#27753 closed Cleanup/optimization (fixed)
Cleanups when no supported version of Django supports Python 2 anymore
Reported by: | Aymeric Augustin | Owned by: | |
---|---|---|---|
Component: | Utilities | Version: | dev |
Severity: | Normal | Keywords: | |
Cc: | Jon Dufresne | Triage Stage: | Accepted |
Has patch: | yes | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description (last modified by )
Compatibility with Python 2 is currently being removed in master as part of the Django 2.0 development cycle: see #23919.
However third-party apps may want to remain compatible with all supported version of Django and the corresponding versions of Python, which includes Python 2.
To allow this, compatibility functions or modules aren't removed. They're still importable. They're usually just aliases or no-ops. Here's a list of things to deprecate and remove eventually:
- django.utils.six
- django.utils.lru_cache
- django.utils._os.abspathu, upath, npath
- django.utils.decorators.available_attrs
- django.utils.encoding.python_2_unicode_compatible
- django.utils.encoding.smart/force_text
- django.utils.http.urlquote/urlquote_plus/urlunquote/urlunquote_plus (unmerged commit)
- django.utils.safestring.SafeBytes
- One of either django.utils.safestring.SafeString or django.utils.safestring.SafeText
- django.test.utils.str_prefix()
- django.utils.encoding.DjangoUnicodeDecodeError
django.utils.functional.curry()
(in favor offunctools.partial()
/partialmethod()
; see 5b1c389603a353625ae1603ba345147356336afb)- django.test.utils.patch_logger()
Change History (40)
comment:1 by , 8 years ago
comment:2 by , 8 years ago
Triage Stage: | Unreviewed → Someday/Maybe |
---|
Those methods don't work on Python 3 (AttributeError: 'dict' object has no attribute 'iterkeys'
) so I don't think they have value for the use case described in the ticket description. A third-party app should use six.iterkeys()
as the old test did (which uses keys()
etc. on Python 3), or am I missing something?
comment:3 by , 8 years ago
The question we discussed on IRC and GitHub with Tim is about deprecating or not deprecating those items.
I would be for a standard deprecation process as I don't see how this is different from any other deprecation. But I'm open to read about use cases which would push for a delayed deprecation process for items related to Python 2.
comment:4 by , 8 years ago
Description: | modified (diff) |
---|
comment:5 by , 8 years ago
Description: | modified (diff) |
---|
comment:6 by , 8 years ago
Description: | modified (diff) |
---|
comment:7 by , 8 years ago
Description: | modified (diff) |
---|
comment:8 by , 8 years ago
Description: | modified (diff) |
---|
comment:9 by , 8 years ago
Owner: | changed from | to
---|---|
Status: | new → assigned |
comment:10 by , 8 years ago
Owner: | removed |
---|---|
Status: | assigned → new |
Please don't work on this ticket -- the timetable for these removals isn't decided but it's not for some years. The "Someday/Maybe" "Triage Stage" of the ticket indicates that it's not appropriate to work on right now.
comment:11 by , 8 years ago
Description: | modified (diff) |
---|
comment:12 by , 8 years ago
Description: | modified (diff) |
---|
comment:13 by , 8 years ago
Description: | modified (diff) |
---|
comment:15 by , 7 years ago
Description: | modified (diff) |
---|
comment:16 by , 6 years ago
I think this was discussed on the mailing list or in IRC but I can't find the discussion so I'll state it here. The reason we don't want to deprecate this stuff now is because that would add lots of non-actionable warnings for Django libraries that want to support Python 2 and 3 until 2020.
comment:17 by , 6 years ago
Description: | modified (diff) |
---|
comment:18 by , 6 years ago
I just wanted to mention that in my opinion it would make more sense for SafeText
to go away instead of SafeString
to be consistent with the (smart|force)_text
removal.
Given we're currently publicly documenting that SafeText
is an alias for SafeString
we should adjust the code and documentation for it to be other way around instead before we move forward with the SafeText
removal or deprecation.
comment:20 by , 6 years ago
Has patch: | set |
---|---|
Triage Stage: | Someday/Maybe → Accepted |
PR for some removals.
comment:21 by , 6 years ago
PR for favoringforce_str()
, smart_str()
, and SafeString
over force_text()
, smart_text()
, and SafeText
. It also deprecates the former functions. No deprecation for SafeText
because it seems more trouble than it's worth right now (deprecation warnings in __init__()
and __isinstance__()
?).
comment:22 by , 6 years ago
Cc: | added |
---|
My PRs address everything in this ticket except DjangoUnicodeDecodeError
-- I'm not sure about the rationale for removing it as it still may provide some value:
DjangoUnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte. You passed in b'\xff' (<class 'bytes'>)
vs. UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte
.
Could you explain your thinking, Jon? Maybe the mistake leading to this exception (#5640) isn't common enough in Python 3 to continue with the special handling?
comment:23 by , 6 years ago
I can't recall exactly, but I suspect my thinking was along the lines of:
- After dropping Python 2,
force_text()
& friends would eventually be removed from Django and replaced by stricter type contracts at the boundaries. This has proved to be time consuming and difficult to accomplish, so this isn't yet complete (but we're getting there).
DjangoUnicodeDecodeError
looks like a close duplicate ofUnicodeDecodeError
. As you pointed out, it is not as it provides more information. Although, if the Python version is insufficient, perhaps effort could be spent improving that message.
So, given the above, I'm OK for it to stick around. No need to remove it.
Maybe the mistake leading to this exception (#5640) isn't common enough in Python 3 to continue with the special handling?
I suspect that is probably true, but I can't say for certain.
comment:35 by , 6 years ago
Resolution: | → fixed |
---|---|
Status: | new → closed |
Should we revert the removal of eb0b921c29ace8643a5a4cd136c433727c53dead in this case?