git 4.4.1 → 5.0.0.beta.1

data/lib/git/command_line/result.rb

@@ -0,0 +1,96 @@
+# frozen_string_literal: true
+
+module Git
+  module CommandLine
+    # The result of running a git command
+    #
+    # This object stores the Git command executed and its status, stdout, and stderr.
+    #
+    # @api public
+    #
+    class Result
+      # Create a Result object
+      #
+      # @example
+      #   git_cmd = %w[git version]
+      #   status = instance_double(ProcessExecuter::Result)
+      #   stdout = "git version 2.39.1\n"
+      #   stderr = ""
+      #   result = Git::CommandLine::Result.new(git_cmd, status, stdout, stderr)
+      #
+      # @param git_cmd [Array<String>] the git command that was executed
+      #
+      # @param status [ProcessExecuter::Result] the process result object returned
+      #   by `ProcessExecuter.run` or `ProcessExecuter.run_with_capture`.
+      #   Responds to `timed_out?`, `signaled?`, and `success?`.
+      #
+      # @param stdout [String] the processed stdout of the process
+      #
+      # @param stderr [String] the processed stderr of the process
+      #
+      def initialize(git_cmd, status, stdout, stderr)
+        @git_cmd = git_cmd
+        @status = status
+        @stdout = stdout
+        @stderr = stderr
+      end
+
+      # @attribute [r] git_cmd
+      #
+      # The git command that was executed
+      #
+      # @example
+      #   git_cmd = %w[git version]
+      #   result = Git::CommandLine::Result.new(git_cmd, nil, '', '')
+      #   result.git_cmd #=> ["git", "version"]
+      #
+      # @return [Array<String>]
+      #
+      attr_reader :git_cmd
+
+      # @attribute [r] status
+      #
+      # The process result object returned by ProcessExecuter
+      #
+      # In practice this is a `ProcessExecuter::ResultWithCapture` (from
+      # {Git::CommandLine::Capturing}) or a `ProcessExecuter::Result` (from
+      # {Git::CommandLine::Streaming}). Both respond to `success?`, `timed_out?`,
+      # and `signaled?`.
+      #
+      # @example
+      #   status = instance_double(ProcessExecuter::Result, success?: true)
+      #   result = Git::CommandLine::Result.new(%w[git version], status, '', '')
+      #   result.status == status #=> true
+      #
+      # @return [ProcessExecuter::Result]
+      #
+      attr_reader :status
+
+      # @attribute [r] stdout
+      #
+      # The output of the process
+      #
+      # @example
+      #   stdout = "git version 2.39.1\n"
+      #   result = Git::CommandLine::Result.new([], nil, stdout, '')
+      #   result.stdout #=> "git version 2.39.1\n"
+      #
+      # @return [String]
+      #
+      attr_reader :stdout
+
+      # @attribute [r] stderr
+      #
+      # The error output of the process
+      #
+      # @example
+      #   stderr = "Tag not found\n"
+      #   result = Git::CommandLine::Result.new([], nil, '', stderr)
+      #   result.stderr #=> "Tag not found\n"
+      #
+      # @return [String]
+      #
+      attr_reader :stderr
+    end
+  end
+end

data/lib/git/command_line/streaming.rb

@@ -0,0 +1,194 @@
+# frozen_string_literal: true
+
+require 'git/command_line'
+require 'stringio'
+
+module Git
+  module CommandLine
+    # Executes a git command in streaming mode without buffering stdout in memory
+    #
+    # {Git::CommandLine::Streaming} is the non-buffering strategy: it calls
+    # `ProcessExecuter.run` and streams stdout directly to the caller-supplied `out:`
+    # IO object.  Stderr is always captured internally in a `StringIO` for error
+    # diagnostics and is available as `result.stderr`.
+    #
+    # Use this class (via {Git::Lib#command_streaming}) for commands such as
+    # `cat-file -p <blob>` whose stdout may be too large to buffer in memory.
+    #
+    # {Git::CommandLine::Capturing} is the complementary strategy for the common case
+    # where buffering stdout is acceptable.
+    #
+    # @example Stream a blob to a file
+    #   streaming = Git::CommandLine::Streaming.new(
+    #     {}, '/usr/bin/git', %w[--git-dir /repo/.git], Logger.new($stdout)
+    #   )
+    #   File.open('/tmp/blob', 'wb') do |f|
+    #     streaming.run('cat-file', 'blob', sha, out: f)
+    #   end
+    #
+    # @see Git::Lib#command_streaming
+    #
+    # @see Git::CommandLine::Capturing
+    #
+    class Streaming < Git::CommandLine::Base
+      # Default options accepted by {#run}
+      #
+      # @api private
+      RUN_OPTION_DEFAULTS = {
+        in: nil,
+        out: nil,
+        err: nil,
+        chdir: nil,
+        timeout: nil,
+        raise_on_failure: true,
+        env: {}
+      }.freeze
+
+      # Execute a git command in streaming mode and return the result
+      #
+      # Unlike {Git::CommandLine::Capturing#run}, this method does **not** buffer
+      # stdout in memory.  Stdout is written only to the IO object provided via the
+      # `out:` option.  Stderr is captured internally via a `StringIO` for error
+      # diagnostics.
+      #
+      # Use this entry point for commands that stream large content (e.g. blobs)
+      # where capturing stdout in memory would be unacceptable.
+      #
+      # @example Stream a blob to a file
+      #   file = File.open('/tmp/blob', 'wb')
+      #   streaming.run('cat-file', 'blob', sha, out: file)
+      #
+      # @param options_hash [Hash] the options to pass to the command
+      #
+      # @option options_hash [IO, nil] :in the IO object to use as stdin for the
+      #   command, or nil to inherit the parent process stdin.  Must be a real IO
+      #   object with a file descriptor (not StringIO).
+      #
+      # @option options_hash [#write, nil] :out the IO/object to stream stdout into.
+      #   Stdout is NOT buffered in the returned result; this is the only way to
+      #   read it.
+      #
+      # @option options_hash [#write, nil] :err an optional additional destination to
+      #   receive stderr output in real time (e.g. `$stderr` or a `File`).  Stderr is
+      #   always captured internally in a `StringIO` for error diagnostics.  When
+      #   `err:` is provided, writes are teed to both the internal buffer and this
+      #   destination.  `result.stderr` always reflects what was captured in the
+      #   internal buffer, regardless of whether `err:` is supplied.
+      #
+      # @option options_hash [String, nil] :chdir the directory to run the command in
+      #
+      # @option options_hash [Numeric, nil] :timeout the maximum seconds to wait for
+      #   the command to complete.  Zero means no timeout.  A timeout kills the
+      #   process via `SIGKILL` and raises {Git::TimeoutError}.
+      #
+      # @option options_hash [Boolean] :raise_on_failure (true) whether to raise
+      #   {Git::FailedError} on non-zero exit status.
+      #   {Git::TimeoutError} and {Git::SignaledError} are always raised regardless.
+      #
+      # @option options_hash [Hash] :env ({}) additional environment variable
+      #   overrides for this command.  String keys map to String values (to set) or
+      #   `nil` (to unset).
+      #
+      # @return [Git::CommandLineResult] the result of the command
+      #
+      #   `result.stdout` will always be `''` (empty) — stdout was streamed to `out:`.
+      #   `result.stderr` contains any stderr output captured for diagnostics.
+      #
+      # @raise [ArgumentError] if `args` contains an array or an unknown option is
+      #   passed
+      #
+      # @raise [Git::SignaledError] if the command was terminated by an uncaught signal
+      #
+      # @raise [Git::FailedError] if the command returned a non-zero exit status
+      #
+      # @raise [Git::ProcessIOError] if an exception was raised while collecting
+      #   subprocess output
+      #
+      # @raise [Git::TimeoutError] if the command times out
+      #
+      def run(*, **options_hash)
+        options = merge_and_validate_options(RUN_OPTION_DEFAULTS, options_hash)
+
+        internal_err = StringIO.new
+        # Tee stderr to the caller-provided destination (if any) AND the internal
+        # StringIO.  This ensures result.stderr is always available even when err:
+        # is a non-StringIO IO object.
+        err_dest = options[:err] ? build_stderr_tee(internal_err, options[:err]) : internal_err
+        result = execute(*, err_io: err_dest, **options)
+        process_result(result, internal_err, options)
+      end
+
+      private
+
+      # @return [ProcessExecuter::Result] the result of running the command (non-capturing)
+      #
+      # @api private
+      def execute(*args, err_io:, **options_hash)
+        git_cmd = build_git_cmd(args)
+        options = execute_options(err_io:, **options_hash)
+        run_process_executer do
+          ProcessExecuter.run(merged_env(options_hash), *git_cmd, **options)
+        end
+      end
+
+      # Build the ProcessExecuter options hash for a streaming run
+      #
+      # @return [Hash]
+      #
+      # @api private
+      def execute_options(err_io:, **options_hash)
+        chdir = options_hash[:chdir] || :not_set
+        timeout_after = options_hash[:timeout]
+
+        { chdir:, timeout_after:, raise_errors: false, err: err_io }.tap do |options|
+          options[:in] = options_hash[:in] unless options_hash[:in].nil?
+          options[:out] = options_hash[:out] unless options_hash[:out].nil?
+        end
+      end
+
+      # Build a tee writer that forwards #write calls to two destinations simultaneously.
+      #
+      # Used to capture stderr in an internal StringIO while also streaming to a
+      # caller-provided destination.
+      #
+      # @param primary [StringIO] the internal capture buffer
+      #
+      # @param secondary [#write] the caller-supplied destination
+      #
+      # @return [#write] an object whose #write method delegates to both destinations
+      #
+      # @api private
+      def build_stderr_tee(primary, secondary)
+        ::Object.new.tap do |tee|
+          tee.define_singleton_method(:write) do |data|
+            primary.write(data)
+            secondary.write(data)
+            data.bytesize
+          end
+        end
+      end
+
+      # Process the result of a streaming command and return a Git::CommandLineResult
+      #
+      # Constructs stdout as `''` (not captured) and stderr from the internal StringIO.
+      #
+      # @param result [ProcessExecuter::Result] the raw process result
+      #
+      # @param err_io [StringIO] the internal StringIO that captured stderr
+      #
+      # @param options [Hash] the merged run options
+      #
+      # @return [Git::CommandLineResult]
+      #
+      # @api private
+      def process_result(result, err_io, options)
+        command = result.command
+        stderr = err_io.string
+        log_result(result, command, '', stderr)
+        command_line_result(
+          command, result, '', stderr, options[:timeout], options[:raise_on_failure]
+        )
+      end
+    end
+  end
+end

data/lib/git/command_line.rb

@@ -1,330 +1,51 @@
 # frozen_string_literal: true
 
-require 'git/base'
-require 'git/command_line_result'
-require 'git/errors'
-require 'stringio'
-
 module Git
-  # Runs a git command and returns the result
+  # Namespace module for git command-line execution strategies
+  #
+  # This module groups the classes responsible for invoking git subprocesses
+  # and handling their output. Choose a concrete class based on your buffering
+  # needs:
+  #
+  # * {Git::CommandLine::Capturing} — buffers stdout and stderr in memory.
+  #   Use this for the vast majority of git commands whose output fits in memory.
+  #
+  # * {Git::CommandLine::Streaming} — streams stdout to a caller-supplied IO.
+  #   Use this for commands (e.g. `cat-file -p <blob>`) whose output may be
+  #   too large to buffer.
+  #
+  # Both classes inherit from {Git::CommandLine::Base} and are instantiated
+  # via factory helpers in {Git::Lib}: {Git::Lib#command_capturing} and
+  # {Git::Lib#command_streaming}.
+  #
+  # Results are returned as {Git::CommandLine::Result} objects (also accessible
+  # as {Git::CommandLineResult} for backward compatibility).
   #
   # @api public
   #
-  class CommandLine
-    # Create a Git::CommandLine object
-    #
-    # @example
-    #   env = { 'GIT_DIR' => '/path/to/git/dir' }
-    #   binary_path = '/usr/bin/git'
-    #   global_opts = %w[--git-dir /path/to/git/dir]
-    #   logger = Logger.new(STDOUT)
-    #   cli = CommandLine.new(env, binary_path, global_opts, logger)
-    #   cli.run('version') #=> #<Git::CommandLineResult:0x00007f9b0c0b0e00
-    #
-    # @param env [Hash<String, String>] environment variables to set
-    # @param global_opts [Array<String>] global options to pass to git
-    # @param logger [Logger] the logger to use
-    #
-    def initialize(env, binary_path, global_opts, logger)
-      @env = env
-      @binary_path = binary_path
-      @global_opts = global_opts
-      @logger = logger
-    end
-
-    # @attribute [r] env
-    #
-    # Variables to set (or unset) in the git command's environment
-    #
-    # @example
-    #   env = { 'GIT_DIR' => '/path/to/git/dir' }
-    #   command_line = Git::CommandLine.new(env, '/usr/bin/git', [], Logger.new(STDOUT))
-    #   command_line.env #=> { 'GIT_DIR' => '/path/to/git/dir' }
-    #
-    # @return [Hash<String, String>]
-    #
-    # @see https://ruby-doc.org/3.2.1/Process.html#method-c-spawn Process.spawn
-    #   for details on how to set environment variables using the `env` parameter
-    #
-    attr_reader :env
-
-    # @attribute [r] binary_path
-    #
-    # The path to the command line binary to run
-    #
-    # @example
-    #   binary_path = '/usr/bin/git'
-    #   command_line = Git::CommandLine.new({}, binary_path, ['version'], Logger.new(STDOUT))
-    #   command_line.binary_path #=> '/usr/bin/git'
-    #
-    # @return [String]
-    #
-    attr_reader :binary_path
-
-    # @attribute [r] global_opts
-    #
-    # The global options to pass to git
-    #
-    # These are options that are passed to git before the command name and
-    # arguments. For example, in `git --git-dir /path/to/git/dir version`, the
-    # global options are %w[--git-dir /path/to/git/dir].
-    #
-    # @example
-    #   env = {}
-    #   global_opts = %w[--git-dir /path/to/git/dir]
-    #   logger = Logger.new(nil)
-    #   cli = CommandLine.new(env, '/usr/bin/git', global_opts, logger)
-    #   cli.global_opts #=> %w[--git-dir /path/to/git/dir]
-    #
-    # @return [Array<String>]
-    #
-    attr_reader :global_opts
-
-    # @attribute [r] logger
-    #
-    # The logger to use for logging git commands and results
-    #
-    # @example
-    #   env = {}
-    #   global_opts = %w[]
-    #   logger = Logger.new(STDOUT)
-    #   cli = CommandLine.new(env, '/usr/bin/git', global_opts, logger)
-    #   cli.logger == logger #=> true
-    #
-    # @return [Logger]
-    #
-    attr_reader :logger
-
-    # Execute a git command, wait for it to finish, and return the result
-    #
-    # Non-option the command line arguements to pass to git. If you collect
-    # the command line arguments in an array, make sure you splat the array
-    # into the parameter list.
-    #
-    # NORMALIZATION
-    #
-    # The command output is returned as a Unicde string containing the binary output
-    # from the command. If the binary output is not valid UTF-8, the output will
-    # cause problems because the encoding will be invalid.
-    #
-    # Normalization is a process that trys to convert the binary output to a valid
-    # UTF-8 string. It uses the `rchardet` gem to detect the encoding of the binary
-    # output and then converts it to UTF-8.
-    #
-    # Normalization is not enabled by default. Pass `normalize: true` to Git::CommandLine#run
-    # to enable it. Normalization will only be performed on stdout and only if the `out:`` option
-    # is nil or is a StringIO object. If the out: option is set to a file or other IO object,
-    # the normalize option will be ignored.
-    #
-    # @example Run a command and return the output
-    #   cli.run('version') #=> "git version 2.39.1\n"
-    #
-    # @example The args array should be splatted into the parameter list
-    #   args = %w[log -n 1 --oneline]
-    #   cli.run(*args) #=> "f5baa11 beginning of Ruby/Git project\n"
-    #
-    # @example Run a command and return the chomped output
-    #   cli.run('version', chomp: true) #=> "git version 2.39.1"
-    #
-    # @example Run a command and without normalizing the output
-    #   cli.run('version', normalize: false) #=> "git version 2.39.1\n"
-    #
-    # @example Capture stdout in a temporary file
-    #   require 'tempfile'
-    #   tempfile = Tempfile.create('git') do |file|
-    #     cli.run('version', out: file)
-    #     file.rewind
-    #     file.read #=> "git version 2.39.1\n"
-    #   end
-    #
-    # @example Capture stderr in a StringIO object
-    #   require 'stringio'
-    #   stderr = StringIO.new
-    #   begin
-    #     cli.run('log', 'nonexistent-branch', err: stderr)
-    #   rescue Git::FailedError => e
-    #     stderr.string #=> "unknown revision or path not in the working tree.\n"
-    #   end
-    #
-    # @param options_hash [Hash] the options to pass to the command
-    #
-    # @option options_hash [#write, nil] :out the object to write stdout to or nil to ignore stdout
-    #
-    #   If this is a 'StringIO' object, then `stdout_writer.string` will be returned.
-    #
-    #   In general, only specify a `stdout_writer` object when you want to redirect
-    #   stdout to a file or some other object that responds to `#write`. The default
-    #   behavior will return the output of the command.
-    #
-    # @option options_hash [#write, nil] :err the object to write stderr to or nil to ignore stderr
-    #
-    #   If this is a 'StringIO' object and `merged_output` is `true`, then
-    #   `stderr_writer.string` will be merged into the output returned by this method.
-    #
-    # @option options_hash [Boolean] :normalize whether to normalize the output of stdout and stderr
-    #
-    # @option options_hash [Boolean] :chomp whether to chomp both stdout and stderr output
-    #
-    # @option options_hash [Boolean] :merge whether to merge stdout and stderr in the string returned
-    #
-    # @option options_hash [String, nil] :chdir the directory to run the command in
-    #
-    # @option options_hash [Numeric, nil] :timeout the maximum seconds to wait for the command to complete
-    #
-    #   If timeout is zero, the timeout will not be enforced.
-    #
-    #   If the command times out, it is killed via a `SIGKILL` signal and `Git::TimeoutError` is raised.
-    #
-    #   If the command does not respond to SIGKILL, it will hang this method.
-    #
-    # @return [Git::CommandLineResult] the output of the command
-    #
-    #   This result of running the command.
-    #
-    # @raise [ArgumentError] if `args` is not an array of strings
-    #
-    # @raise [Git::SignaledError] if the command was terminated because of an uncaught signal
-    #
-    # @raise [Git::FailedError] if the command returned a non-zero exitstatus
-    #
-    # @raise [Git::ProcessIOError] if an exception was raised while collecting subprocess output
-    #
-    # @raise [Git::TimeoutError] if the command times out
-    #
-    def run(*, **options_hash)
-      options_hash = RUN_ARGS.merge(options_hash)
-      extra_options = options_hash.keys - RUN_ARGS.keys
-      raise ArgumentError, "Unknown options: #{extra_options.join(', ')}" if extra_options.any?
-
-      result = run_with_capture(*, **options_hash)
-      process_result(result, options_hash[:normalize], options_hash[:chomp], options_hash[:timeout])
-    end
-
-    # @return [Git::CommandLineResult] the result of running the command
-    #
-    # @api private
-    #
-    def run_with_capture(*args, **options_hash)
-      git_cmd = build_git_cmd(args)
-      options = run_with_capture_options(**options_hash)
-      ProcessExecuter.run_with_capture(env, *git_cmd, **options)
-    rescue ProcessExecuter::ProcessIOError => e
-      raise Git::ProcessIOError.new(e.message), cause: e.exception.cause
-    end
-
-    def run_with_capture_options(**options_hash)
-      chdir = options_hash[:chdir] || :not_set
-      timeout_after = options_hash[:timeout]
-      out = options_hash[:out]
-      err = options_hash[:err]
-      merge_output = options_hash[:merge] || false
-
-      { chdir:, timeout_after:, merge_output:, raise_errors: false }.tap do |options|
-        options[:out] = out unless out.nil?
-        options[:err] = err unless err.nil?
-      end
-    end
-
-    RUN_ARGS = {
-      normalize: false,
-      chomp: false,
-      merge: false,
-      out: nil,
-      err: nil,
-      chdir: nil,
-      timeout: nil
-    }.freeze
-
-    private
-
-    # Build the git command line from the available sources to send to `Process.spawn`
-    # @return [Array<String>]
-    # @api private
-    #
-    def build_git_cmd(args)
-      raise ArgumentError, 'The args array can not contain an array' if args.any?(Array)
-
-      [binary_path, *global_opts, *args].map(&:to_s)
-    end
-
-    # Process the result of the command and return a Git::CommandLineResult
-    #
-    # Post process output, log the command and result, and raise an error if the
-    # command failed.
-    #
-    # @param result [ProcessExecuter::Command::Result] the result it is a
-    #   Process::Status and include command, stdout, and stderr
-    #
-    # @param normalize [Boolean] whether to normalize the output of each writer
-    #
-    # @param chomp [Boolean] whether to chomp the output of each writer
-    #
-    # @param timeout [Numeric, nil] the maximum seconds to wait for the command to
-    #   complete
-    #
-    # @return [Git::CommandLineResult] the result of the command to return to the
-    #   caller
-    #
-    # @raise [Git::FailedError] if the command failed
-    #
-    # @raise [Git::SignaledError] if the command was signaled
-    #
-    # @raise [Git::TimeoutError] if the command times out
-    #
-    # @raise [Git::ProcessIOError] if an exception was raised while collecting
-    #   subprocess output
-    #
-    # @api private
-    #
-    def process_result(result, normalize, chomp, timeout)
-      command = result.command
-      processed_out, processed_err = post_process_output(result, normalize, chomp)
-      log_result(result, command, processed_out, processed_err)
-      command_line_result(command, result, processed_out, processed_err, timeout)
-    end
-
-    def log_result(result, command, processed_out, processed_err)
-      logger.info { "#{command} exited with status #{result}" }
-      logger.debug { "stdout:\n#{processed_out.inspect}\nstderr:\n#{processed_err.inspect}" }
-    end
-
-    def command_line_result(command, result, processed_out, processed_err, timeout)
-      Git::CommandLineResult.new(command, result, processed_out, processed_err).tap do |processed_result|
-        raise Git::TimeoutError.new(processed_result, timeout) if result.timeout?
-
-        raise Git::SignaledError, processed_result if result.signaled?
-
-        raise Git::FailedError, processed_result unless result.success?
-      end
-    end
-
-    # Post-process and return an array of raw output strings
-    #
-    # For each raw output string:
-    #
-    # * If normalize: is true, normalize the encoding by transcoding each line from
-    #   the detected encoding to UTF-8.
-    # * If chomp: is true chomp the output after normalization.
-    #
-    # Even if no post-processing is done based on the options, the strings returned
-    # are a copy of the raw output strings. The raw output strings are not modified.
-    #
-    # @param result [ProcessExecuter::ResultWithCapture] the command's output to post-process
-    #
-    # @param normalize [Boolean] whether to normalize the output of each writer
-    # @param chomp [Boolean] whether to chomp the output of each writer
-    #
-    # @return [Array<String>]
-    #
-    # @api private
-    #
-    def post_process_output(result, normalize, chomp)
-      [result.stdout, result.stderr].map do |raw_output|
-        output = raw_output.dup
-        output = output.lines.map { |l| Git::EncodingUtils.normalize_encoding(l) }.join if normalize
-        output.chomp! if chomp
-        output
-      end
-    end
+  # @example Buffered command via Git::CommandLine::Capturing
+  #   cli = Git::CommandLine::Capturing.new(
+  #     {}, '/usr/bin/git', [], Logger.new(nil)
+  #   )
+  #   result = cli.run('version')
+  #   result.stdout #=> "git version 2.39.1\n"
+  #
+  # @example Streaming command via Git::CommandLine::Streaming
+  #   cli = Git::CommandLine::Streaming.new(
+  #     {}, '/usr/bin/git', [], Logger.new(nil)
+  #   )
+  #   File.open('/tmp/blob', 'wb') do |f|
+  #     cli.run('cat-file', 'blob', sha, out: f)
+  #   end
+  #
+  # @see Git::CommandLine::Base
+  # @see Git::CommandLine::Result
+  #
+  module CommandLine
   end
 end
+
+require 'git/command_line/result'
+require 'git/command_line/base'
+require 'git/command_line/capturing'
+require 'git/command_line/streaming'