jrf 0.1.5 → 0.1.6

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 408c1f9706af5efaa1bf0125201d6647b4c108aa4aa28c99a93b59fb9cc94f02
-  data.tar.gz: 702f2fb14dc9d498292b02c41f0cdb4a91c0fa3e093ad9a71435d9a2604532fa
+  metadata.gz: e9bb2a3a16d2bbe8cfb463267ff74d7d582511d4b4891e56ad3dfa6eee75fceb
+  data.tar.gz: a13b2e9c8517c3da997452166556505b24fc4d5f898765ad33495eafd57c3081
 SHA512:
-  metadata.gz: 80dfa6d2bb7c9304e779a3e80815efbde9c599d66665708738b833b08daa1918ae54bc5b170c8b90c60399fe18b0df06d576e2c8c3d8b76b74f9daa826efcfa8
-  data.tar.gz: 597b715fd3ebd31a49cb2839f7dda814b845cd5aa87a3ac9a9cf551553792b453af749e287652553903de851ea7b06a9e5940abc7c25fccd319a9e7e72d75840
+  metadata.gz: 54b400cdaba584896f2511acfe9a41ef10af25033bf88cfc6e0386eaa840df9395fb0d008c320b3193d55a9c3fad444a7f54bd29f52c34f69bc9a9cf392a7809
+  data.tar.gz: 80c72675e179da483316bfeaee7114da6edb49dc66ae179aa072d48907c4c9caf74113c6681b2f4a83f4b97da6faac436f5d6af5bd31e82605b122d85892cede

data/lib/jrf/cli/runner.rb

@@ -0,0 +1,126 @@
+# frozen_string_literal: true
+
+require "json"
+require_relative "../pipeline"
+require_relative "../pipeline_parser"
+
+module Jrf
+  class CLI
+    class Runner
+      RS_CHAR = "\x1e"
+      DEFAULT_OUTPUT_BUFFER_LIMIT = 4096
+
+      class RsNormalizer
+        def initialize(input)
+          @input = input
+        end
+
+        def read(length = nil, outbuf = nil)
+          chunk = @input.read(length)
+          return nil if chunk.nil?
+
+          chunk = chunk.tr(RS_CHAR, "\n")
+          if outbuf
+            outbuf.replace(chunk)
+          else
+            chunk
+          end
+        end
+      end
+
+      def initialize(inputs:, out: $stdout, err: $stderr, lax: false, pretty: false, atomic_write_bytes: DEFAULT_OUTPUT_BUFFER_LIMIT)
+        @inputs = inputs
+        @out = out
+        @err = err
+        @lax = lax
+        @pretty = pretty
+        @atomic_write_bytes = atomic_write_bytes
+        @output_buffer = +""
+      end
+
+      def run(expression, verbose: false)
+        parsed = PipelineParser.new(expression).parse
+        stages = parsed[:stages]
+        dump_stages(stages) if verbose
+
+        blocks = stages.map { |stage|
+          eval("proc { #{stage[:src]} }", nil, "(jrf stage)", 1) # rubocop:disable Security/Eval
+        }
+        pipeline = Pipeline.new(*blocks)
+
+        input_enum = Enumerator.new { |y| each_input_value { |v| y << v } }
+        pipeline.call(input_enum) do |value|
+          emit_output(value)
+        end
+      ensure
+        write_output(@output_buffer)
+      end
+
+      private
+
+      def each_input_value
+        return each_input_value_lax { |value| yield value } if @lax
+
+        each_input_value_ndjson { |value| yield value }
+      end
+
+      def each_input_value_ndjson
+        each_input do |source|
+          source.each_line do |raw_line|
+            line = raw_line.strip
+            next if line.empty?
+
+            yield JSON.parse(line)
+          end
+        end
+      end
+
+      def each_input_value_lax
+        require "oj"
+        handler = Class.new(Oj::ScHandler) do
+          def initialize(&emit)
+            @emit = emit
+          end
+
+          def hash_start = {}
+          def hash_key(key) = key
+          def hash_set(hash, key, value) = hash[key] = value
+          def array_start = []
+          def array_append(array, value) = array << value
+          def add_value(value) = @emit.call(value)
+        end
+        each_input do |source|
+          Oj.sc_parse(handler.new { |value| yield value }, RsNormalizer.new(source))
+        end
+      rescue LoadError
+        raise "oj is required for --lax mode (gem install oj)"
+      rescue Oj::ParseError => e
+        raise JSON::ParserError, e.message
+      end
+
+      def dump_stages(stages)
+        stages.each_with_index do |stage, i|
+          @err.puts "stage[#{i}]: #{stage[:src]}"
+        end
+      end
+
+      def each_input
+        @inputs.each { |source| yield source }
+      end
+
+      def emit_output(value)
+        record = (@pretty ? JSON.pretty_generate(value) : JSON.generate(value)) << "\n"
+        if @output_buffer.bytesize + record.bytesize <= @atomic_write_bytes
+          @output_buffer << record
+        else
+          write_output(@output_buffer)
+          @output_buffer = record
+        end
+      end
+
+      def write_output(str)
+        @out.syswrite(str)
+      end
+    end
+  end
+end

data/lib/jrf/cli.rb

@@ -1,13 +1,13 @@
 # frozen_string_literal: true
 
-require_relative "runner"
+require_relative "cli/runner"
 
 module Jrf
   class CLI
-    USAGE = "usage: jrf [-v] [--lax] [--pretty] [--help] 'STAGE >> STAGE >> ...'"
+    USAGE = "usage: jrf [options] 'STAGE >> STAGE >> ...'"
 
     HELP_TEXT = <<~'TEXT'
-      usage: jrf [-v] [--lax] [--pretty] [--help] 'STAGE >> STAGE >> ...'
+      usage: jrf [options] 'STAGE >> STAGE >> ...'
 
       JSON filter with the power and speed of Ruby.
 
@@ -15,6 +15,8 @@ module Jrf
         -v, --verbose  print parsed stage expressions
         --lax          allow multiline JSON texts; split inputs by whitespace (also detects JSON-SEQ RS 0x1e)
         -p, --pretty   pretty-print JSON output instead of compact NDJSON
+        --atomic-write-bytes N
+                       group short outputs into atomic writes of up to N bytes
         -h, --help     show this help and exit
 
       Pipeline:
@@ -36,6 +38,7 @@ module Jrf
       verbose = false
       lax = false
       pretty = false
+      atomic_write_bytes = Runner::DEFAULT_OUTPUT_BUFFER_LIMIT
 
       while argv.first&.start_with?("-")
         case argv.first
@@ -48,6 +51,14 @@ module Jrf
         when "-p", "--pretty"
           pretty = true
           argv.shift
+        when /\A--atomic-write-bytes=(.+)\z/
+          atomic_write_bytes = parse_atomic_write_bytes(Regexp.last_match(1), err)
+          return 1 unless atomic_write_bytes
+          argv.shift
+        when "--atomic-write-bytes"
+          argv.shift
+          atomic_write_bytes = parse_atomic_write_bytes(argv.shift, err)
+          return 1 unless atomic_write_bytes
         when "-h", "--help"
           out.puts HELP_TEXT
           return 0
@@ -64,8 +75,43 @@ module Jrf
       end
 
       expression = argv.shift
-      Runner.new(input: input, out: out, err: err, lax: lax, pretty: pretty).run(expression, verbose: verbose)
+      inputs = Enumerator.new do |y|
+        if argv.empty?
+          y << input
+        else
+          argv.each do |path|
+            if path == "-"
+              y << input
+            elsif path.end_with?(".gz")
+              require "zlib"
+              Zlib::GzipReader.open(path) do |source|
+                y << source
+              end
+            else
+              File.open(path, "rb") do |source|
+                y << source
+              end
+            end
+          end
+        end
+      end
+      Runner.new(
+        inputs: inputs,
+        out: out,
+        err: err,
+        lax: lax,
+        pretty: pretty,
+        atomic_write_bytes: atomic_write_bytes
+      ).run(expression, verbose: verbose)
       0
     end
+
+    def self.parse_atomic_write_bytes(value, err)
+      bytes = Integer(value, exception: false)
+      return bytes if bytes && bytes.positive?
+
+      err.puts "--atomic-write-bytes requires a positive integer"
+      nil
+    end
   end
 end

data/lib/jrf/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Jrf
-  VERSION = "0.1.5"
+  VERSION = "0.1.6"
 end

data/test/jrf_test.rb

@@ -1,7 +1,17 @@
 # frozen_string_literal: true
 
+begin
+  require "bundler/setup"
+rescue LoadError
+  # Allow running tests in plain Ruby environments with globally installed gems.
+end
+
 require "json"
 require "open3"
+require "stringio"
+require "tmpdir"
+require "zlib"
+require_relative "../lib/jrf/cli/runner"
 
 def run_jrf(expr, input, *opts)
   Open3.capture3("./exe/jrf", *opts, expr, stdin_data: input)
@@ -41,6 +51,45 @@ def lines(str)
   str.lines.map(&:strip).reject(&:empty?)
 end
 
+class RecordingRunner < Jrf::CLI::Runner
+  attr_reader :writes
+
+  def initialize(**kwargs)
+    super
+    @writes = []
+  end
+
+  private
+
+  def write_output(str)
+    return if str.empty?
+
+    @writes << str
+  end
+end
+
+class ChunkedSource
+  def initialize(str, chunk_size: 5)
+    @str = str
+    @chunk_size = chunk_size
+    @offset = 0
+  end
+
+  def read(length = nil, outbuf = nil)
+    raise "expected chunked reads" if length.nil?
+
+    chunk = @str.byteslice(@offset, [length, @chunk_size].min)
+    return nil unless chunk
+
+    @offset += chunk.bytesize
+    if outbuf
+      outbuf.replace(chunk)
+    else
+      chunk
+    end
+  end
+end
+
 File.chmod(0o755, "./exe/jrf")
 
 input = <<~NDJSON
@@ -92,10 +141,11 @@ assert_includes(stderr, 'stage[1]: _["hello"]')
 
 stdout, stderr, status = Open3.capture3("./exe/jrf", "--help")
 assert_success(status, stderr, "help option")
-assert_includes(stdout, "usage: jrf [-v] [--lax] [--pretty] [--help] 'STAGE >> STAGE >> ...'")
+assert_includes(stdout, "usage: jrf [options] 'STAGE >> STAGE >> ...'")
 assert_includes(stdout, "JSON filter with the power and speed of Ruby.")
 assert_includes(stdout, "--lax")
 assert_includes(stdout, "--pretty")
+assert_includes(stdout, "--atomic-write-bytes N")
 assert_includes(stdout, "Pipeline:")
 assert_includes(stdout, "Connect stages with top-level >>.")
 assert_includes(stdout, "The current value in each stage is available as _.")
@@ -103,11 +153,72 @@ assert_includes(stdout, "See Also:")
 assert_includes(stdout, "https://github.com/kazuho/jrf#readme")
 assert_equal([], lines(stderr), "help stderr output")
 
+threshold_input = StringIO.new((1..4).map { |i| "{\"foo\":\"#{'x' * 1020}\",\"i\":#{i}}\n" }.join)
+buffered_runner = RecordingRunner.new(inputs: [threshold_input], out: StringIO.new, err: StringIO.new)
+buffered_runner.run('_')
+expected_line = JSON.generate({"foo" => "x" * 1020, "i" => 1}) + "\n"
+assert_equal(2, buffered_runner.writes.length, "default atomic write limit buffers records until the configured threshold")
+assert_equal(expected_line.bytesize * 3, buffered_runner.writes.first.bytesize, "default atomic write limit flushes before the next record would exceed the threshold")
+assert_equal(expected_line.bytesize, buffered_runner.writes.last.bytesize, "final buffer flush emits the remaining record")
+
+small_limit_runner = RecordingRunner.new(inputs: [StringIO.new("{\"foo\":1}\n{\"foo\":2}\n")], out: StringIO.new, err: StringIO.new, atomic_write_bytes: 1)
+small_limit_runner.run('_["foo"]')
+assert_equal(["1\n", "2\n"], small_limit_runner.writes, "small atomic write limit emits oversized records directly")
+
+error_runner = RecordingRunner.new(inputs: [StringIO.new("{\"foo\":1}\n{\"foo\":")], out: StringIO.new, err: StringIO.new)
+begin
+  error_runner.run('_["foo"]')
+  raise "expected parse error for buffered flush test"
+rescue JSON::ParserError
+  assert_equal(["1\n"], error_runner.writes, "buffer flushes pending output before parse errors escape")
+end
+
 stdout, stderr, status = run_jrf('select(_["hello"] == 123) >> _["hello"]', input_hello, "--verbose")
 assert_success(status, stderr, "dump stages verbose alias")
 assert_equal(%w[123], lines(stdout), "dump stages verbose alias output")
 assert_includes(stderr, 'stage[0]: select(_["hello"] == 123)')
 
+stdout, stderr, status = run_jrf('_["hello"]', input_hello, "--atomic-write-bytes", "512")
+assert_success(status, stderr, "atomic write bytes option")
+assert_equal(%w[123 456], lines(stdout), "atomic write bytes option output")
+
+stdout, stderr, status = run_jrf('_["hello"]', input_hello, "--atomic-write-bytes=512")
+assert_success(status, stderr, "atomic write bytes equals form")
+assert_equal(%w[123 456], lines(stdout), "atomic write bytes equals form output")
+
+stdout, stderr, status = Open3.capture3("./exe/jrf", "--atomic-write-bytes", "0", '_["hello"]', stdin_data: input_hello)
+assert_failure(status, "atomic write bytes rejects zero")
+assert_includes(stderr, "--atomic-write-bytes requires a positive integer")
+
+Dir.mktmpdir do |dir|
+  gz_path = File.join(dir, "input.ndjson.gz")
+  Zlib::GzipWriter.open(gz_path) do |io|
+    io.write("{\"foo\":10}\n{\"foo\":20}\n")
+  end
+
+  stdout, stderr, status = Open3.capture3("./exe/jrf", '_["foo"]', gz_path)
+  assert_success(status, stderr, "compressed input by suffix")
+  assert_equal(%w[10 20], lines(stdout), "compressed input output")
+
+  lax_gz_path = File.join(dir, "input-lax.json.gz")
+  Zlib::GzipWriter.open(lax_gz_path) do |io|
+    io.write("{\"foo\":30}\n\x1e{\"foo\":40}\n")
+  end
+
+  stdout, stderr, status = Open3.capture3("./exe/jrf", "--lax", '_["foo"]', lax_gz_path)
+  assert_success(status, stderr, "compressed lax input by suffix")
+  assert_equal(%w[30 40], lines(stdout), "compressed lax input output")
+
+  second_gz_path = File.join(dir, "input2.ndjson.gz")
+  Zlib::GzipWriter.open(second_gz_path) do |io|
+    io.write("{\"foo\":50}\n")
+  end
+
+  stdout, stderr, status = Open3.capture3("./exe/jrf", '_["foo"]', gz_path, second_gz_path)
+  assert_success(status, stderr, "multiple compressed inputs by suffix")
+  assert_equal(%w[10 20 50], lines(stdout), "multiple compressed input output")
+end
+
 stdout, stderr, status = run_jrf('_', input_hello, "--pretty")
 assert_success(status, stderr, "pretty output")
 assert_equal(
@@ -493,6 +604,26 @@ stdout, stderr, status = run_jrf('_["foo"]', input_lax_trailing_rs, "--lax")
 assert_success(status, stderr, "lax ignores trailing separator")
 assert_equal(%w[9], lines(stdout), "lax trailing separator output")
 
+chunked_lax_out = RecordingRunner.new(
+  inputs: [ChunkedSource.new("{\"foo\":1}\n\x1e{\"foo\":2}\n\t{\"foo\":3}\n")],
+  out: StringIO.new,
+  err: StringIO.new,
+  lax: true
+)
+chunked_lax_out.run('_["foo"]')
+assert_equal(%w[1 2 3], lines(chunked_lax_out.writes.join), "lax mode streams chunked input without whole-input reads")
+
+Dir.mktmpdir do |dir|
+  one = File.join(dir, "one.json")
+  two = File.join(dir, "two.json")
+  File.write(one, "1")
+  File.write(two, "2")
+
+  stdout, stderr, status = Open3.capture3("./exe/jrf", "--lax", "_", one, two)
+  assert_success(status, stderr, "lax keeps file boundaries")
+  assert_equal(%w[1 2], lines(stdout), "lax does not merge JSON across file boundaries")
+end
+
 stdout, stderr, status = run_jrf('select(_["x"] > ) >> _["foo"]', "")
 assert_failure(status, "syntax error should fail before row loop")
 assert_includes(stderr, "syntax error")

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: jrf
 version: !ruby/object:Gem::Version
-  version: 0.1.5
+  version: 0.1.6
 platform: ruby
 authors:
 - kazuho
@@ -40,12 +40,12 @@ files:
 - jrf.gemspec
 - lib/jrf.rb
 - lib/jrf/cli.rb
+- lib/jrf/cli/runner.rb
 - lib/jrf/control.rb
 - lib/jrf/pipeline.rb
 - lib/jrf/pipeline_parser.rb
 - lib/jrf/reducers.rb
 - lib/jrf/row_context.rb
-- lib/jrf/runner.rb
 - lib/jrf/stage.rb
 - lib/jrf/version.rb
 - test/jrf_test.rb

data/lib/jrf/runner.rb

@@ -1,81 +0,0 @@
-# frozen_string_literal: true
-
-require "json"
-require_relative "pipeline"
-require_relative "pipeline_parser"
-
-module Jrf
-  class Runner
-    RS_CHAR = "\x1e"
-
-    def initialize(input: ARGF, out: $stdout, err: $stderr, lax: false, pretty: false)
-      @input = input
-      @out = out
-      @err = err
-      @lax = lax
-      @pretty = pretty
-    end
-
-    def run(expression, verbose: false)
-      parsed = PipelineParser.new(expression).parse
-      stages = parsed[:stages]
-      dump_stages(stages) if verbose
-
-      blocks = stages.map { |stage|
-        eval("proc { #{stage[:src]} }", nil, "(jrf stage)", 1) # rubocop:disable Security/Eval
-      }
-      pipeline = Pipeline.new(*blocks)
-
-      input_enum = Enumerator.new { |y| each_input_value { |v| y << v } }
-      pipeline.call(input_enum) do |value|
-        @out.puts(@pretty ? JSON.pretty_generate(value) : JSON.generate(value))
-      end
-    end
-
-    private
-
-    def each_input_value
-      return each_input_value_lax { |value| yield value } if @lax
-
-      each_input_value_ndjson { |value| yield value }
-    end
-
-    def each_input_value_ndjson
-      @input.each_line do |raw_line|
-        line = raw_line.strip
-        next if line.empty?
-
-        yield JSON.parse(line)
-      end
-    end
-
-    def each_input_value_lax
-      require "oj"
-      source = @input.read.to_s
-      source = source.include?(RS_CHAR) ? source.tr(RS_CHAR, "\n") : source
-      handler = Class.new(Oj::ScHandler) do
-        def initialize(&emit)
-          @emit = emit
-        end
-
-        def hash_start = {}
-        def hash_key(key) = key
-        def hash_set(hash, key, value) = hash[key] = value
-        def array_start = []
-        def array_append(array, value) = array << value
-        def add_value(value) = @emit.call(value)
-      end.new { |value| yield value }
-      Oj.sc_parse(handler, source)
-    rescue LoadError
-      raise "oj is required for --lax mode (gem install oj)"
-    rescue Oj::ParseError => e
-      raise JSON::ParserError, e.message
-    end
-
-    def dump_stages(stages)
-      stages.each_with_index do |stage, i|
-        @err.puts "stage[#{i}]: #{stage[:src]}"
-      end
-    end
-  end
-end