Fix URL pattern parsing in schema generation #5689
Conversation
- Call `str(pattern)` to get non-escaped route - Strip converters from path to comply with uritemplate format Fixes encode#5675
|
@encode/django-rest-framework-core-team 3.7.4 is due to go tomorrow. I can review in the morning but @tiltec has capacity to continue work today. Thus if any of you have the capacity to review nowish that would help greatly. (My initial thought was Looks Good, but it’s all new and I haven’t had time to explore properly.) |
|
@tiltec That failure will be the isort change. There’s a weirdness such that isort reports different results for Python 2.7. Revert it and it will pass. We need to work out why this is. We also need to stop using Python 2.7 to run the lint build. 🙂 |
|
Oh, a lint failure for Py2.7. Let's see where it is. @carltongibson |
tests/test_renderers.py
Outdated
| @@ -4,6 +4,7 @@ | |||
| import re | |||
| from collections import MutableMapping, OrderedDict | |||
|
|
|||
| import coreapi | |||
There was a problem hiding this comment.
This change will be the one causing the failure. Not sure what’s causing the isort inconsistency.
There was a problem hiding this comment.
https://travis-ci.org/encode/django-rest-framework/jobs/318806673#L499
It’s reporting for test_schemas too so you’lol need to confirm that, but the test_renderers one is a glitch with isort and Python 2.7 (and possibly our configuration, I’m not sure...🙂)
| @@ -135,6 +136,11 @@ def endpoint_ordering(endpoint): | |||
| return (path, method_priority) | |||
|
|
|||
|
|
|||
| _PATH_PARAMETER_COMPONENT_RE = re.compile( | |||
| r'<(?:(?P<converter>[^>:]+):)?(?P<parameter>\w+)>' | |||
There was a problem hiding this comment.
This is the full regex copied from Django 2. I didn't dare to import it, since it had an underscore in front of it.
|
lint passed! ✨ |
There was a problem hiding this comment.
OK. This looks great. Lets have it.
@tiltec Thanks for the timely input! 💃🏼
| @@ -32,7 +32,7 @@ | |||
| def get_regex_pattern(urlpattern): | |||
| if hasattr(urlpattern, 'pattern'): | |||
| # Django 2.0 | |||
| return urlpattern.pattern.regex.pattern | |||
| return str(urlpattern.pattern) | |||
There was a problem hiding this comment.
I think this has broken #5672 even further - instead of getting the expected regex it now gets a path/regex hybrid like 'convtest/<int:pk>\\.(?P<format>[a-z0-9]+)/?$' when trying to add suffix patterns.
There was a problem hiding this comment.
Indeed, str(urlpattern.pattern) is not always a regex pattern anymore. Maybe we need to undo this change and have an if-else in EndpointEnumerator, depending if urlpattern is a RoutePattern or a RegexPattern?
There was a problem hiding this comment.
That would still leave #5672 broken, just in a different way.
Perhaps renaming the method to something like get_original_route would help avoid confusion, but a separate fix would still be needed for the other issue.
There was a problem hiding this comment.
No problem adding a similar, not quite duplicate, function to compat. (We can always abstract again later.)
|
Well done! #5691 🙂 |
|
@carltongibson just a test for now; I need to think a bit about possible ways to fix it and more tests/edge cases |
|
@axnsan12 All good! 😉 |
* Fix url parsing in schema generation - Call `str(pattern)` to get non-escaped route - Strip converters from path to comply with uritemplate format. Background: encode#5675 (comment) Fixes encode#5675
str(pattern)to get non-escaped route.include_docs_urls#5675 (comment)Closes #5675