csvops 0.3.0.alpha → 0.4.0.alpha

data/lib/csvtool/application/use_cases/run_extraction.rb

@@ -19,8 +19,8 @@ require "csvtool/domain/column_session/column_selection"
 require "csvtool/domain/column_session/extraction_options"
 require "csvtool/domain/column_session/extraction_value"
 require "csvtool/domain/column_session/preview"
-require "csvtool/domain/column_session/output_destination"
 require "csvtool/domain/column_session/column_session"
+require "csvtool/domain/shared/output_destination"
 
 module Csvtool
   module Application
@@ -79,9 +79,9 @@ module Csvtool
           return if output_destination.nil?
           domain_destination =
             if output_destination[:mode] == :file
-              Domain::ColumnSession::OutputDestination.file(path: output_destination[:path])
+              Domain::Shared::OutputDestination.file(path: output_destination[:path])
             else
-              Domain::ColumnSession::OutputDestination.console
+              Domain::Shared::OutputDestination.console
             end
           session = session.with_output_destination(domain_destination)
 

data/lib/csvtool/application/use_cases/run_row_extraction.rb

@@ -11,8 +11,8 @@ require "csvtool/infrastructure/output/csv_row_console_writer"
 require "csvtool/infrastructure/output/csv_row_file_writer"
 require "csvtool/domain/row_session/row_range"
 require "csvtool/domain/row_session/row_source"
-require "csvtool/domain/row_session/row_output_destination"
 require "csvtool/domain/row_session/row_session"
+require "csvtool/domain/shared/output_destination"
 
 module Csvtool
   module Application
@@ -56,9 +56,9 @@ module Csvtool
           return if output_destination.nil?
           destination =
             if output_destination[:mode] == :file
-              Domain::RowSession::RowOutputDestination.file(path: output_destination[:path])
+              Domain::Shared::OutputDestination.file(path: output_destination[:path])
             else
-              Domain::RowSession::RowOutputDestination.console
+              Domain::Shared::OutputDestination.console
             end
           session = session.with_output_destination(destination)
 

data/lib/csvtool/application/use_cases/run_row_randomization.rb

@@ -11,8 +11,8 @@ require "csvtool/infrastructure/csv/header_reader"
 require "csvtool/infrastructure/csv/row_randomizer"
 require "csvtool/domain/row_randomization_session/randomization_source"
 require "csvtool/domain/row_randomization_session/randomization_options"
-require "csvtool/domain/row_randomization_session/randomization_output_destination"
 require "csvtool/domain/row_randomization_session/randomization_session"
+require "csvtool/domain/shared/output_destination"
 
 module Csvtool
   module Application
@@ -55,9 +55,9 @@ module Csvtool
           return if output_destination.nil?
           destination =
             if output_destination[:mode] == :file
-              Domain::RowRandomizationSession::RandomizationOutputDestination.file(path: output_destination[:path])
+              Domain::Shared::OutputDestination.file(path: output_destination[:path])
             else
-              Domain::RowRandomizationSession::RandomizationOutputDestination.console
+              Domain::Shared::OutputDestination.console
             end
           session = session.with_output_destination(destination)
 

data/lib/csvtool/cli.rb

@@ -5,6 +5,7 @@ require "csvtool/interface/cli/menu_loop"
 require "csvtool/application/use_cases/run_extraction"
 require "csvtool/application/use_cases/run_row_extraction"
 require "csvtool/application/use_cases/run_row_randomization"
+require "csvtool/interface/cli/workflows/run_cross_csv_dedupe_workflow"
 require "csvtool/interface/cli/errors/presenter"
 require "csvtool/infrastructure/csv/header_reader"
 require "csvtool/infrastructure/csv/value_streamer"
@@ -16,6 +17,7 @@ module Csvtool
       "Extract column",
       "Extract rows (range)",
       "Randomize rows",
+      "Dedupe using another CSV",
       "Exit"
     ].freeze
 
@@ -48,13 +50,15 @@ module Csvtool
       extract_column_action = -> { Application::UseCases::RunExtraction.new(stdin: @stdin, stdout: @stdout).call }
       extract_rows_action = -> { Application::UseCases::RunRowExtraction.new(stdin: @stdin, stdout: @stdout).call }
       randomize_rows_action = -> { Application::UseCases::RunRowRandomization.new(stdin: @stdin, stdout: @stdout).call }
+      dedupe_action = -> { Interface::CLI::Workflows::RunCrossCsvDedupeWorkflow.new(stdin: @stdin, stdout: @stdout).call }
       Interface::CLI::MenuLoop.new(
         stdin: @stdin,
         stdout: @stdout,
         menu_options: MENU_OPTIONS,
         extract_column_action: extract_column_action,
         extract_rows_action: extract_rows_action,
-        randomize_rows_action: randomize_rows_action
+        randomize_rows_action: randomize_rows_action,
+        dedupe_action: dedupe_action
       ).run
     end
 

data/lib/csvtool/domain/cross_csv_dedupe_session/column_selector.rb

@@ -0,0 +1,44 @@
+# frozen_string_literal: true
+
+module Csvtool
+  module Domain
+    module CrossCsvDedupeSession
+      class ColumnSelector
+        attr_reader :value
+
+        def self.from_input(headers_present:, input:)
+          if headers_present
+            raise ArgumentError, "column name cannot be empty" if input.to_s.empty?
+
+            new(value: input.to_s, headers_present: true)
+          else
+            raise ArgumentError, "column index must be a positive integer" unless /\A[1-9]\d*\z/.match?(input.to_s)
+
+            new(value: input.to_i, headers_present: false)
+          end
+        end
+
+        def initialize(value:, headers_present:)
+          @value = value
+          @headers_present = !!headers_present
+        end
+
+        def headers_present?
+          @headers_present
+        end
+
+        def index?
+          !@headers_present
+        end
+
+        def extract_from(row)
+          if headers_present?
+            row[@value].to_s
+          else
+            row[@value - 1].to_s
+          end
+        end
+      end
+    end
+  end
+end

data/lib/csvtool/domain/cross_csv_dedupe_session/cross_csv_dedupe_session.rb

@@ -0,0 +1,46 @@
+# frozen_string_literal: true
+
+require "csvtool/domain/cross_csv_dedupe_session/csv_profile"
+require "csvtool/domain/cross_csv_dedupe_session/key_mapping"
+require "csvtool/domain/cross_csv_dedupe_session/match_options"
+require "csvtool/domain/shared/output_destination"
+
+module Csvtool
+  module Domain
+    module CrossCsvDedupeSession
+      class CrossCsvDedupeSession
+        attr_reader :source, :reference, :key_mapping, :match_options, :output_destination
+
+        def self.start(source:, reference:, key_mapping:, match_options:)
+          new(source: source, reference: reference, key_mapping: key_mapping, match_options: match_options)
+        end
+
+        def initialize(source:, reference:, key_mapping:, match_options:, output_destination: nil)
+          raise ArgumentError, "source must be CsvProfile" unless source.is_a?(CsvProfile)
+          raise ArgumentError, "reference must be CsvProfile" unless reference.is_a?(CsvProfile)
+          raise ArgumentError, "key_mapping must be KeyMapping" unless key_mapping.is_a?(KeyMapping)
+          raise ArgumentError, "match_options must be MatchOptions" unless match_options.is_a?(MatchOptions)
+          unless output_destination.nil? || output_destination.is_a?(Domain::Shared::OutputDestination)
+            raise ArgumentError, "output_destination must be OutputDestination or nil"
+          end
+
+          @source = source
+          @reference = reference
+          @key_mapping = key_mapping
+          @match_options = match_options
+          @output_destination = output_destination
+        end
+
+        def with_output_destination(destination)
+          self.class.new(
+            source: @source,
+            reference: @reference,
+            key_mapping: @key_mapping,
+            match_options: @match_options,
+            output_destination: destination
+          )
+        end
+      end
+    end
+  end
+end

data/lib/csvtool/domain/cross_csv_dedupe_session/csv_profile.rb

@@ -0,0 +1,24 @@
+# frozen_string_literal: true
+
+module Csvtool
+  module Domain
+    module CrossCsvDedupeSession
+      class CsvProfile
+        attr_reader :path, :separator
+
+        def initialize(path:, separator:, headers_present:)
+          raise ArgumentError, "path cannot be empty" if path.to_s.empty?
+          raise ArgumentError, "separator cannot be empty" if separator.to_s.empty?
+
+          @path = path
+          @separator = separator
+          @headers_present = !!headers_present
+        end
+
+        def headers_present?
+          @headers_present
+        end
+      end
+    end
+  end
+end

data/lib/csvtool/domain/cross_csv_dedupe_session/key_mapping.rb

@@ -0,0 +1,22 @@
+# frozen_string_literal: true
+
+require "csvtool/domain/cross_csv_dedupe_session/column_selector"
+
+module Csvtool
+  module Domain
+    module CrossCsvDedupeSession
+      class KeyMapping
+        attr_reader :source_selector, :reference_selector
+
+        def initialize(source_selector:, reference_selector:)
+          unless source_selector.is_a?(ColumnSelector) && reference_selector.is_a?(ColumnSelector)
+            raise ArgumentError, "selectors must be ColumnSelector"
+          end
+
+          @source_selector = source_selector
+          @reference_selector = reference_selector
+        end
+      end
+    end
+  end
+end

data/lib/csvtool/domain/cross_csv_dedupe_session/match_options.rb

@@ -0,0 +1,29 @@
+# frozen_string_literal: true
+
+module Csvtool
+  module Domain
+    module CrossCsvDedupeSession
+      class MatchOptions
+        attr_reader :trim_whitespace, :case_insensitive
+
+        def initialize(trim_whitespace:, case_insensitive:)
+          @trim_whitespace = !!trim_whitespace
+          @case_insensitive = !!case_insensitive
+        end
+
+        def trim_whitespace?
+          @trim_whitespace
+        end
+
+        def case_insensitive?
+          @case_insensitive
+        end
+
+        def normalize(value)
+          normalized = trim_whitespace? ? value.to_s.strip : value.to_s
+          case_insensitive? ? normalized.downcase : normalized
+        end
+      end
+    end
+  end
+end

data/lib/csvtool/domain/row_randomization_session/randomization_source.rb

@@ -7,6 +7,7 @@ module Csvtool
         attr_reader :path, :separator
 
         def initialize(path:, separator:, headers_present:)
+          raise ArgumentError, "path cannot be empty" if path.to_s.empty?
           raise ArgumentError, "separator cannot be empty" if separator.to_s.empty?
 
           @path = path

data/lib/csvtool/domain/row_session/row_source.rb

@@ -7,6 +7,9 @@ module Csvtool
         attr_reader :path, :separator
 
         def initialize(path:, separator:)
+          raise ArgumentError, "path cannot be empty" if path.to_s.empty?
+          raise ArgumentError, "separator cannot be empty" if separator.to_s.empty?
+
           @path = path
           @separator = separator
         end

data/lib/csvtool/domain/{column_session → shared}/output_destination.rb

@@ -2,7 +2,7 @@
 
 module Csvtool
   module Domain
-    module ColumnSession
+    module Shared
       class OutputDestination
         attr_reader :mode, :path
 

data/lib/csvtool/infrastructure/csv/cross_csv_deduper.rb

@@ -0,0 +1,85 @@
+# frozen_string_literal: true
+
+require "csv"
+require "set"
+require "csvtool/domain/cross_csv_dedupe_session/match_options"
+
+module Csvtool
+  module Infrastructure
+    module CSV
+      class CrossCsvDeduper
+        def call(
+          source_path:,
+          reference_path:,
+          source_selector:,
+          reference_selector:,
+          source_col_sep: ",",
+          reference_col_sep: ",",
+          match_options: Domain::CrossCsvDedupeSession::MatchOptions.new(trim_whitespace: true, case_insensitive: false)
+        )
+          kept_rows = []
+          stats = each_retained(
+            source_path: source_path,
+            reference_path: reference_path,
+            source_selector: source_selector,
+            reference_selector: reference_selector,
+            source_col_sep: source_col_sep,
+            reference_col_sep: reference_col_sep,
+            match_options: match_options
+          ) do |fields|
+            kept_rows << fields
+          end
+
+          stats.merge(kept_rows: kept_rows)
+        end
+
+        def each_retained(
+          source_path:,
+          reference_path:,
+          source_selector:,
+          reference_selector:,
+          source_col_sep: ",",
+          reference_col_sep: ",",
+          match_options: Domain::CrossCsvDedupeSession::MatchOptions.new(trim_whitespace: true, case_insensitive: false)
+        )
+          source_has_headers = source_selector.headers_present?
+          reference_has_headers = reference_selector.headers_present?
+          reference_keys = Set.new
+          ::CSV.foreach(reference_path, headers: reference_has_headers, col_sep: reference_col_sep) do |row|
+            reference_keys << extract_key(row, selector: reference_selector, match_options: match_options)
+          end
+
+          source_header_row = nil
+          source_rows = 0
+          removed_rows = 0
+          kept_rows_count = 0
+
+          ::CSV.foreach(source_path, headers: source_has_headers, col_sep: source_col_sep) do |row|
+            source_header_row ||= row.headers if source_has_headers
+            source_rows += 1
+            key = extract_key(row, selector: source_selector, match_options: match_options)
+            if reference_keys.include?(key)
+              removed_rows += 1
+            else
+              kept_rows_count += 1
+              yield(source_has_headers ? row.fields : row) if block_given?
+            end
+          end
+
+          {
+            headers: source_has_headers ? (source_header_row || []) : nil,
+            source_rows: source_rows,
+            removed_rows: removed_rows,
+            kept_rows_count: kept_rows_count
+          }
+        end
+
+        private
+
+        def extract_key(row, selector:, match_options:)
+          match_options.normalize(selector.extract_from(row))
+        end
+      end
+    end
+  end
+end

data/lib/csvtool/infrastructure/csv/selector_validator.rb

@@ -0,0 +1,30 @@
+# frozen_string_literal: true
+
+require "csv"
+require "csvtool/infrastructure/csv/header_reader"
+
+module Csvtool
+  module Infrastructure
+    module CSV
+      class SelectorValidator
+        def initialize(header_reader: HeaderReader.new)
+          @header_reader = header_reader
+        end
+
+        def valid?(profile:, selector:)
+          if selector.headers_present?
+            headers = @header_reader.call(file_path: profile.path, col_sep: profile.separator)
+            return false if headers.empty?
+
+            headers.include?(selector.value)
+          else
+            first_row = ::CSV.open(profile.path, "r", headers: false, col_sep: profile.separator, &:first)
+            return false if first_row.nil?
+
+            selector.value <= first_row.length
+          end
+        end
+      end
+    end
+  end
+end

data/lib/csvtool/interface/cli/menu_loop.rb

@@ -4,13 +4,14 @@ module Csvtool
   module Interface
     module CLI
       class MenuLoop
-        def initialize(stdin:, stdout:, menu_options:, extract_column_action:, extract_rows_action:, randomize_rows_action:)
+        def initialize(stdin:, stdout:, menu_options:, extract_column_action:, extract_rows_action:, randomize_rows_action:, dedupe_action:)
           @stdin = stdin
           @stdout = stdout
           @menu_options = menu_options
           @extract_column_action = extract_column_action
           @extract_rows_action = extract_rows_action
           @randomize_rows_action = randomize_rows_action
+          @dedupe_action = dedupe_action
         end
 
         def run
@@ -28,9 +29,11 @@ module Csvtool
             when "3"
               @randomize_rows_action.call
             when "4"
+              @dedupe_action.call
+            when "5"
               return 0
             else
-              @stdout.puts "Please choose 1, 2, 3, or 4."
+              @stdout.puts "Please choose 1, 2, 3, 4, or 5."
             end
           end
         end

data/lib/csvtool/interface/cli/workflows/run_cross_csv_dedupe_workflow.rb

@@ -0,0 +1,163 @@
+# frozen_string_literal: true
+
+require "csv"
+require "csvtool/application/use_cases/run_cross_csv_dedupe"
+require "csvtool/interface/cli/errors/presenter"
+require "csvtool/interface/cli/prompts/file_path_prompt"
+require "csvtool/interface/cli/prompts/separator_prompt"
+require "csvtool/interface/cli/prompts/output_destination_prompt"
+require "csvtool/domain/cross_csv_dedupe_session/csv_profile"
+require "csvtool/domain/cross_csv_dedupe_session/column_selector"
+require "csvtool/domain/cross_csv_dedupe_session/key_mapping"
+require "csvtool/domain/cross_csv_dedupe_session/match_options"
+require "csvtool/domain/cross_csv_dedupe_session/cross_csv_dedupe_session"
+require "csvtool/domain/shared/output_destination"
+
+module Csvtool
+  module Interface
+    module CLI
+      module Workflows
+        class RunCrossCsvDedupeWorkflow
+          def initialize(stdin:, stdout:, use_case: Application::UseCases::RunCrossCsvDedupe.new)
+            @stdin = stdin
+            @stdout = stdout
+            @use_case = use_case
+            @errors = Interface::CLI::Errors::Presenter.new(stdout: stdout)
+          end
+
+          def call
+            source_path = Interface::CLI::Prompts::FilePathPrompt.new(stdin: @stdin, stdout: @stdout).call
+            return @errors.file_not_found(source_path) unless File.file?(source_path)
+
+            @stdout.puts "Source CSV separator:"
+            source_col_sep = Interface::CLI::Prompts::SeparatorPrompt.new(stdin: @stdin, stdout: @stdout, errors: @errors).call
+            return if source_col_sep.nil?
+            @stdout.print "Source headers present? [Y/n]: "
+            source_headers_present = !%w[n no].include?(@stdin.gets&.strip.to_s.downcase)
+            source = Domain::CrossCsvDedupeSession::CsvProfile.new(
+              path: source_path,
+              separator: source_col_sep,
+              headers_present: source_headers_present
+            )
+
+            @stdout.print "Reference CSV file path: "
+            reference_path = @stdin.gets&.strip.to_s
+            return @errors.file_not_found(reference_path) unless File.file?(reference_path)
+
+            @stdout.puts "Reference CSV separator:"
+            reference_col_sep = Interface::CLI::Prompts::SeparatorPrompt.new(stdin: @stdin, stdout: @stdout, errors: @errors).call
+            return if reference_col_sep.nil?
+            @stdout.print "Reference headers present? [Y/n]: "
+            reference_headers_present = !%w[n no].include?(@stdin.gets&.strip.to_s.downcase)
+            reference = Domain::CrossCsvDedupeSession::CsvProfile.new(
+              path: reference_path,
+              separator: reference_col_sep,
+              headers_present: reference_headers_present
+            )
+
+            source_selector = prompt_selector("Source", source.headers_present?)
+            return @errors.column_not_found if source_selector.nil?
+            reference_selector = prompt_selector("Reference", reference.headers_present?)
+            return @errors.column_not_found if reference_selector.nil?
+
+            @stdout.print "Trim whitespace before matching? [Y/n]: "
+            trim_whitespace = read_yes_no(default: true)
+            @stdout.print "Case-insensitive matching? [y/N]: "
+            case_insensitive = read_yes_no(default: false)
+
+            key_mapping = Domain::CrossCsvDedupeSession::KeyMapping.new(
+              source_selector: source_selector,
+              reference_selector: reference_selector
+            )
+            match_options = Domain::CrossCsvDedupeSession::MatchOptions.new(
+              trim_whitespace: trim_whitespace,
+              case_insensitive: case_insensitive
+            )
+            session = Domain::CrossCsvDedupeSession::CrossCsvDedupeSession.start(
+              source: source,
+              reference: reference,
+              key_mapping: key_mapping,
+              match_options: match_options
+            )
+
+            output_destination = Interface::CLI::Prompts::OutputDestinationPrompt.new(
+              stdin: @stdin,
+              stdout: @stdout,
+              errors: @errors
+            ).call
+            return if output_destination.nil?
+            session = session.with_output_destination(
+              if output_destination[:mode] == :file
+                Domain::Shared::OutputDestination.file(path: output_destination[:path])
+              else
+                Domain::Shared::OutputDestination.console
+              end
+            )
+
+            result = @use_case.call(
+              session: session,
+              on_header: ->(headers) { print_header(headers, col_sep: session.source.separator) },
+              on_row: ->(fields) { print_row(fields, col_sep: session.source.separator) }
+            )
+            return handle_error(result) unless result.ok?
+
+            @stdout.puts "Wrote output to #{result.data[:output_path]}" if session.output_destination.file?
+            stats = result.data[:stats]
+            @stdout.puts "Summary: source_rows=#{stats[:source_rows]} removed_rows=#{stats[:removed_rows]} kept_rows=#{stats[:kept_rows_count]}"
+            @stdout.puts "No rows removed; no matching keys found." if stats[:removed_rows].zero?
+            @stdout.puts "All source rows were removed by dedupe." if stats[:source_rows].positive? && stats[:kept_rows_count].zero?
+          rescue ArgumentError => e
+            return @errors.empty_output_path if e.message == "file output path cannot be empty"
+
+            raise e
+          end
+
+          private
+
+          def prompt_selector(label, headers_present)
+            if headers_present
+              @stdout.print "#{label} key column name: "
+            else
+              @stdout.print "#{label} key column index (1-based): "
+            end
+            input = @stdin.gets&.strip.to_s
+            Domain::CrossCsvDedupeSession::ColumnSelector.from_input(headers_present: headers_present, input: input)
+          rescue ArgumentError
+            nil
+          end
+
+          def print_header(headers, col_sep:)
+            @stdout.puts
+            @stdout.puts ::CSV.generate_line(headers, row_sep: "", col_sep: col_sep).chomp
+          end
+
+          def print_row(fields, col_sep:)
+            @stdout.puts ::CSV.generate_line(fields, row_sep: "", col_sep: col_sep).chomp
+          end
+
+          def handle_error(result)
+            case result.error
+            when :column_not_found
+              @errors.column_not_found
+            when :could_not_parse_csv
+              @errors.could_not_parse_csv
+            when :cannot_read_file
+              @errors.cannot_read_file(result.data[:path])
+            when :cannot_write_output_file
+              @errors.cannot_write_output_file(result.data[:path], result.data[:error_class])
+            end
+          end
+
+          def read_yes_no(default:)
+            answer = @stdin.gets&.strip.to_s.downcase
+            return default if answer.empty?
+            return true if %w[y yes].include?(answer)
+            return false if %w[n no].include?(answer)
+
+            default
+          end
+        end
+      end
+    end
+  end
+end

data/lib/csvtool/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Csvtool
-  VERSION = "0.3.0.alpha"
+  VERSION = "0.4.0.alpha"
 end