git 4.3.2 → 5.0.0.beta.1

data/lib/git/command_line/capturing.rb

@@ -0,0 +1,249 @@
+# frozen_string_literal: true
+
+require 'git/command_line'
+
+module Git
+  module CommandLine
+    # Executes a git command and captures both stdout and stderr in memory
+    #
+    # {Git::CommandLine::Capturing} is the buffering strategy: it calls
+    # `ProcessExecuter.run_with_capture`, which reads all subprocess output into
+    # `String` objects before returning.  Use this class (via
+    # {Git::Lib#command_capturing}) for the vast majority of git subcommands whose
+    # output fits comfortably in memory.
+    #
+    # {Git::CommandLine::Streaming} is the complementary strategy for commands
+    # (such as `cat-file -p <blob>`) whose stdout may be too large to buffer.
+    #
+    # @example
+    #   capturing = Git::CommandLine::Capturing.new(
+    #     {}, '/usr/bin/git', %w[--git-dir /repo/.git], Logger.new($stdout)
+    #   )
+    #   result = capturing.run('log', '--oneline', '-5')
+    #   result.stdout   # => "abc1234 Initial commit\n..."
+    #   result.stderr   # => ""
+    #
+    # @see Git::Lib#command_capturing
+    #
+    # @see Git::CommandLine::Streaming
+    #
+    class Capturing < 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: {},
+        normalize: false,
+        chomp: false,
+        merge: false
+      }.freeze
+
+      # Execute a git command, capture stdout and stderr, and return the result
+      #
+      # Non-option command-line arguments to pass to git. If you collect the
+      # arguments in an array, splat the array into the parameter list.
+      #
+      # NORMALIZATION
+      #
+      # The command output is returned as a Unicode 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 tries 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 enable
+      # it. When enabled, normalization is applied to both stdout and stderr in
+      # the returned result object, regardless of the `out:` or `err:` options.
+      # Only the captured in-memory strings are normalized; any external IO you
+      # provide will receive the raw subprocess output.
+      #
+      # @example Run a command and return the output
+      #   result = capturing.run('version')
+      #   result.stdout #=> "git version 2.39.1\n"
+      #
+      # @example The args array should be splatted into the parameter list
+      #   args = %w[log -n 1 --oneline]
+      #   result = capturing.run(*args)
+      #   result.stdout #=> "f5baa11 beginning of Ruby/Git project\n"
+      #
+      # @example Run a command and return the chomped output
+      #   result = capturing.run('version', chomp: true)
+      #   result.stdout #=> "git version 2.39.1"
+      #
+      # @example Run a command without normalizing the output
+      #   capturing.run('version', normalize: false) #=> "git version 2.39.1\n"
+      #
+      # @example Capture stdout in a temporary file
+      #   require 'tempfile'
+      #   Tempfile.create('git') do |file|
+      #     capturing.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
+      #     capturing.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 [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 object to write stdout to, or
+      #   nil to capture stdout in the returned result.
+      #
+      #   If this is a `StringIO` object, `stdout_writer.string` will be returned.
+      #
+      #   In general, only specify a `stdout_writer` when you want to redirect
+      #   stdout to a file or other `#write`-responding object.  The default
+      #   behaviour returns the command output.
+      #
+      # @option options_hash [#write, nil] :err the object to write stderr to, or
+      #   nil to capture stderr in the returned result.
+      #
+      # @option options_hash [Boolean] :normalize (false) whether to normalize the
+      #   encoding of stdout and stderr output
+      #
+      # @option options_hash [Boolean] :chomp (false) whether to chomp both stdout
+      #   and stderr output
+      #
+      # @option options_hash [Boolean] :merge (false) whether to merge stdout and
+      #   stderr in the returned string
+      #
+      # @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
+      #
+      # @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)
+
+        result = execute(*, **options)
+        process_result(result, options)
+      end
+
+      private
+
+      # @return [ProcessExecuter::ResultWithCapture] the process result with captured output
+      #
+      # @api private
+      def execute(*args, **options_hash)
+        git_cmd = build_git_cmd(args)
+        options = execute_options(**options_hash)
+        run_process_executer do
+          ProcessExecuter.run_with_capture(merged_env(options_hash), *git_cmd, **options)
+        end
+      end
+
+      # Build the ProcessExecuter options hash for a capturing run
+      #
+      # @return [Hash]
+      #
+      # @api private
+      def execute_options(**options_hash)
+        chdir = options_hash[:chdir] || :not_set
+        timeout_after = options_hash[:timeout]
+        merge_output = options_hash[:merge] || false
+
+        { chdir:, timeout_after:, merge_output:, raise_errors: false }.tap do |options|
+          redirect_options(options_hash).each { |k, v| options[k] = v }
+        end
+      end
+
+      # Extract non-nil redirect options (`:in`, `:out`, `:err`) from options_hash
+      #
+      # @return [Hash]
+      #
+      # @api private
+      def redirect_options(options_hash)
+        %i[in out err].filter_map do |key|
+          val = options_hash[key]
+          [key, val] unless val.nil?
+        end.to_h
+      end
+
+      # Post-process and return the stdout/stderr strings from the captured result,
+      # then log and raise on failure if required.
+      #
+      # @param result [ProcessExecuter::ResultWithCapture] the raw result
+      #
+      # @param options [Hash] the merged run options
+      #
+      # @return [Git::CommandLineResult]
+      #
+      # @raise [Git::FailedError] if the command failed and raise_on_failure is true
+      #
+      # @raise [Git::SignaledError] if the command was signaled
+      #
+      # @raise [Git::TimeoutError] if the command timed out
+      #
+      # @api private
+      def process_result(result, options)
+        command = result.command
+        processed_out, processed_err = post_process_output(result, options[:normalize], options[:chomp])
+        log_result(result, command, processed_out, processed_err)
+        command_line_result(
+          command, result, processed_out, processed_err, options[:timeout], options[:raise_on_failure]
+        )
+      end
+
+      # Normalize and/or chomp the raw stdout and stderr strings.
+      #
+      # @param result [ProcessExecuter::ResultWithCapture] the raw result
+      #
+      # @param normalize [Boolean]
+      #
+      # @param chomp [Boolean]
+      #
+      # @return [Array<String>] two-element array: [processed_stdout, processed_stderr]
+      #
+      # @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
+    end
+  end
+end

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