotes open_quotes += l if close_quotes == 0 && old_open_quotes == 0 && open_quotes % 2 == 0 && post_match !~ /'/ ret << m[2..-2] open_quotes = 0 ar.pop next end elsif m =~ /\A[^"]""+\z/ l = m[1..-1].strip.length ar << [:dqcblock, index+l-1, l] # index+l-1 is the index of the last closing quote: ''"'['] old_close_quotes = close_quotes close_quotes += l if open_quotes == 0 && old_close_quotes == 0 && close_quotes % 2 == 0 && post_match !~ /"/ ret << m[2..-2] close_quotes = 0 ar.pop next end elsif m =~ /\A[^']''+\z/ l = m[1..-1].strip.length ar << [:sqcblock, index+l-1, l] old_close_quotes = close_quotes close_quotes += l if open_quotes == 0 && old_close_quotes == 0 && close_quotes % 2 == 0 && post_match !~ /'/ ret << m[2..-2] close_quotes = 0 ar.pop next end elsif m =~ /\A\s'\z/ ar << [:sqo, index, 1] open_quotes += 1 elsif m =~ /\A\S'\z/ ar << [:sqc, index, 1] close_quotes += 1 elsif m =~ /\A\s"\z/ ar << [:dqo, index, 1] open_quotes += 1 elsif m =~ /\A\S"\z/ ar << [:dqc, index, 1] close_quotes += 1 else if m =~ /\A\s"\s\z/ # " surrounded by whitespace if open_quotes > close_quotes ar << [:dqc, index, 1] close_quotes += 1 # avoid :sqo followed by :dqc or :sqc followed by :dqc if post_match =~ /"/ && open_quotes == close_quotes && (ar.at(-2).first == :sqo || ar.at(-2).first == :sqc) ar.pop ar << [:dqo, index, 1] close_quotes -= 1 open_quotes += 1 end else ar << [:dqo, index, 1] open_quotes += 1 end elsif m =~ /\A\s'\s\z/ # ' surrounded by whitespace if open_quotes > close_quotes ar << [:sqc, index, 1] close_quotes += 1 # avoid :dqo followed by :sqc or :dqc followed by :sqc if post_match =~ /'/ && open_quotes == close_quotes && (ar.at(-2).first == :dqo || ar.at(-2).first == :dqc) ar.pop ar << [:sqo, index, 1] close_quotes -= 1 open_quotes += 1 end else ar << [:sqo, index, 1] open_quotes += 1 end elsif m =~ /\A\s["']["']+\z/ # :qoblock: xxx "'""'abc yyy l = m[1..-1].strip.length ar << [:qoblock, index, l] old_open_quotes = open_quotes open_quotes += l if close_quotes == 0 && old_open_quotes == 0 && open_quotes % 2 == 0 && post_match !~ /["']/ ret << m[2..-2] open_quotes = 0 ar.pop next end elsif m =~ /\A[^"'\s]["']["']+\z/ # :qcblock: xxx abc"'""' yyy l = m[1..-1].strip.length ar << [:qcblock, index+l-1, l] old_close_quotes = close_quotes close_quotes += l if open_quotes == 0 && old_close_quotes == 0 && close_quotes % 2 == 0 && post_match !~ /["']/ ret << m[2..-2] close_quotes = 0 ar.pop next end elsif m =~ /\A\s*["'].*?["']\s*\z/ # last try ret.concat(m.split(/"(.+?)"|\s+/).reject {|s| s.empty? }) next elsif m =~ /\A[^"']+[^"'\s]\z/ # part of quoted substring contains neither " nor ' next unless open_quotes == 0 && close_quotes == 0 next if m.strip.empty? ret << m.gsub(/\x20\x20/, "\x20").strip; next end end puts puts "open_quotes: #{open_quotes}\nclose_quotes: #{close_quotes}\n" #puts "ar: #{ar.inspect}" if open_quotes == close_quotes #puts "open_quotes & close_quotes: #{close_quotes}" puts "ar: #{ar.inspect}" ret << s[ar.first[1]..ar.last[1]].gsub(/\x20\x20/, "\x20")[1..-2] unless ar.empty? ar.clear open_quotes = 0 close_quotes = 0 end end # scan 2 unless open_quotes.zero? && close_quotes.zero? error_code = 1 puts "\e[1mparsing error\e[m for the quoted string: #{str.strip.squeeze[0..20]}" #raise "\e[1mparsing error\e[m for the quoted string: #{str.strip.squeeze[0..20]}" end end # if end # scan 1 num_of_chars2 = ret.join.count('a-zA-Z_0-9', "^\000ds") unless num_of_chars1 == num_of_chars2 error_code = 1 puts "\n\e[1mparsing error due to wrong number of characters a-zA-Z_0-9\e[m: \n#{num_of_chars2} instead of #{num_of_chars1}\n" #raise "\e[1mparsing error due to wrong number of characters a-zA-Z_0-9\e[m: \n#{num_of_chars2} instead of #{num_of_chars1}\n in #{str.strip.squeeze[0..20]}" end # use Shellwords in case the quote matching above failed if error_code == 1 #if error_code == 1 || ret.join =~ /\A["']+\z/ require 'shellwords' ret.clear ret.concat(Shellwords::shellwords(str)) #str =~ /\A\S+\z/ ? ret.concat(str.split(/"(.+?)"|\s+/).reject {|s| s.empty? }) : ret.concat(Shellwords::shellwords(str)) end puts "\n\e[1mResult\e[m:\n\n" ret.each_with_index do |t,i| # decode encoded escaped quotes t = t.gsub(/\000d\000|\000s\000/) { |m| m =~ /^\000d\000$/ ? '\"' : "\\'" } puts "#{i+1}: #{t.inspect}" end puts "\n\e[1mShellwords\e[m:\n\n" require 'shellwords' Shellwords::shellwords(str2).each_with_index { |t,i| puts "#{i+1}: #{t.inspect}" } #---------------------- # matching quoted strings using backreferences # See: Regexes in Depth: Advanced Quoted String Matching, # http://blog.stevenlevithan.com/archives/match-quoted-string str = '"abc"' regex = %r{(["'])([^"']*)(\1)} regex = %r{(["'])([^\1]*)(\1)} p regex str.scan(regex) { |m| p m; p $1 << $2 << $3 }