POSIX::Spawn::Child includes logic for executing child processes and reading/writing from their standard input, output, and error streams. It's designed to take all input in a single string and provides all output (stderr and stdout) as single strings and is therefore not well-suited to streaming large quantities of data in and out of commands.
Create and run a process to completion:
>> child = POSIX::Spawn::Child.new('git', '--help')
Retrieve stdout or stderr output:
>> child.out => "usage: git [--version] [--exec-path[=GIT_EXEC_PATH]]\n ..." >> child.err => ""
Check process exit status information:
>> child.status => #<Process::Status: pid=80718,exited(0)>
To write data on the new process’s stdin immediately after spawning:
>> child = POSIX::Spawn::Child.new('bc', :input => '40 + 2') >> child.out "42\n"
To access output from the process even if an exception was raised:
>> child = POSIX::Spawn::Child.build('git', 'log', :max => 1000) >> begin ?> child.exec! ?> rescue POSIX::Spawn::MaximumOutputExceeded ?> # just so you know ?> end >> child.out "... first 1000 characters of log output ..."
Q: Why use POSIX::Spawn::Child instead of popen3, hand rolled fork/exec code, or Process::spawn?
It’s more efficient than popen3 and provides meaningful process hierarchies because it performs a single fork/exec. (popen3 double forks to avoid needing to collect the exit status and also calls Process::detach which creates a Ruby Thread!!!!).
It handles all max pipe buffer (PIPE_BUF) hang cases when reading and writing semi-large amounts of data. This is non-trivial to implement correctly and must be accounted for with popen3, spawn, or hand rolled fork/exec code.
It’s more portable than hand rolled pipe, fork, exec code because fork(2) and exec aren’t available on all platforms. In those cases, POSIX::Spawn::Child falls back to using whatever janky substitutes the platform provides.
Maximum buffer size for reading
Set up a new process to spawn, but do not actually spawn it.
Invoke this just like the normal constructor to set up a process to be run. Call `exec!` to actually run the child process, send the input, read the output, and wait for completion. Use this alternative way of constructing a POSIX::Spawn::Child if you want to read any partial output from the child process even after an exception.
child = POSIX::Spawn::Child.build(... arguments ...) child.exec!
The arguments are the same as the regular constructor.
Returns a new Child instance but does not run the underlying process.
# File lib/posix/spawn/child.rb, line 109 def self.build(*args) options = if args[-1].respond_to?(:to_hash) args.pop.to_hash else {} end new(*args, { :noexec => true }.merge(options)) end
Spawn a new process, write all input and read all output, and wait for the program to exit. Supports the standard spawn interface as described in the POSIX::Spawn module documentation:
new([env], command, [argv1, ...], [options])
The following options are supported in addition to the standard POSIX::Spawn options:
:input => str Write str to the new process's standard input. :timeout => int Maximum number of seconds to allow the process to execute before aborting with a TimeoutExceeded exception. :max => total Maximum number of bytes of output to allow the process to generate before aborting with a MaximumOutputExceeded exception.
Returns a new Child instance whose underlying process has already executed to completion. The out, err, and status attributes are immediately available.
# File lib/posix/spawn/child.rb, line 84 def initialize(*args) @env, @argv, options = extract_process_spawn_arguments(*args) @options = options.dup @input = @options.delete(:input) @timeout = @options.delete(:timeout) @max = @options.delete(:max) @options.delete(:chdir) if @options[:chdir].nil? exec! if !@options.delete(:noexec) end
Execute command, write input, and read output. This is called immediately when a new instance of this object is created, or can be called explicitly when creating the Child via `build`.
# File lib/posix/spawn/child.rb, line 139 def exec! # spawn the process and hook up the pipes pid, stdin, stdout, stderr = popen4(@env, *(@argv + [@options])) # async read from all streams into buffers read_and_write(@input, stdin, stdout, stderr, @timeout, @max) # grab exit status @status = waitpid(pid) rescue Object => boom [stdin, stdout, stderr].each { |fd| fd.close rescue nil } if @status.nil? ::Process.kill('TERM', pid) rescue nil @status = waitpid(pid) rescue nil end raise ensure # let's be absolutely certain these are closed [stdin, stdout, stderr].each { |fd| fd.close rescue nil } end
Generated with the Darkfish Rdoc Generator 2.