Don't build quoted_fields array when not needed (#312)

``` N_ROWS=5000 rake benchmark:write benchmark:parse benchmark:parse_liberal_parsing benchmark:parse_quote_char_nil benchmark:parse_strip ``` ``` RUBYLIB= BUNDLER_ORIG_RUBYLIB= /Users/vladimirkochnev/.asdf/installs/ruby/3.3.3/bin/ruby -v -S benchmark-driver /Users/vladimirkochnev/code/csv/benchmark/write.yaml ruby 3.3.3 (2024-06-12 revision f1c7b6f435) [arm64-darwin23] Calculating ------------------------------------- csv 3.3.0 master generate_line: fields 29.211 29.755 i/s - 100.000 times in 3.423319s 3.360759s generate_line: Row 28.090 28.121 i/s - 100.000 times in 3.560007s 3.556013s generate_line: Hash 26.398 26.888 i/s - 100.000 times in 3.788145s 3.719147s << fields 130.692 142.421 i/s - 100.000 times in 0.765156s 0.702142s << Row 103.416 107.886 i/s - 100.000 times in 0.966972s 0.926906s << Hash 109.760 114.806 i/s - 100.000 times in 0.911082s 0.871038s << fields: write headers 131.147 141.668 i/s - 100.000 times in 0.762501s 0.705878s << Row: write headers 102.956 108.919 i/s - 100.000 times in 0.971286s 0.918117s << Hash: write headers 109.498 115.403 i/s - 100.000 times in 0.913259s 0.866528s Comparison: generate_line: fields master: 29.8 i/s csv 3.3.0: 29.2 i/s - 1.02x slower generate_line: Row master: 28.1 i/s csv 3.3.0: 28.1 i/s - 1.00x slower generate_line: Hash master: 26.9 i/s csv 3.3.0: 26.4 i/s - 1.02x slower << fields master: 142.4 i/s csv 3.3.0: 130.7 i/s - 1.09x slower << Row master: 107.9 i/s csv 3.3.0: 103.4 i/s - 1.04x slower << Hash master: 114.8 i/s csv 3.3.0: 109.8 i/s - 1.05x slower << fields: write headers master: 141.7 i/s csv 3.3.0: 131.1 i/s - 1.08x slower << Row: write headers master: 108.9 i/s csv 3.3.0: 103.0 i/s - 1.06x slower << Hash: write headers master: 115.4 i/s csv 3.3.0: 109.5 i/s - 1.05x slower ``` ``` RUBYLIB= BUNDLER_ORIG_RUBYLIB= /Users/vladimirkochnev/.asdf/installs/ruby/3.3.3/bin/ruby -v -S benchmark-driver /Users/vladimirkochnev/code/csv/benchmark/parse.yaml ruby 3.3.3 (2024-06-12 revision f1c7b6f435) [arm64-darwin23] Calculating ------------------------------------- csv 3.3.0 master unquoted 21.798 22.176 i/s - 100.000 times in 4.587570s 4.509469s quoted 11.580 12.896 i/s - 100.000 times in 8.635392s 7.754641s mixed 14.082 14.139 i/s - 100.000 times in 7.101360s 7.072725s include_col_sep 5.206 5.191 i/s - 100.000 times in 19.209061s 19.265310s include_row_sep 5.125 5.179 i/s - 100.000 times in 19.513305s 19.307953s encode_utf-8 16.247 16.221 i/s - 100.000 times in 6.154900s 6.165029s encode_sjis 16.811 16.442 i/s - 100.000 times in 5.948591s 6.082152s Comparison: unquoted master: 22.2 i/s csv 3.3.0: 21.8 i/s - 1.02x slower quoted master: 12.9 i/s csv 3.3.0: 11.6 i/s - 1.11x slower mixed master: 14.1 i/s csv 3.3.0: 14.1 i/s - 1.00x slower include_col_sep csv 3.3.0: 5.2 i/s master: 5.2 i/s - 1.00x slower include_row_sep master: 5.2 i/s csv 3.3.0: 5.1 i/s - 1.01x slower encode_utf-8 csv 3.3.0: 16.2 i/s master: 16.2 i/s - 1.00x slower encode_sjis csv 3.3.0: 16.8 i/s master: 16.4 i/s - 1.02x slower ``` ``` RUBYLIB= BUNDLER_ORIG_RUBYLIB= /Users/vladimirkochnev/.asdf/installs/ruby/3.3.3/bin/ruby -v -S benchmark-driver /Users/vladimirkochnev/code/csv/benchmark/parse_liberal_parsing.yaml ruby 3.3.3 (2024-06-12 revision f1c7b6f435) [arm64-darwin23] Calculating ------------------------------------- csv 3.3.0 master unquoted 8.135 8.270 i/s - 100.000 times in 12.291808s 12.091261s unquoted_backslash_quote 3.865 3.854 i/s - 100.000 times in 25.872675s 25.946134s quoted 3.627 3.598 i/s - 100.000 times in 27.572211s 27.789378s quoted_double_quote_outside_quote 2.260 2.216 i/s - 100.000 times in 44.241118s 45.117111s quoted_backslash_quote 1.795 1.789 i/s - 100.000 times in 55.721082s 55.903782s include_col_sep 3.622 3.615 i/s - 100.000 times in 27.606966s 27.664617s include_row_sep 3.575 3.611 i/s - 100.000 times in 27.970871s 27.694692s encode_utf-8 8.041 8.175 i/s - 100.000 times in 12.436682s 12.232314s encode_sjis 8.515 8.147 i/s - 100.000 times in 11.744171s 12.274468s Comparison: unquoted master: 8.3 i/s csv 3.3.0: 8.1 i/s - 1.02x slower unquoted_backslash_quote csv 3.3.0: 3.9 i/s master: 3.9 i/s - 1.00x slower quoted csv 3.3.0: 3.6 i/s master: 3.6 i/s - 1.01x slower quoted_double_quote_outside_quote csv 3.3.0: 2.3 i/s master: 2.2 i/s - 1.02x slower quoted_backslash_quote csv 3.3.0: 1.8 i/s master: 1.8 i/s - 1.00x slower include_col_sep csv 3.3.0: 3.6 i/s master: 3.6 i/s - 1.00x slower include_row_sep master: 3.6 i/s csv 3.3.0: 3.6 i/s - 1.01x slower encode_utf-8 master: 8.2 i/s csv 3.3.0: 8.0 i/s - 1.02x slower encode_sjis csv 3.3.0: 8.5 i/s master: 8.1 i/s - 1.05x slower ``` ``` RUBYLIB= BUNDLER_ORIG_RUBYLIB= /Users/vladimirkochnev/.asdf/installs/ruby/3.3.3/bin/ruby -v -S benchmark-driver /Users/vladimirkochnev/code/csv/benchmark/parse_quote_char_nil.yaml ruby 3.3.3 (2024-06-12 revision f1c7b6f435) [arm64-darwin23] Calculating ------------------------------------- csv 3.3.0 master without_quote_char 22.806 22.552 i/s - 100.000 times in 4.384858s 4.434139s quote_char_nil 32.576 44.911 i/s - 100.000 times in 3.069777s 2.226621s col_sep_space 12.182 12.341 i/s - 100.000 times in 8.208668s 8.102909s Comparison: without_quote_char csv 3.3.0: 22.8 i/s master: 22.6 i/s - 1.01x slower quote_char_nil master: 44.9 i/s csv 3.3.0: 32.6 i/s - 1.38x slower col_sep_space master: 12.3 i/s csv 3.3.0: 12.2 i/s - 1.01x slower ``` ``` RUBYLIB= BUNDLER_ORIG_RUBYLIB= /Users/vladimirkochnev/.asdf/installs/ruby/3.3.3/bin/ruby -v -S benchmark-driver /Users/vladimirkochnev/code/csv/benchmark/parse_strip.yaml ruby 3.3.3 (2024-06-12 revision f1c7b6f435) [arm64-darwin23] Calculating ------------------------------------- csv 3.3.0 master default 13.025 13.075 i/s - 100.000 times in 7.677346s 7.648185s no_quote_strip 8.823 8.866 i/s - 100.000 times in 11.333992s 11.279182s Comparison: default master: 13.1 i/s csv 3.3.0: 13.0 i/s - 1.00x slower no_quote_strip master: 8.9 i/s csv 3.3.0: 8.8 i/s - 1.00x slower ```
ruby · Jul 21, 2024 · bb93c28 · bb93c28
1 parent e75132e
commit bb93c28
Show file tree

Hide file tree

Showing 3 changed files with 15 additions and 11 deletions.
diff --git a/lib/csv/fields_converter.rb b/lib/csv/fields_converter.rb
@@ -4,6 +4,13 @@ class CSV
   # Note: Don't use this class directly. This is an internal class.
   class FieldsConverter
     include Enumerable
+
+    NO_QUOTED_FIELDS = [] # :nodoc:
+    def NO_QUOTED_FIELDS.[](_index)
+      false
+    end
+    NO_QUOTED_FIELDS.freeze
+
     #
     # A CSV::FieldsConverter is a data structure for storing the
     # fields converter properties to be passed as a parameter
@@ -44,7 +51,7 @@ def empty?
       @converters.empty?
     end
 
-    def convert(fields, headers, lineno, quoted_fields)
+    def convert(fields, headers, lineno, quoted_fields=NO_QUOTED_FIELDS)
       return fields unless need_convert?
 
       fields.collect.with_index do |field, index|

diff --git a/lib/csv/parser.rb b/lib/csv/parser.rb
@@ -757,7 +757,7 @@ def prepare_header
       case headers
       when Array
         @raw_headers = headers
-        quoted_fields = [false] * @raw_headers.size
+        quoted_fields = FieldsConverter::NO_QUOTED_FIELDS
         @use_headers = true
       when String
         @raw_headers, quoted_fields = parse_headers(headers)
@@ -941,11 +941,9 @@ def parse_no_quote(&block)
         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)
@@ -959,7 +957,7 @@ def parse_no_quote(&block)
           end
         end
         @last_line = original_line
-        emit_row(row, quoted_fields, &block)
+        emit_row(row, &block)
       end
     end
 
@@ -981,7 +979,7 @@ def parse_quotable_loose(&block)
             next
           end
           row = []
-          quoted_fields = []
+          quoted_fields = FieldsConverter::NO_QUOTED_FIELDS
         elsif line.include?(@cr) or line.include?(@lf)
           @scanner.keep_back
           @parse_method = :parse_quotable_robust
@@ -1043,13 +1041,13 @@ def parse_quotable_robust(&block)
           quoted_fields << @quoted_column_value
         elsif parse_row_end
           if row.empty? and value.nil?
-            emit_row([], [], &block) unless @skip_blanks
+            emit_row(row, &block) unless @skip_blanks
           else
             row << value
             quoted_fields << @quoted_column_value
             emit_row(row, quoted_fields, &block)
             row = []
-            quoted_fields = []
+            quoted_fields.clear
           end
           skip_needless_lines
           start_row
@@ -1254,7 +1252,7 @@ def start_row
       @scanner.keep_start
     end
 
-    def emit_row(row, quoted_fields, &block)
+    def emit_row(row, quoted_fields=FieldsConverter::NO_QUOTED_FIELDS, &block)
       @lineno += 1
 
       raw_row = row

diff --git a/lib/csv/writer.rb b/lib/csv/writer.rb
@@ -40,8 +40,7 @@ def <<(row)
       @lineno += 1
 
       if @fields_converter
-        quoted_fields = [false] * row.size
-        row = @fields_converter.convert(row, nil, lineno, quoted_fields)
+        row = @fields_converter.convert(row, nil, lineno)
       end
 
       i = -1