Error changes #307
Error changes #307
Conversation
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## main #307 +/- ##
=======================================
Coverage 97.36% 97.37%
=======================================
Files 55 55
Lines 6344 6352 +8
Branches 45 44 -1
=======================================
+ Hits 6177 6185 +8
Misses 167 167
Continue to review full report at Codecov.
|
|
Also reverted |
|
Makes sense, and as I was saying on DM, I don't have a personal preference, but maybe there's a way to gauge usage and how much would it break people's code. But it also makes sense to try and keep compatibility, if it's doable enough. 🤷 |
|
I think it's expected for v2 to be a big change! People want to have the same features (and more) but already know it won't be free. I would go with whatever seems the best for the future (tuple looks good, naming too) even if it changes more compared to v1. TLDR: I'm in favour of "get it right" |
Agreed. But on this occasion, I think what we have here is right. I'll work on this tomorrow. |
Feedback on this welcome, in particular @tiangolo and @PrettyWood.
Basically this is reverting the keys in error information to match pydantic V1.
While I'm in favour of "get it right" over "keep it compatible", I don't think the win here is worth the headache this this will cause users.
With this change, errors like like:
[ { 'type': 'foobar', # new values but this is a marked of what type of error we've got, mostly for machines to read 'loc': ('field_a', 'field_b'), # loc remains a tuple, this gives a slight performance improve over using a list 'msg': 'Whatever, # message template with context substituted 'input': '...', # this is new, it's the value that was submitted to this field/item 'ctx': {'foo': 123}, # not always included, the context used to render the msg }, { ... }, ... ]Before this change (e.g. on
mainof pydantic-core atm.), errors looked like:[ { 'kind': 'foobar', 'loc': ['field_a', 'field_b'], 'message': 'Whatever', 'input_value': '...', 'context': {'foo': 123}, } ]While this might be a bit better, I don't think it's enough better to make up for the hassle this change will cause people.
WDYT?
If we decide to change this, I'd rather do it in pydantic/pydantic#4516 than after to minimise the changes there.