bpo-36077: Update handling of defaulted fields #17322
Conversation
When a non-defaulted dataclass field follows a defaulted field, the behaviour has changed from raising a `TypeError` to making all arguments in `__init__` following the defaulted field keyword-only.
|
Looking forward to this! |
|
This is waiting on PEP 557 to be changed to allow this in the first place. I've submitted a feature-request in Discourse but there's been no discussion. |
|
We don't typically update PEPs when new features are added. |
|
Other way around: this pull-request is in direct contradiction with the PEP (it says that non-defaulted fields following defaulted fields cause a TypeError) |
|
It doesn't matter: PEPs are not meant to describe current behavior. |
|
For those wanting a patch: https://gist.github.com/EpicWink/195f43286e47c26eef7d5eb3263fee75 |
|
This is so useful! Thanks for the patch. |
|
Thank you for the patch. Any way to know if this will get into 3.10? |
|
There hasn't been much discussion in the mailing list entry I posted or the Discourse post, so it seems the core devs aren't interested. This suggestion from 2018 suggests allowing it in a backwards-compatible way by introducing a There is an existing solution in the mailing list: pass |
|
I've posted a different proposal in python-ideas at https://mail.python.org/archives/list/python-ideas@python.org/message/FI6KS4O67XDEIDYOFWCXMDLDOSCNSEYG/ I'd be interested in any opinions on it. Please comment on python-ideas for a wider audience. Thanks! |
|
This has been superseded by the more explicit solution to be included in Python 3.10 #24909 |
When a non-defaulted dataclass field follows a defaulted field, the behaviour has changed from raising a
TypeErrorto making all arguments in__init__following the defaulted field keyword-only.I'm not sold on the prettiness of the
versionchangedarea in the docs change: please comment.https://bugs.python.org/issue36077