Server IP : 66.29.132.122 / Your IP : 3.149.233.221 Web Server : LiteSpeed System : Linux business142.web-hosting.com 4.18.0-553.lve.el8.x86_64 #1 SMP Mon May 27 15:27:34 UTC 2024 x86_64 User : admazpex ( 531) PHP Version : 7.2.34 Disable Function : NONE MySQL : OFF | cURL : ON | WGET : ON | Perl : ON | Python : ON | Sudo : OFF | Pkexec : OFF Directory : /proc/self/root/proc/self/root/proc/thread-self/root/proc/thread-self/root/proc/self/root/proc/thread-self/root/proc/self/root/proc/self/root/proc/self/root/proc/self/root/opt/alt/ruby33/share/ruby/csv/ |
Upload File : |
# frozen_string_literal: true require "strscan" require_relative "input_record_separator" require_relative "row" require_relative "table" class CSV # Note: Don't use this class directly. This is an internal class. class Parser # # A CSV::Parser is m17n aware. The parser works in the Encoding of the IO # or String object being read from or written to. Your data is never transcoded # (unless you ask Ruby to transcode it for you) and will literally be parsed in # the Encoding it is in. Thus CSV will return Arrays or Rows of Strings in the # Encoding of your data. This is accomplished by transcoding the parser itself # into your Encoding. # # Raised when encoding is invalid. class InvalidEncoding < StandardError end # Raised when unexpected case is happen. class UnexpectedError < StandardError end # # CSV::Scanner receives a CSV output, scans it and return the content. # It also controls the life cycle of the object with its methods +keep_start+, # +keep_end+, +keep_back+, +keep_drop+. # # Uses StringScanner (the official strscan gem). Strscan provides lexical # scanning operations on a String. We inherit its object and take advantage # on the methods. For more information, please visit: # https://ruby-doc.org/stdlib-2.6.1/libdoc/strscan/rdoc/StringScanner.html # class Scanner < StringScanner alias_method :scan_all, :scan def initialize(*args) super @keeps = [] end def each_line(row_separator) position = pos rest.each_line(row_separator) do |line| position += line.bytesize self.pos = position yield(line) end end def keep_start @keeps.push(pos) end def keep_end start = @keeps.pop string.byteslice(start, pos - start) end def keep_back self.pos = @keeps.pop end def keep_drop @keeps.pop end end # # CSV::InputsScanner receives IO inputs, encoding and the chunk_size. # It also controls the life cycle of the object with its methods +keep_start+, # +keep_end+, +keep_back+, +keep_drop+. # # CSV::InputsScanner.scan() tries to match with pattern at the current position. # If there's a match, the scanner advances the "scan pointer" and returns the matched string. # Otherwise, the scanner returns nil. # # CSV::InputsScanner.rest() returns the "rest" of the string (i.e. everything after the scan pointer). # If there is no more data (eos? = true), it returns "". # class InputsScanner def initialize(inputs, encoding, row_separator, chunk_size: 8192) @inputs = inputs.dup @encoding = encoding @row_separator = row_separator @chunk_size = chunk_size @last_scanner = @inputs.empty? @keeps = [] read_chunk end def each_line(row_separator) return enum_for(__method__, row_separator) unless block_given? buffer = nil input = @scanner.rest position = @scanner.pos offset = 0 n_row_separator_chars = row_separator.size # trace(__method__, :start, input) while true input.each_line(row_separator) do |line| @scanner.pos += line.bytesize if buffer if n_row_separator_chars == 2 and buffer.end_with?(row_separator[0]) and line.start_with?(row_separator[1]) buffer << line[0] line = line[1..-1] position += buffer.bytesize + offset @scanner.pos = position offset = 0 yield(buffer) buffer = nil next if line.empty? else buffer << line line = buffer buffer = nil end end if line.end_with?(row_separator) position += line.bytesize + offset @scanner.pos = position offset = 0 yield(line) else buffer = line end end break unless read_chunk input = @scanner.rest position = @scanner.pos offset = -buffer.bytesize if buffer end yield(buffer) if buffer end def scan(pattern) # trace(__method__, pattern, :start) value = @scanner.scan(pattern) # trace(__method__, pattern, :done, :last, value) if @last_scanner return value if @last_scanner read_chunk if value and @scanner.eos? # trace(__method__, pattern, :done, value) value end def scan_all(pattern) # trace(__method__, pattern, :start) value = @scanner.scan(pattern) # trace(__method__, pattern, :done, :last, value) if @last_scanner return value if @last_scanner # trace(__method__, pattern, :done, :nil) if value.nil? return nil if value.nil? while @scanner.eos? and read_chunk and (sub_value = @scanner.scan(pattern)) # trace(__method__, pattern, :sub, sub_value) value << sub_value end # trace(__method__, pattern, :done, value) value end def eos? @scanner.eos? end def keep_start # trace(__method__, :start) adjust_last_keep @keeps.push([@scanner, @scanner.pos, nil]) # trace(__method__, :done) end def keep_end # trace(__method__, :start) scanner, start, buffer = @keeps.pop if scanner == @scanner keep = @scanner.string.byteslice(start, @scanner.pos - start) else keep = @scanner.string.byteslice(0, @scanner.pos) end if buffer buffer << keep keep = buffer end # trace(__method__, :done, keep) keep end def keep_back # trace(__method__, :start) scanner, start, buffer = @keeps.pop if buffer # trace(__method__, :rescan, start, buffer) string = @scanner.string if scanner == @scanner keep = string.byteslice(start, string.bytesize - @scanner.pos - start) else keep = string end if keep and not keep.empty? @inputs.unshift(StringIO.new(keep)) @last_scanner = false end @scanner = StringScanner.new(buffer) else if @scanner != scanner message = "scanners are different but no buffer: " message += "#{@scanner.inspect}(#{@scanner.object_id}): " message += "#{scanner.inspect}(#{scanner.object_id})" raise UnexpectedError, message end # trace(__method__, :repos, start, buffer) @scanner.pos = start end read_chunk if @scanner.eos? end def keep_drop _, _, buffer = @keeps.pop # trace(__method__, :done, :empty) unless buffer return unless buffer last_keep = @keeps.last # trace(__method__, :done, :no_last_keep) unless last_keep return unless last_keep if last_keep[2] last_keep[2] << buffer else last_keep[2] = buffer end # trace(__method__, :done) end def rest @scanner.rest end def check(pattern) @scanner.check(pattern) end private def trace(*args) pp([*args, @scanner, @scanner&.string, @scanner&.pos, @keeps]) end def adjust_last_keep # trace(__method__, :start) keep = @keeps.last # trace(__method__, :done, :empty) if keep.nil? return if keep.nil? scanner, start, buffer = keep string = @scanner.string if @scanner != scanner start = 0 end if start == 0 and @scanner.eos? keep_data = string else keep_data = string.byteslice(start, @scanner.pos - start) end if keep_data if buffer buffer << keep_data else keep[2] = keep_data.dup end end # trace(__method__, :done) end def read_chunk return false if @last_scanner adjust_last_keep input = @inputs.first case input when StringIO string = input.read raise InvalidEncoding unless string.valid_encoding? # trace(__method__, :stringio, string) @scanner = StringScanner.new(string) @inputs.shift @last_scanner = @inputs.empty? true else chunk = input.gets(@row_separator, @chunk_size) if chunk raise InvalidEncoding unless chunk.valid_encoding? # trace(__method__, :chunk, chunk) @scanner = StringScanner.new(chunk) if input.respond_to?(:eof?) and input.eof? @inputs.shift @last_scanner = @inputs.empty? end true else # trace(__method__, :no_chunk) @scanner = StringScanner.new("".encode(@encoding)) @inputs.shift @last_scanner = @inputs.empty? if @last_scanner false else read_chunk end end end end end def initialize(input, options) @input = input @options = options @samples = [] prepare end def column_separator @column_separator end def row_separator @row_separator end def quote_character @quote_character end def field_size_limit @max_field_size&.succ end def max_field_size @max_field_size end def skip_lines @skip_lines end def unconverted_fields? @unconverted_fields end def headers @headers end def header_row? @use_headers and @headers.nil? end def return_headers? @return_headers end def skip_blanks? @skip_blanks end def liberal_parsing? @liberal_parsing end def lineno @lineno end def line last_line end def parse(&block) return to_enum(__method__) unless block_given? if @return_headers and @headers and @raw_headers headers = Row.new(@headers, @raw_headers, true) if @unconverted_fields headers = add_unconverted_fields(headers, []) end yield headers end begin @scanner ||= build_scanner if quote_character.nil? parse_no_quote(&block) elsif @need_robust_parsing parse_quotable_robust(&block) else parse_quotable_loose(&block) end rescue InvalidEncoding if @scanner ignore_broken_line lineno = @lineno else lineno = @lineno + 1 end raise InvalidEncodingError.new(@encoding, lineno) rescue UnexpectedError => error if @scanner ignore_broken_line lineno = @lineno else lineno = @lineno + 1 end message = "This should not be happen: #{error.message}: " message += "Please report this to https://github.com/ruby/csv/issues" raise MalformedCSVError.new(message, lineno) end end def use_headers? @use_headers end private # A set of tasks to prepare the file in order to parse it def prepare prepare_variable prepare_quote_character prepare_backslash prepare_skip_lines prepare_strip prepare_separators validate_strip_and_col_sep_options prepare_quoted prepare_unquoted prepare_line prepare_header prepare_parser end def prepare_variable @need_robust_parsing = false @encoding = @options[:encoding] liberal_parsing = @options[:liberal_parsing] if liberal_parsing @liberal_parsing = true if liberal_parsing.is_a?(Hash) @double_quote_outside_quote = liberal_parsing[:double_quote_outside_quote] @backslash_quote = liberal_parsing[:backslash_quote] else @double_quote_outside_quote = false @backslash_quote = false end @need_robust_parsing = true else @liberal_parsing = false @backslash_quote = false end @unconverted_fields = @options[:unconverted_fields] @max_field_size = @options[:max_field_size] @skip_blanks = @options[:skip_blanks] @fields_converter = @options[:fields_converter] @header_fields_converter = @options[:header_fields_converter] end def prepare_quote_character @quote_character = @options[:quote_character] if @quote_character.nil? @escaped_quote_character = nil @escaped_quote = nil else @quote_character = @quote_character.to_s.encode(@encoding) if @quote_character.length != 1 message = ":quote_char has to be nil or a single character String" raise ArgumentError, message end @escaped_quote_character = Regexp.escape(@quote_character) @escaped_quote = Regexp.new(@escaped_quote_character) end end def prepare_backslash return unless @backslash_quote @backslash_character = "\\".encode(@encoding) @escaped_backslash_character = Regexp.escape(@backslash_character) @escaped_backslash = Regexp.new(@escaped_backslash_character) if @quote_character.nil? @backslash_quote_character = nil else @backslash_quote_character = @backslash_character + @escaped_quote_character end end def prepare_skip_lines skip_lines = @options[:skip_lines] case skip_lines when String @skip_lines = skip_lines.encode(@encoding) when Regexp, nil @skip_lines = skip_lines else unless skip_lines.respond_to?(:match) message = ":skip_lines has to respond to \#match: #{skip_lines.inspect}" raise ArgumentError, message end @skip_lines = skip_lines end end def prepare_strip @strip = @options[:strip] @escaped_strip = nil @strip_value = nil @rstrip_value = nil if @strip.is_a?(String) case @strip.length when 0 raise ArgumentError, ":strip must not be an empty String" when 1 # ok else raise ArgumentError, ":strip doesn't support 2 or more characters yet" end @strip = @strip.encode(@encoding) @escaped_strip = Regexp.escape(@strip) if @quote_character @strip_value = Regexp.new(@escaped_strip + "+".encode(@encoding)) @rstrip_value = Regexp.new(@escaped_strip + "+\\z".encode(@encoding)) end @need_robust_parsing = true elsif @strip strip_values = " \t\f\v" @escaped_strip = strip_values.encode(@encoding) if @quote_character @strip_value = Regexp.new("[#{strip_values}]+".encode(@encoding)) @rstrip_value = Regexp.new("[#{strip_values}]+\\z".encode(@encoding)) end @need_robust_parsing = true end end begin StringScanner.new("x").scan("x") rescue TypeError STRING_SCANNER_SCAN_ACCEPT_STRING = false else STRING_SCANNER_SCAN_ACCEPT_STRING = true end def prepare_separators column_separator = @options[:column_separator] @column_separator = column_separator.to_s.encode(@encoding) if @column_separator.size < 1 message = ":col_sep must be 1 or more characters: " message += column_separator.inspect raise ArgumentError, message end @row_separator = resolve_row_separator(@options[:row_separator]).encode(@encoding) @escaped_column_separator = Regexp.escape(@column_separator) @escaped_first_column_separator = Regexp.escape(@column_separator[0]) if @column_separator.size > 1 @column_end = Regexp.new(@escaped_column_separator) @column_ends = @column_separator.each_char.collect do |char| Regexp.new(Regexp.escape(char)) end @first_column_separators = Regexp.new(@escaped_first_column_separator + "+".encode(@encoding)) else if STRING_SCANNER_SCAN_ACCEPT_STRING @column_end = @column_separator else @column_end = Regexp.new(@escaped_column_separator) end @column_ends = nil @first_column_separators = nil end escaped_row_separator = Regexp.escape(@row_separator) @row_end = Regexp.new(escaped_row_separator) if @row_separator.size > 1 @row_ends = @row_separator.each_char.collect do |char| Regexp.new(Regexp.escape(char)) end else @row_ends = nil end @cr = "\r".encode(@encoding) @lf = "\n".encode(@encoding) @line_end = Regexp.new("\r\n|\n|\r".encode(@encoding)) @not_line_end = Regexp.new("[^\r\n]+".encode(@encoding)) end # This method verifies that there are no (obvious) ambiguities with the # provided +col_sep+ and +strip+ parsing options. For example, if +col_sep+ # and +strip+ were both equal to +\t+, then there would be no clear way to # parse the input. def validate_strip_and_col_sep_options return unless @strip if @strip.is_a?(String) if @column_separator.start_with?(@strip) || @column_separator.end_with?(@strip) raise ArgumentError, "The provided strip (#{@escaped_strip}) and " \ "col_sep (#{@escaped_column_separator}) options are incompatible." end else if Regexp.new("\\A[#{@escaped_strip}]|[#{@escaped_strip}]\\z").match?(@column_separator) raise ArgumentError, "The provided strip (true) and " \ "col_sep (#{@escaped_column_separator}) options are incompatible." end end end def prepare_quoted if @quote_character @quotes = Regexp.new(@escaped_quote_character + "+".encode(@encoding)) no_quoted_values = @escaped_quote_character.dup if @backslash_quote no_quoted_values << @escaped_backslash_character end @quoted_value = Regexp.new("[^".encode(@encoding) + no_quoted_values + "]+".encode(@encoding)) end if @escaped_strip @split_column_separator = Regexp.new(@escaped_strip + "*".encode(@encoding) + @escaped_column_separator + @escaped_strip + "*".encode(@encoding)) else if @column_separator == " ".encode(@encoding) @split_column_separator = Regexp.new(@escaped_column_separator) else @split_column_separator = @column_separator end end end def prepare_unquoted return if @quote_character.nil? no_unquoted_values = "\r\n".encode(@encoding) no_unquoted_values << @escaped_first_column_separator unless @liberal_parsing no_unquoted_values << @escaped_quote_character end @unquoted_value = Regexp.new("[^".encode(@encoding) + no_unquoted_values + "]+".encode(@encoding)) end def resolve_row_separator(separator) if separator == :auto cr = "\r".encode(@encoding) lf = "\n".encode(@encoding) if @input.is_a?(StringIO) pos = @input.pos separator = detect_row_separator(@input.read, cr, lf) @input.seek(pos) elsif @input.respond_to?(:gets) if @input.is_a?(File) chunk_size = 32 * 1024 else chunk_size = 1024 end begin while separator == :auto # # if we run out of data, it's probably a single line # (ensure will set default value) # break unless sample = @input.gets(nil, chunk_size) # extend sample if we're unsure of the line ending if sample.end_with?(cr) sample << (@input.gets(nil, 1) || "") end @samples << sample separator = detect_row_separator(sample, cr, lf) end rescue IOError # do nothing: ensure will set default end end separator = InputRecordSeparator.value if separator == :auto end separator.to_s.encode(@encoding) end def detect_row_separator(sample, cr, lf) lf_index = sample.index(lf) if lf_index cr_index = sample[0, lf_index].index(cr) else cr_index = sample.index(cr) end if cr_index and lf_index if cr_index + 1 == lf_index cr + lf elsif cr_index < lf_index cr else lf end elsif cr_index cr elsif lf_index lf else :auto end end def prepare_line @lineno = 0 @last_line = nil @scanner = nil end def last_line if @scanner @last_line ||= @scanner.keep_end else @last_line end end def prepare_header @return_headers = @options[:return_headers] headers = @options[:headers] case headers when Array @raw_headers = headers quoted_fields = [false] * @raw_headers.size @use_headers = true when String @raw_headers, quoted_fields = parse_headers(headers) @use_headers = true when nil, false @raw_headers = nil @use_headers = false else @raw_headers = nil @use_headers = true end if @raw_headers @headers = adjust_headers(@raw_headers, quoted_fields) else @headers = nil end end def parse_headers(row) quoted_fields = [] converter = lambda do |field, info| quoted_fields << info.quoted? field end headers = CSV.parse_line(row, col_sep: @column_separator, row_sep: @row_separator, quote_char: @quote_character, converters: [converter]) [headers, quoted_fields] end def adjust_headers(headers, quoted_fields) adjusted_headers = @header_fields_converter.convert(headers, nil, @lineno, quoted_fields) adjusted_headers.each {|h| h.freeze if h.is_a? String} adjusted_headers end def prepare_parser @may_quoted = may_quoted? end def may_quoted? return false if @quote_character.nil? if @input.is_a?(StringIO) pos = @input.pos sample = @input.read @input.seek(pos) else return false if @samples.empty? sample = @samples.first end sample[0, 128].index(@quote_character) end class UnoptimizedStringIO # :nodoc: def initialize(string) @io = StringIO.new(string, "rb:#{string.encoding}") end def gets(*args) @io.gets(*args) end def each_line(*args, &block) @io.each_line(*args, &block) end def eof? @io.eof? end end SCANNER_TEST = (ENV["CSV_PARSER_SCANNER_TEST"] == "yes") if SCANNER_TEST SCANNER_TEST_CHUNK_SIZE_NAME = "CSV_PARSER_SCANNER_TEST_CHUNK_SIZE" SCANNER_TEST_CHUNK_SIZE_VALUE = ENV[SCANNER_TEST_CHUNK_SIZE_NAME] def build_scanner inputs = @samples.collect do |sample| UnoptimizedStringIO.new(sample) end if @input.is_a?(StringIO) inputs << UnoptimizedStringIO.new(@input.read) else inputs << @input end begin chunk_size_value = ENV[SCANNER_TEST_CHUNK_SIZE_NAME] rescue # Ractor::IsolationError # Ractor on Ruby 3.0 can't read ENV value. chunk_size_value = SCANNER_TEST_CHUNK_SIZE_VALUE end chunk_size = Integer((chunk_size_value || "1"), 10) InputsScanner.new(inputs, @encoding, @row_separator, chunk_size: chunk_size) end else def build_scanner string = nil if @samples.empty? and @input.is_a?(StringIO) string = @input.read elsif @samples.size == 1 and @input != ARGF and @input.respond_to?(:eof?) and @input.eof? string = @samples[0] end if string unless string.valid_encoding? index = string.lines(@row_separator).index do |line| !line.valid_encoding? end if index raise InvalidEncodingError.new(@encoding, @lineno + index + 1) end end Scanner.new(string) else inputs = @samples.collect do |sample| StringIO.new(sample) end inputs << @input InputsScanner.new(inputs, @encoding, @row_separator) end end end def skip_needless_lines return unless @skip_lines until @scanner.eos? @scanner.keep_start line = @scanner.scan_all(@not_line_end) || "".encode(@encoding) line << @row_separator if parse_row_end if skip_line?(line) @lineno += 1 @scanner.keep_drop else @scanner.keep_back return end end end def skip_line?(line) line = line.delete_suffix(@row_separator) case @skip_lines when String line.include?(@skip_lines) when Regexp @skip_lines.match?(line) else @skip_lines.match(line) end end def validate_field_size(field) return unless @max_field_size return if field.size <= @max_field_size ignore_broken_line message = "Field size exceeded: #{field.size} > #{@max_field_size}" raise MalformedCSVError.new(message, @lineno) end def parse_no_quote(&block) @scanner.each_line(@row_separator) do |line| next if @skip_lines and skip_line?(line) original_line = line line = line.delete_suffix(@row_separator) if line.empty? next if @skip_blanks row = [] quoted_fields = [] else line = strip_value(line) row = line.split(@split_column_separator, -1) quoted_fields = [false] * row.size if @max_field_size row.each do |column| validate_field_size(column) end end n_columns = row.size i = 0 while i < n_columns row[i] = nil if row[i].empty? i += 1 end end @last_line = original_line emit_row(row, quoted_fields, &block) end end def parse_quotable_loose(&block) @scanner.keep_start @scanner.each_line(@row_separator) do |line| if @skip_lines and skip_line?(line) @scanner.keep_drop @scanner.keep_start next end original_line = line line = line.delete_suffix(@row_separator) if line.empty? if @skip_blanks @scanner.keep_drop @scanner.keep_start next end row = [] quoted_fields = [] elsif line.include?(@cr) or line.include?(@lf) @scanner.keep_back @need_robust_parsing = true return parse_quotable_robust(&block) else row = line.split(@split_column_separator, -1) quoted_fields = [] n_columns = row.size i = 0 while i < n_columns column = row[i] if column.empty? quoted_fields << false row[i] = nil else n_quotes = column.count(@quote_character) if n_quotes.zero? quoted_fields << false # no quote elsif n_quotes == 2 and column.start_with?(@quote_character) and column.end_with?(@quote_character) quoted_fields << true row[i] = column[1..-2] else @scanner.keep_back @need_robust_parsing = true return parse_quotable_robust(&block) end validate_field_size(row[i]) end i += 1 end end @scanner.keep_drop @scanner.keep_start @last_line = original_line emit_row(row, quoted_fields, &block) end @scanner.keep_drop end def parse_quotable_robust(&block) row = [] quoted_fields = [] skip_needless_lines start_row while true @quoted_column_value = false @unquoted_column_value = false @scanner.scan_all(@strip_value) if @strip_value value = parse_column_value if value @scanner.scan_all(@strip_value) if @strip_value validate_field_size(value) end if parse_column_end row << value quoted_fields << @quoted_column_value elsif parse_row_end if row.empty? and value.nil? emit_row([], [], &block) unless @skip_blanks else row << value quoted_fields << @quoted_column_value emit_row(row, quoted_fields, &block) row = [] quoted_fields = [] end skip_needless_lines start_row elsif @scanner.eos? break if row.empty? and value.nil? row << value quoted_fields << @quoted_column_value emit_row(row, quoted_fields, &block) break else if @quoted_column_value if liberal_parsing? and (new_line = @scanner.check(@line_end)) message = "Illegal end-of-line sequence outside of a quoted field " + "<#{new_line.inspect}>" else message = "Any value after quoted field isn't allowed" end ignore_broken_line raise MalformedCSVError.new(message, @lineno) elsif @unquoted_column_value and (new_line = @scanner.scan(@line_end)) ignore_broken_line message = "Unquoted fields do not allow new line " + "<#{new_line.inspect}>" raise MalformedCSVError.new(message, @lineno) elsif @scanner.rest.start_with?(@quote_character) ignore_broken_line message = "Illegal quoting" raise MalformedCSVError.new(message, @lineno) elsif (new_line = @scanner.scan(@line_end)) ignore_broken_line message = "New line must be <#{@row_separator.inspect}> " + "not <#{new_line.inspect}>" raise MalformedCSVError.new(message, @lineno) else ignore_broken_line raise MalformedCSVError.new("TODO: Meaningful message", @lineno) end end end end def parse_column_value if @liberal_parsing quoted_value = parse_quoted_column_value if quoted_value @scanner.scan_all(@strip_value) if @strip_value unquoted_value = parse_unquoted_column_value if unquoted_value if @double_quote_outside_quote unquoted_value = unquoted_value.gsub(@quote_character * 2, @quote_character) if quoted_value.empty? # %Q{""...} case return @quote_character + unquoted_value end end @quote_character + quoted_value + @quote_character + unquoted_value else quoted_value end else parse_unquoted_column_value end elsif @may_quoted parse_quoted_column_value || parse_unquoted_column_value else parse_unquoted_column_value || parse_quoted_column_value end end def parse_unquoted_column_value value = @scanner.scan_all(@unquoted_value) return nil unless value @unquoted_column_value = true if @first_column_separators while true @scanner.keep_start is_column_end = @column_ends.all? do |column_end| @scanner.scan(column_end) end @scanner.keep_back break if is_column_end sub_separator = @scanner.scan_all(@first_column_separators) break if sub_separator.nil? value << sub_separator sub_value = @scanner.scan_all(@unquoted_value) break if sub_value.nil? value << sub_value end end value.gsub!(@backslash_quote_character, @quote_character) if @backslash_quote if @rstrip_value value.gsub!(@rstrip_value, "") end value end def parse_quoted_column_value quotes = @scanner.scan_all(@quotes) return nil unless quotes @quoted_column_value = true n_quotes = quotes.size if (n_quotes % 2).zero? quotes[0, (n_quotes - 2) / 2] else value = quotes[0, n_quotes / 2] while true quoted_value = @scanner.scan_all(@quoted_value) value << quoted_value if quoted_value if @backslash_quote if @scanner.scan(@escaped_backslash) if @scanner.scan(@escaped_quote) value << @quote_character else value << @backslash_character end next end end quotes = @scanner.scan_all(@quotes) unless quotes ignore_broken_line message = "Unclosed quoted field" raise MalformedCSVError.new(message, @lineno) end n_quotes = quotes.size if n_quotes == 1 break else value << quotes[0, n_quotes / 2] break if (n_quotes % 2) == 1 end end value end end def parse_column_end return true if @scanner.scan(@column_end) return false unless @column_ends @scanner.keep_start if @column_ends.all? {|column_end| @scanner.scan(column_end)} @scanner.keep_drop true else @scanner.keep_back false end end def parse_row_end return true if @scanner.scan(@row_end) return false unless @row_ends @scanner.keep_start if @row_ends.all? {|row_end| @scanner.scan(row_end)} @scanner.keep_drop true else @scanner.keep_back false end end def strip_value(value) return value unless @strip return value if value.nil? case @strip when String while value.delete_prefix!(@strip) # do nothing end while value.delete_suffix!(@strip) # do nothing end else value.strip! end value end def ignore_broken_line @scanner.scan_all(@not_line_end) @scanner.scan_all(@line_end) @lineno += 1 end def start_row if @last_line @last_line = nil else @scanner.keep_drop end @scanner.keep_start end def emit_row(row, quoted_fields, &block) @lineno += 1 raw_row = row if @use_headers if @headers.nil? @headers = adjust_headers(row, quoted_fields) return unless @return_headers row = Row.new(@headers, row, true) else row = Row.new(@headers, @fields_converter.convert(raw_row, @headers, @lineno, quoted_fields)) end else # convert fields, if needed... row = @fields_converter.convert(raw_row, nil, @lineno, quoted_fields) end # inject unconverted fields and accessor, if requested... if @unconverted_fields and not row.respond_to?(:unconverted_fields) add_unconverted_fields(row, raw_row) end yield(row) end # This method injects an instance variable <tt>unconverted_fields</tt> into # +row+ and an accessor method for +row+ called unconverted_fields(). The # variable is set to the contents of +fields+. def add_unconverted_fields(row, fields) class << row attr_reader :unconverted_fields end row.instance_variable_set(:@unconverted_fields, fields) row end end end