Hi dev-team, I'm reopening the discussion of PEP 473, which was closed due to: *The steering council felt the PEP was too broad and not focused enough. Discussions about adding more attributes to built-in exceptions can continue on the issue tracker on a per-exception basis (and obviously here for any broader points, e.g. performance implications as I know that has come up before when the idea of storing relevant objects on exceptions).* Before the PEP was rejected by the SC, I had submitted https://github.com/python/cpython/pull/6271 and the last response said: *You may propose a newly revised PEP which overcomes the last SC feedback.But without the accepted PEP, this PR can not be merged.* Please clarify how can I proceed with adding some of the specified fields to the mentioned exceptions. Should I create a new PEP as suggested in the PR or should we create individual BPO issues for each exception modification. -- Sebastian Kreft
Context: PEP 473 "Adding structured data to built-in exceptions" proposed in 2014. https://www.python.org/dev/peps/pep-0473/ Abstract: "Exceptions like AttributeError, IndexError, KeyError, LookupError, NameError, TypeError, and ValueError do not provide all information required by programmers to debug and better understand what caused them. Furthermore, in some cases the messages even have slightly different formats, which makes it really difficult for tools to automatically provide additional information to diagnose the problem. To tackle the former and to lay ground for the latter, it is proposed to expand these exceptions so to hold both the offending and affected entities." For example the PEP proposes to add "target" and "attribute" attributes to AttributeError. PS: Please give at least the PEP title for me when you mention a PEP only by its number, I'm unable to remember what are the 500+ PEP :-( Victor On Sat, Mar 13, 2021 at 7:49 PM Sebastian Kreft <skreft@gmail.com> wrote:
Hi dev-team, I'm reopening the discussion of PEP 473, which was closed due to:
The steering council felt the PEP was too broad and not focused enough. Discussions about adding more attributes to built-in exceptions can continue on the issue tracker on a per-exception basis (and obviously here for any broader points, e.g. performance implications as I know that has come up before when the idea of storing relevant objects on exceptions).
Before the PEP was rejected by the SC, I had submitted https://github.com/python/cpython/pull/6271 and the last response said:
You may propose a newly revised PEP which overcomes the last SC feedback. But without the accepted PEP, this PR can not be merged.
Please clarify how can I proceed with adding some of the specified fields to the mentioned exceptions. Should I create a new PEP as suggested in the PR or should we create individual BPO issues for each exception modification.
-- Sebastian Kreft _______________________________________________ Python-Dev mailing list -- python-dev@python.org To unsubscribe send an email to python-dev-leave@python.org https://mail.python.org/mailman3/lists/python-dev.python.org/ Message archived at https://mail.python.org/archives/list/python-dev@python.org/message/4XVFFNBH... Code of Conduct: http://python.org/psf/codeofconduct/
-- Night gathers, and now my watch begins. It shall not end until my death.
Since tweaking the built-in exceptions is so far-reaching, probably at least discussing each proposed change (one at a time, not 5 at once) would be the minimum. Otherwise you could do a PEP, but once again you're looking at a PEP per exception. I think it's really going to come down to how big of an addition it is per exception, how easy it would be to have a memory leak, what the API to setting the attribute(s) would look like, etc. On Sat, Mar 13, 2021 at 1:22 PM Victor Stinner <vstinner@python.org> wrote:
Context: PEP 473 "Adding structured data to built-in exceptions" proposed in 2014. https://www.python.org/dev/peps/pep-0473/
Abstract: "Exceptions like AttributeError, IndexError, KeyError, LookupError, NameError, TypeError, and ValueError do not provide all information required by programmers to debug and better understand what caused them. Furthermore, in some cases the messages even have slightly different formats, which makes it really difficult for tools to automatically provide additional information to diagnose the problem. To tackle the former and to lay ground for the latter, it is proposed to expand these exceptions so to hold both the offending and affected entities."
For example the PEP proposes to add "target" and "attribute" attributes to AttributeError.
PS: Please give at least the PEP title for me when you mention a PEP only by its number, I'm unable to remember what are the 500+ PEP :-(
Victor
On Sat, Mar 13, 2021 at 7:49 PM Sebastian Kreft <skreft@gmail.com> wrote:
Hi dev-team, I'm reopening the discussion of PEP 473, which was closed
due to:
The steering council felt the PEP was too broad and not focused enough.
Discussions about adding more attributes to built-in exceptions can continue on the issue tracker on a per-exception basis (and obviously here for any broader points, e.g. performance implications as I know that has come up before when the idea of storing relevant objects on exceptions).
Before the PEP was rejected by the SC, I had submitted
https://github.com/python/cpython/pull/6271 and the last response said:
You may propose a newly revised PEP which overcomes the last SC feedback. But without the accepted PEP, this PR can not be merged.
Please clarify how can I proceed with adding some of the specified
fields to the mentioned exceptions. Should I create a new PEP as suggested in the PR or should we create individual BPO issues for each exception modification.
-- Sebastian Kreft _______________________________________________ Python-Dev mailing list -- python-dev@python.org To unsubscribe send an email to python-dev-leave@python.org https://mail.python.org/mailman3/lists/python-dev.python.org/ Message archived at
https://mail.python.org/archives/list/python-dev@python.org/message/4XVFFNBH...
Code of Conduct: http://python.org/psf/codeofconduct/
-- Night gathers, and now my watch begins. It shall not end until my death. _______________________________________________ Python-Dev mailing list -- python-dev@python.org To unsubscribe send an email to python-dev-leave@python.org https://mail.python.org/mailman3/lists/python-dev.python.org/ Message archived at https://mail.python.org/archives/list/python-dev@python.org/message/PLPUG6MF... Code of Conduct: http://python.org/psf/codeofconduct/
I am not a Python developer, but I am very interested in this topic. While I would definitely make use of any additional fields for exceptions as discussed, it is already possible to find most, if not all of the information discussed in PEP 473. To give just a concrete example taken from the PEP: ``` Friendly Console version 0.3.1. [Python version: 3.8.4]
a = {'foo': 1} a['fo']
Traceback (most recent call last): File "<friendly-console:2>", line 1, in <module> a['fo'] KeyError: 'fo' Did you mean foo?
foo = 1 fo
Traceback (most recent call last): File "<friendly-console:5>", line 1, in <module> fo NameError: name 'fo' is not defined Did you mean foo?
The above two examples are the actual output done by a program I work on called friendly (formerly known as friendly-traceback) and are essentially identical to what is given in PEP 473 as a desired outcome.
Much information about IndexError, AttributeError, etc, mentioned as desirable in PEP 473 can already be obtained by such automated tools. Using friendly as an example, a sample of what's possible can be found at https://aroberge.github.io/friendly-traceback-docs/docs/html/tracebacks_en_3.8.html
Of course, having information more easily available about specific exceptions would be welcome and make it easier and likely more error-proof to write automated tools. But since much of the information is already available (if one looks hard enough...), the cost of adding fields to existing exceptions should be carefully evaluated.
André Roberge
For best performance (reduce time to build an exception object), I would be interested to only format the error message on demand. For example, when str(exception) is called. The problem is the Exception.args attribute. Example: $ ./python Python 3.10.0a6+
exc=AttributeError("%s object has no attribute %s" % ("MyObject", "name")) str(exc) 'MyObject object has no attribute name' exc.args ('MyObject object has no attribute name',)
Currently, args is the raw positional arguments passed to the Exception constructor, and they are very likely many applications relying on Exception.args. For backward compatibility, we could store "%s object has no attribute %s" and ("MyObject", "name"), and build the args tuple on demand and format the string at the first str() call. At the C level, args is exposed directly as PyBaseExceptionObject.args and the PyBaseExceptionObject structure is part of the public C API. Changing it would be a C API incompatible change. By the way, the PEP 473 doesn't say anything about it and has no "Backward compatibility" section :-( Victor On Sat, Mar 13, 2021 at 7:49 PM Sebastian Kreft <skreft@gmail.com> wrote:
Hi dev-team, I'm reopening the discussion of PEP 473, which was closed due to:
The steering council felt the PEP was too broad and not focused enough. Discussions about adding more attributes to built-in exceptions can continue on the issue tracker on a per-exception basis (and obviously here for any broader points, e.g. performance implications as I know that has come up before when the idea of storing relevant objects on exceptions).
Before the PEP was rejected by the SC, I had submitted https://github.com/python/cpython/pull/6271 and the last response said:
You may propose a newly revised PEP which overcomes the last SC feedback. But without the accepted PEP, this PR can not be merged.
Please clarify how can I proceed with adding some of the specified fields to the mentioned exceptions. Should I create a new PEP as suggested in the PR or should we create individual BPO issues for each exception modification.
-- Sebastian Kreft _______________________________________________ Python-Dev mailing list -- python-dev@python.org To unsubscribe send an email to python-dev-leave@python.org https://mail.python.org/mailman3/lists/python-dev.python.org/ Message archived at https://mail.python.org/archives/list/python-dev@python.org/message/4XVFFNBH... Code of Conduct: http://python.org/psf/codeofconduct/
-- Night gathers, and now my watch begins. It shall not end until my death.
participants (4)
-
André Roberge -
Brett Cannon -
Sebastian Kreft -
Victor Stinner