Class: Toys::Utils::Exec

Inherits:
Object
  • Object
show all
Defined in:
lib/toys/utils/exec.rb

Overview

A service that executes subprocesses.

This service provides a convenient interface for controlling spawned processes and their streams. It also provides shortcuts for common cases such as invoking Ruby in a subprocess or capturing output in a string.

This class is not loaded by default. Before using it directly, you should require "toys/utils/exec"

Controlling processes

A process can be started in the foreground or the background. If you start a foreground process, it will "take over" your standard input and output streams by default, and it will keep control until it completes. If you start a background process, its streams will be redirected to null by default, and control will be returned to you immediately.

When a process is running, you can control it using a Controller object. Use a controller to interact with the process's input and output streams, send it signals, or wait for it to complete.

When running a process in the foreground, the controller will be yielded to an optional block. For example, the following code starts a process in the foreground and passes its output stream to a controller.

exec_service.exec(["git", "init"], out: :controller) do |controller|
  loop do
    line = controller.out.gets
    break if line.nil?
    puts "Got line: #{line}"
  end
end

When running a process in the background, the controller is returned from the method that starts the process:

controller = exec_service.exec(["git", "init"], background: true)

Stream handling

By default, subprocess streams are connected to the corresponding streams in the parent process. You can change this behavior, redirecting streams or providing ways to control them, using the :in, :out, and :err options.

Three general strategies are available for custom stream handling. First, you can redirect to other streams such as files, IO objects, or Ruby strings. Some of these options map directly to options provided by the Process#spawn method. Second, you can use a controller to manipulate the streams programmatically. Third, you can capture output stream data and make it available in the result.

Following is a full list of the stream handling options, along with how to specify them using the :in, :out, and :err options.

  • Inherit parent stream: You can inherit the corresponding stream in the parent process by passing :inherit as the option value. This is the default if the subprocess is not run in the background.

  • Redirect to null: You can redirect to a null stream by passing :null as the option value. This connects to a stream that is not closed but contains no data, i.e. /dev/null on unix systems. This is the default if the subprocess is run in the background.

  • Close the stream: You can close the stream by passing :close as the option value. This is the same as passing :close to Process#spawn.

  • Redirect to a file: You can redirect to a file. This reads from an existing file when connected to :in, and creates or appends to a file when connected to :out or :err. To specify a file, use the setting [:file, "/path/to/file"]. You can also, when writing a file, append an optional mode and permission code to the array. For example, [:file, "/path/to/file", "a", 0644].

  • Redirect to an IO object: You can redirect to an IO object in the parent process, by passing the IO object as the option value. You can use any IO object. For example, you could connect the child's output to the parent's error using out: $stderr, or you could connect to an existing File stream. Unlike Process#spawn, this works for IO objects that do not have a corresponding file descriptor (such as StringIO objects). In such a case, a thread will be spawned to pipe the IO data through to the child process.

  • Redirect to a pipe: You can redirect to a pipe created using IO.pipe (i.e. a two-element array of read and write IO objects) by passing the array as the option value. This will connect the appropriate IO (either read or write), and close it in the parent. Thus, you can connect only one process to each end. If you want more direct control over IO closing behavior, pass the IO object (i.e. the element of the pipe array) directly.

  • Combine with another child stream: You can redirect one child output stream to another, to combine them. To merge the child's error stream into its output stream, use err: [:child, :out].

  • Read from a string: You can pass a string to the input stream by setting [:string, "the string"]. This works only for :in.

  • Capture output stream: You can capture a stream and make it available on the Result object, using the setting :capture. This works only for the :out and :err streams.

  • Use the controller: You can hook a stream to the controller using the setting :controller. You can then manipulate the stream via the controller. If you pass a block to #exec, it yields the Controller, giving you access to streams.

  • Make copies of an output stream: You can "tee," or duplicate the :out or :err stream and redirect those copies to various destinations. To specify a tee, use the setting [:tee, ...] where the additional array elements include two or more of the following. See the corresponding documentation above for more detail.

    • :inherit to direct to the parent process's stream.
    • :capture to capture the stream and store it in the result.
    • :controller to direct the stream to the controller.
    • [:file, "/path/to/file"] to write to a file.
    • An IO or StringIO object.
    • An array of two IO objects representing a pipe

    Additionally, the last element of the array can be a hash of options. Supported options include:

    • :buffer_size The size of the memory buffer for each element of the tee. Larger buffers may allow higher throughput. The default is 65536.

Result handling

A subprocess result is represented by a Result object, which includes the exit code, the content of any captured output streams, and any exeption raised when attempting to run the process. When you run a process in the foreground, the method will return a result object. When you run a process in the background, you can obtain the result from the controller once the process completes.

The following example demonstrates running a process in the foreground and getting the exit code:

result = exec_service.exec(["git", "init"])
puts "exit code: #{result.exit_code}"

The following example demonstrates starting a process in the background, waiting for it to complete, and getting its exit code:

controller = exec_service.exec(["git", "init"], background: true)
result = controller.result(timeout: 1.0)
if result
  puts "exit code: #{result.exit_code}"
else
  puts "timed out"
end

You can also provide a callback that is executed once a process completes. For example:

my_callback = proc do |result|
  puts "exit code: #{result.exit_code}"
end
exec_service.exec(["git", "init"], result_callback: my_callback)

Configuration options

A variety of options can be used to control subprocesses. These can be provided to any method that starts a subprocess. Youc an also set defaults by calling #configure_defaults.

Options that affect the behavior of subprocesses:

  • :env (Hash) Environment variables to pass to the subprocess. Keys represent variable names and should be strings. Values should be either strings or nil, which unsets the variable.

  • :background (Boolean) Runs the process in the background if true.

  • :result_callback (Proc) Called and passed the result object when the subprocess exits.

Options for connecting input and output streams. See the section above on stream handling for info on the values that can be passed.

  • :in Connects the input stream of the subprocess. See the section on stream handling.

  • :out Connects the standard output stream of the subprocess. See the section on stream handling.

  • :err Connects the standard error stream of the subprocess. See the section on stream handling.

Options related to logging and reporting:

  • :logger (Logger) Logger to use for logging the actual command. If not present, the command is not logged.

  • :log_level (Integer,false) Level for logging the actual command. Defaults to Logger::INFO if not present. You can also pass false to disable logging of the command.

  • :log_cmd (String) The string logged for the actual command. Defaults to the inspect representation of the command.

  • :name (Object) An optional object that can be used to identify this subprocess. It is available in the controller and result objects.

In addition, the following options recognized by Process#spawn are supported.

  • :chdir (String) Set the working directory for the command.

  • :close_others (Boolean) Whether to close non-redirected non-standard file descriptors.

  • :new_pgroup (Boolean) Create new process group (Windows only).

  • :pgroup (Integer,true,nil) The process group setting.

  • :umask (Integer) Umask setting for the new process.

  • :unsetenv_others (Boolean) Clear environment variables except those explicitly set.

Any other option key will result in an ArgumentError.

Defined Under Namespace

Classes: Controller, Result

Instance Method Summary collapse

Constructor Details

#initialize(**opts, &block) ⇒ Exec

Create an exec service.

Parameters:

  • block (Proc) —

    A block that is called if a key is not found. It is passed the unknown key, and expected to return a default value (which can be nil).

  • opts (keywords) —

    Initial default options. See Toys::Utils::Exec for a description of the options.



247
248
249
 
# File 'lib/toys/utils/exec.rb', line 247

def initialize(**opts, &block)
  @default_opts = Opts.new(&block).add(opts)
end

Instance Method Details

#capture(cmd, **opts) {|controller| ... } ⇒ String

Execute a command. The command can be given as a single string to pass to a shell, or an array of strings indicating a posix command.

Captures standard out and returns it as a string. Cannot be run in the background.

If a block is provided, a Controller will be yielded to it.

Parameters:

  • cmd (String, Array<String>) —

    The command to execute.

  • opts (keywords) —

    The command options. See the section on configuration options in the Toys::Utils::Exec class docs.

Yield Parameters:

Returns:

  • (String) —

    What was written to standard out.



364
365
366
367
 
# File 'lib/toys/utils/exec.rb', line 364

def capture(cmd, **opts, &block)
  opts = opts.merge(out: :capture, background: false)
  exec(cmd, **opts, &block).captured_out
end

#capture_proc(func, **opts) {|controller| ... } ⇒ String

Execute a proc in a fork.

Captures standard out and returns it as a string. Cannot be run in the background.

If a block is provided, a Controller will be yielded to it.

Parameters:

  • func (Proc) —

    The proc to call.

  • opts (keywords) —

    The command options. See the section on configuration options in the Toys::Utils::Exec class docs.

Yield Parameters:

Returns:

  • (String) —

    What was written to standard out.



408
409
410
411
 
# File 'lib/toys/utils/exec.rb', line 408

def capture_proc(func, **opts, &block)
  opts = opts.merge(out: :capture, background: false)
  exec_proc(func, **opts, &block).captured_out
end

#capture_ruby(args, **opts) {|controller| ... } ⇒ String

Spawn a ruby process and pass the given arguments to it.

Captures standard out and returns it as a string. Cannot be run in the background.

If a block is provided, a Controller will be yielded to it.

Parameters:

  • args (String, Array<String>) —

    The arguments to ruby.

  • opts (keywords) —

    The command options. See the section on configuration options in the Toys::Utils::Exec class docs.

Yield Parameters:

Returns:

  • (String) —

    What was written to standard out.



386
387
388
389
 
# File 'lib/toys/utils/exec.rb', line 386

def capture_ruby(args, **opts, &block)
  opts = opts.merge(out: :capture, background: false)
  ruby(args, **opts, &block).captured_out
end

#configure_defaults(**opts) ⇒ self

Set default options. See Toys::Utils::Exec for a description of the options.

Parameters:

  • opts (keywords) —

    New default options to set

Returns:

  • (self) 


258
259
260
261
 
# File 'lib/toys/utils/exec.rb', line 258

def configure_defaults(**opts)
  @default_opts.add(opts)
  self
end

#exec(cmd, **opts) {|controller| ... } ⇒ Toys::Utils::Exec::Controller, Toys::Utils::Exec::Result

Execute a command. The command can be given as a single string to pass to a shell, or an array of strings indicating a posix command.

If the process is not set to run in the background, and a block is provided, a Controller will be yielded to it.

Parameters:

  • cmd (String, Array<String>) —

    The command to execute.

  • opts (keywords) —

    The command options. See the section on configuration options in the Toys::Utils::Exec class docs.

Yield Parameters:

Returns:



281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
 
# File 'lib/toys/utils/exec.rb', line 281

def exec(cmd, **opts, &block)
  exec_opts = Opts.new(@default_opts).add(opts)
  spawn_cmd =
    if cmd.is_a?(::Array)
      if cmd.size > 1
        binary = canonical_binary_spec(cmd.first, exec_opts)
        [binary] + cmd[1..-1].map(&:to_s)
      else
        [canonical_binary_spec(Array(cmd.first), exec_opts)]
      end
    else
      [cmd.to_s]
    end
  executor = Executor.new(exec_opts, spawn_cmd, block)
  executor.execute
end

#exec_proc(func, **opts) {|controller| ... } ⇒ Toys::Utils::Exec::Controller, Toys::Utils::Exec::Result

Execute a proc in a fork.

If the process is not set to run in the background, and a block is provided, a Controller will be yielded to it.

Parameters:

  • func (Proc) —

    The proc to call.

  • opts (keywords) —

    The command options. See the section on configuration options in the Toys::Utils::Exec class docs.

Yield Parameters:

Returns:



340
341
342
343
344
 
# File 'lib/toys/utils/exec.rb', line 340

def exec_proc(func, **opts, &block)
  exec_opts = Opts.new(@default_opts).add(opts)
  executor = Executor.new(exec_opts, func, block)
  executor.execute
end

#exec_ruby(args, **opts) {|controller| ... } ⇒ Toys::Utils::Exec::Controller, Toys::Utils::Exec::Result Also known as: ruby

Spawn a ruby process and pass the given arguments to it.

If the process is not set to run in the background, and a block is provided, a Controller will be yielded to it.

Parameters:

  • args (String, Array<String>) —

    The arguments to ruby.

  • opts (keywords) —

    The command options. See the section on configuration options in the Toys::Utils::Exec class docs.

Yield Parameters:

Returns:



315
316
317
318
319
320
 
# File 'lib/toys/utils/exec.rb', line 315

def exec_ruby(args, **opts, &block)
  cmd = args.is_a?(::Array) ? [::RbConfig.ruby] + args : "#{::RbConfig.ruby} #{args}"
  log_cmd = "exec ruby: #{args.inspect}"
  opts = {argv0: "ruby", log_cmd: log_cmd}.merge(opts)
  exec(cmd, **opts, &block)
end

#sh(cmd, **opts) {|controller| ... } ⇒ Integer

Execute the given string in a shell. Returns the exit code. Cannot be run in the background.

If a block is provided, a Controller will be yielded to it.

Parameters:

  • cmd (String) —

    The shell command to execute.

  • opts (keywords) —

    The command options. See the section on configuration options in the Toys::Utils::Exec class docs.

Yield Parameters:

Returns:

  • (Integer) —

    The exit code



428
429
430
431
 
# File 'lib/toys/utils/exec.rb', line 428

def sh(cmd, **opts, &block)
  opts = opts.merge(background: false)
  exec(cmd, **opts, &block).exit_code || -1
end