Awesome
subx: A Data Structure for Results of Subprocesses
SubprocessResult
This library gives you a data structure called
SubprocessResult
. It combines stdout, stderr and ret.
This is handy if you do "one shot" calling of subprocesses.
Since Python 3.5 the subprocess module has the method run()
which returns the datastructure CompletedProcess. This means
the subx
library is not needed any more.
Why?
If subx fails, you get a meaningful exception message that helps you. You see the first bytes of stdout and stderr. This more convinient than the standard library subprocess.
Gracefull handling of timeouts. You get a meaningful error message, even if timeout happens: You see stdin and stdout which were emitted until the timeout occurred.
Passing in a string as stdin of a subprocess is easy. Just use the kwarg data
.
Examples
The method call()
returns an instance of SubprocessResult
result = subx.call(['date'])
Just replace subprocess.check_call(cmd)
with subx.call(cmd)
and you get all you want plus a helpful
exception messages.
Or replace subprocess.check_output(cmd)
with
subx.call(cmd).stdout
.
If you want to ignore the status code like shell scripts do, and you want to see the head of stdout/stderr you can use this:
logging.info(subx.call(assert_zero_exit_status=False))
This will use repr(result). Which looks like roughly this:
<SubprocessResult cmd='my-command' ret=0 stdout='....' stderr='...'>
By default subx.call()
raises subx.SubprocessError
if the
exit status is non-zero.
If you want to handle non-zero exist status yourself, then you can do it like this:
result = subx.call(cmd, assert_zero_exit_status=False)
if result.ret:
print('Failed: {}\n{}\n{}'.format(result.cmd, result.stderr, result.stdout))
...
Method subx.call()
Arguments:
call(cmd, data=None, assert_zero_exit_status=True, warn_on_non_zero_exist_status=False, **kwargs)
data: String which gets send to stdin of the subprocess.
assert_zero_exit_status: raise an exception if exist status is non-zero?
warn_on_non_zero_exist_status: warn on non zero exit status?
Returns: SubprocessResult instance
Class SubprocessResult
The class SubprocessResult
has the following attributes:
- stdout
- stderr
- ret (exit status)
- cmd
Not suited for ...
This library is not useful if you want to read streamed data from your subprocess. But the library is useful, if you want to stream data to your subprocess.
Install
Install from pypi:
pip install subx
subprocess.check_output() vs subx.call()
Look, compare, think and decide what message helps your more.
subprocess.check_output():
CalledProcessError: Command '['cat', 'some-file']' returned non-zero exit status 1
sub.call():
SubprocessError: Command '['cat', 'some-file']' returned non-zero exit status 1:
stdout='' stderr='cat: some-file: No such file or directory'
... especially if the code fails in a production environment where reproducing the error is not easy, subx can call help you to spot the source of the failure. In above case you see "No such file or directory" which gives you a hint about the root cause.
Development Install on Python3
Install subx for development on Python3:
python3 -m venv subx-py3env
cd subx-py3env
. ./bin/activate
pip install --upgrade pip
pip install -e git+https://github.com/guettli/subx.git#egg=subx
Development Testing
Testing:
pip install -r src/subx/requirements.txt
cd src/subx
pytest # all test ok?
pyCharm src/subx/...
pytest # all test still ok?
.... I am waiting for your pull request :-)
Python 2
Python 2 is not supported any more. Please use version
2019.36.0
if you need it.
Reflections
Creating subprocesses should be avoided. It is slow and error prone. In the past you could not avoid it. Today there is a library for almost everything, and that's great.
More from Thomas Güttler
You can find more of me on my Working-out-Loud list.