| | | | Pivotal Tracker Project
Makes it easy to spawn Ruby sub-processes with guaranteed exit status handling, capturing and/or suppressing combined STDOUT and STDERR streams, providing STDIN input, timeouts, and running via a pseudo terminal.
- Always raise an exception on unexpected exit status (i.e. return code or
$!
) - Combine and interleave STDOUT and STDERR streams into STDOUT (using Open3.popen2e) or PTY.spawn, so you don't have to worry about how to capture the output of both streams.
- Provide useful options for suppressing output and including output when an exception is raised due to an unexpected exit status
- Provide real-time streaming of combined STDOUT/STDERR streams in addition to returning full combined output as a string returned from the method and/or in the exception.
- Support passing input to the STDIN stream via string or StringIO.
- Support running commands in a pseudo terminal.
- Allow override of the expected exit status(es) (zero is expected by default)
- Provide short forms of all options for terse, concise usage.
- Any explicit support for process forks, multiple threads, or anything other than a single direct child process.
- Any support for separate handling of STDOUT and STDERR streams
There's many other libraries to make it easier to work with processes in Ruby (see the Resources section).
However, process_helper
was created because none of them made it easy to run processes while meeting all of these requirements (redundant details are repeated above in Goals section):
- Combine STDOUT/STDERR output streams interleaved chronologically as emitted
- Stream STDOUT/STDERR real-time while process is still running, in addition to returning full output as a string and/or in an exception
- Guarantee an exception is always raised on an unexpected exit status (and allow specification of multiple nonzero values as expected exit statuses)
- Can be used very concisely. I.e. All behavior can be invoked via a single mixed-in module with single public method call using terse options with sensible defaults, no need to use IO streams directly or have any blocks or local variables declared.
- Goals
- Non-Goals
- Why Yet Another Ruby Process Wrapper Library
- Installation
- Usage
- Options
- Warnings if failure output will be suppressed based on options
- Example Helper Script
- Version
- Contributing
- (Un)License
- Resources
Add this line to your application's Gemfile:
gem 'process_helper'
And then execute:
$ bundle
Or install it yourself as:
$ gem install process_helper
Or, if you want to avoid a Gem/RubyGems dependency, you can copy the process_helper.rb
file from the lib
folder directly into your project and require it:
# from a ruby file:
require_relative 'process_helper'
# from IRB:
require './process_helper'
# or put the directory containing it on the load path:
$LOAD_PATH.unshift('.')
require 'process_helper'
ProcessHelper
is a Ruby module you can include in any Ruby code,
and then call #process
to run a command, like this:
require 'process_helper'
include ProcessHelper
process('echo "Hello"')
By default, ProcessHelper will combine any STDERR and STDOUT, and output it to STDOUT,
and also return it as the result of the #process
method.
Expected Integer exit status, or array of expected Integer exit statuses.
Default value is [0]
.
An exception will be raised by the ProcessHelper#process
method if the
actual exit status of the processed command is not (one of) the
expected exit status(es).
Here's an example of expecting a nonzero failure exit status which matches the actual exit status
(the actual exit status of a failed ls
command will be 1 on OSX, 2 on Linux):
# The following will NOT raise an exception:
process('ls /does_not_exist', expected_exit_status: [1,2])
...but it WILL still print the output (in this case STDERR output from the failed ls
command) to STDOUT:
ls: /does_not_exist: No such file or directory
Here's a second example of expecting a nonzero failure exit status but the command succeeds:
# The following WILL raise an exception:
process('printf FAIL', expected_exit_status: 1)
Here's the output of the above example:
FAIL
ProcessHelper::UnexpectedExitStatusError: Command succeeded but was expected to fail, pid 62974 exit 0 (expected [1]). Command: `printf FAIL`. Command Output: "FAIL"
Boolean flag indicating whether output should be included in the message of the Exception (error)
which will be raised by the ProcessHelper#process
method if the command fails (has an unexpected exit status).
Here's an example of a failing command:
process('ls /does_not_exist', include_output_in_exception: true)
Here's the exception generated by the above example. Notice the "Command Output" with the "...No such file or directory" STDERR output of the failed command:
ProcessHelper::UnexpectedExitStatusError: Command failed, pid 64947 exit 1. Command: `ls /does_not_exist`. Command Output: "ls: /does_not_exist: No such file or directory
"
A String or StringIO object which will be supplied as standard input to the command.
The entire string will be read and piped to the command prior to outputting any output, but this behavior may be changed in the future to allow a separator character for processing input as "lines".
Valid values are true
and false
. Default value is false
.
When this option is true
, it will cause the command to be processed via
PTY.spawn
in a pseudo-terminal.
Some commands require a terminal, or "tty" to work properly, or to work at all. For example, some commands may not emit colored output unless they detect that they are running via a terminal.
It is important to note that this can change the behavior of a command.
For example, in most default Linux and OSX (BSD) terminals, newlines (\n
) in output
will be translated to carriage-return + newline (\r\n
). This will normally
have no effect, and can be controlled by the onlcr (-onlcr)
option
of the stty command.
Use stty -a
to get info on the current terminal.
Also, any input given to the command may be echoed to the output as well.
In some cases, the PTY will exit and return, even though the child process PID
for which it was spawned is still running. In this case, process_helper
will wait for the the child PID to exit. If it does not exit after
the specified :timeout
, or by default
60 seconds if :timeout
is unspecified, an exception will be raised.
Valid values are :always
, :error
, and :never
. Default value is :always
.
:always
will always print output to STDOUT:error
will only print output to STDOUT if command has an error - i.e. non-zero or unexpected exit status:never
will never print output to STDOUT
NOTE: This option will be changed in a future release.
Valid value is a float, e.g. 1.5
. Default value is nil/undefined.
- Controls how long
process_helper
will wait to read from a blocked IO stream before timing out (via IO.select). For example, invokingcat
with no arguments, which by default will continue accepting input until killed. - Will also kill long running processes which are not in blocked waiting on an IO stream read (i.e. kill process regardless of any IO state, not just via IO.selects timeout support).
- If undefined (default), there will be no timeout, and
process_helper
will hang if a process hangs while waiting to read from IO. - This option will also override how long a
:pseudo_terminal
will wait for a child PID to exit after the terminal has already exited.
The following changes are planned for this option:
- Add validation of value (enforced to be a float).
- Have both types of timeouts raise different and unique exception classes.
- Possibly have different option names to allow different timeout values for the two types of timeouts.
See https://www.pivotaltracker.com/story/show/93303096 for more details.
Valid value is a boolean. Default value is false.
- Prints command to STDOUT before it is executed.
- This is useful to have a "log" to know what commands are executing,
especially when using
puts_output: :error
orputs_output: :never
, and when the command might not finish immediately.
ProcessHelper will give you a warning if you pass a combination of options that would prevent ANY output from being printed or included in the Exception message if a command were to fail.
For example, in this case there is no output, and the expected exit status includes
the actual failure status which is returned, so the warning is printed, and the only
place the output will be seen is in the return value of the ProcessHelper#process
method:
> process('ls /does_not_exist', expected_exit_status: [1,2], puts_output: :never, include_output_in_exception: false)
WARNING: Check your ProcessHelper options - :puts_output is :never, and :include_output_in_exception is false, so all error output will be suppressed if process fails.
=> "ls: /does_not_exist: No such file or directory\n"
You can DRY up your usage of common process helper options by making a helper script like this that you can require:
setup_process_helper.rb
require 'process_helper'
include ProcessHelper
def process_without_output(cmd, options = {})
process(cmd, {out: :error, include_output_in_exception: false}.merge(default_options_for_process_helper).merge(options))
end
def process_with_output(cmd, options = {})
process(cmd, default_options_for_process_helper.merge(options))
end
def default_options_for_process_helper
# IMPORTANT NOTE: Running under a pseudo-terminal (pty) gives coloring and other
# benefits, but will result in end-of-line being "\r\n" instead of just "\n".
# Be aware of this if attempting to split return value into lines.
{log: true, pty: true}
end
You can see the version of ProcessHelper in the ProcessHelper::VERSION
constant:
ProcessHelper::VERSION
=> "0.0.1"
- Fork it ( https://github.com/thewoolleyman/process_helper/fork )
- Create your feature branch (
git checkout -b my-new-feature
) - Commit your changes (
git commit -am 'Add some feature'
) - Push to the branch (
git push origin my-new-feature
) - If you are awesome, use
git rebase --interactive
to ensure you have a single atomic commit on your branch. - Create a new Pull Request
process_helper
is (un)licensed under the unlicense. So, feel free to copy it into your project (it's usable as a single required file and module), use it, abuse it, do what you will, no attribution required. :)
Other Ruby Process tools/libraries:
- a comprehensive StackOverflow article describing, in detail, the myriad methods of launching Ruby subprocesses.
- open4 - a solid and useful library - the main thing I missed in it was easily combining real-time streaming interleaved STDOUT/STDERR streams
- open4 on ruby toolbox - see if there's some useful higher-level gem that depends on it and gives you functionality you may need
- A great series of blog posts by Devver:
- A gist exploring ruby PTY behavior
Some notes on why you should use printf over echo: