use function name in the log output if possible

[parkouss]

Hey there, first thanks for the redo package!

So basically when using the retry function, We have this kind of lines:

INFO | retry: Calling <function _download at 0x7fad39b54b18> with args: (), kwargs: {}, attempt #2

It would be good to just print "Calling _download ..." - i.e. the name of the function (if available) instead of the function object.

What do you think @bhearsum ? I can create a PR for that.

[bhearsum]

If you're only talking about getting right of the function identity part, I agree. I do think it's important to continue printing the args and kwargs, though maybe only at DEBUG level logging? Could be something like:
INFO | retry: Calling _download, attempt #2
DEBUG | retry: args: (), kwargs: {}

[parkouss]

Yes, I was only talking about the function part. But since I was at it, I made a minor change to only display args/kwargs if we have at least one of them. I just created the PR #36.

Seems to work well on python 2.7.10 and 3.5.0.

(venv)dev/redo [ retry -v -s 1 cat something
cat: something: Aucun fichier ou dossier de ce type
INFO:redo:retry: calling check_call with args: (['cat', 'something'],), kwargs: {}, attempt #2
cat: something: Aucun fichier ou dossier de ce type
INFO:redo:retry: calling check_call with args: (['cat', 'something'],), kwargs: {}, attempt #3
cat: something: Aucun fichier ou dossier de ce type
INFO:redo:retry: calling check_call with args: (['cat', 'something'],), kwargs: {}, attempt #4
cat: something: Aucun fichier ou dossier de ce type
INFO:redo:retry: calling check_call with args: (['cat', 'something'],), kwargs: {}, attempt #5
cat: something: Aucun fichier ou dossier de ce type
INFO:redo:retry: Giving up on check_call
ERROR:redo.cmd:Unable to run command after 5 attempts
Traceback (most recent call last):
  File "/home/jp/dev/redo/redo/cmd.py", line 44, in main
    r_check_call(args.cmd)
  File "/home/jp/dev/redo/redo/__init__.py", line 206, in _retriable_wrapper
    **retry_kwargs)
  File "/home/jp/dev/redo/redo/__init__.py", line 161, in retry
    return action(*args, **kwargs)
  File "/usr/lib64/python2.7/subprocess.py", line 540, in check_call
    raise CalledProcessError(retcode, cmd)
CalledProcessError: Command '['cat', 'something']' returned non-zero exit status 1

I also tried with a lambda function (without a name attribute, and broken on purpose):

DEBUG:redo:attempt 5/5
INFO:redo:retry: calling <lambda>, attempt #5
DEBUG:redo:retry: Caught exception: 
Traceback (most recent call last):
  File "redo/__init__.py", line 161, in retry
    return action(*args, **kwargs)
TypeError: <lambda>() takes exactly 1 argument (0 given)
INFO:redo:retry: Giving up on <lambda>
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "redo/__init__.py", line 161, in retry
    return action(*args, **kwargs)
TypeError: <lambda>() takes exactly 1 argument (0 given)

[bhearsum]

Breaking support for lambda's is a regression, so that case will need to be handled. It's probably OK to fallback to printing the function identity instead of the name in that case?

[parkouss]

Sorry that was not explicit enough, this is what the patch does already:

https://github.com/bhearsum/redo/pull/36/files#diff-9adb60c4e52347a6dae527b7cd3c6ee6R141

The example above just show that it works (what is inside the lambda test function is broken so it is retried, that's all).

[bhearsum]

Ah, I see! Sorry, I hadn't noticed the PR!

[Deleted user]

This was fixed in #36