email.policy.Compat32(max_line_length=None) not as documented

[vadmium]

BPO	29478
Nosy	@warsaw, @bitdancer, @vadmium, @Mariatta, @mircea-cosbuc
PRs	[email] bpo-29478: Fix passing max_line_length=None from Compat32 policy #595 [3.6] bpo-29478: Fix passing max_line_length=None from Compat32 policy #2024 [3.5] bpo-29478: Fix passing max_line_length=None from Compat32 policy #2025 [3.6] [email] bpo-29478: Fix passing max_line_length=None from Compat… #2233 [3.5] [email] bpo-29478: Fix passing max_line_length=None from Compat… #2234

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2017-06-16.14:21:56.977>
created_at = <Date 2017-02-08.10:10:45.594>
labels = ['type-bug', '3.7', 'expert-email']
title = 'email.policy.Compat32(max_line_length=None) not as documented'
updated_at = <Date 2017-06-16.14:21:56.976>
user = 'https://github.com/vadmium'

bugs.python.org fields:

activity = <Date 2017-06-16.14:21:56.976>
actor = 'Mariatta'
assignee = 'none'
closed = True
closed_date = <Date 2017-06-16.14:21:56.977>
closer = 'Mariatta'
components = ['email']
creation = <Date 2017-02-08.10:10:45.594>
creator = 'martin.panter'
dependencies = []
files = []
hgrepos = []
issue_num = 29478
keywords = []
message_count = 12.0
messages = ['287294', '287310', '289368', '289383', '289385', '289387', '289421', '295491', '295526', '295748', '296146', '296194']
nosy_count = 5.0
nosy_names = ['barry', 'r.david.murray', 'martin.panter', 'Mariatta', 'mcosbuc']
pr_nums = ['595', '2024', '2025', '2233', '2234']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue29478'
versions = ['Python 3.5', 'Python 3.6', 'Python 3.7']

[vadmium]

By default, the email package turns single-line header fields into multi-line ones to try and limit the length of each line. The documentation <https://docs.python.org/release/3.5.2/library/email.policy.html#email.policy.Policy.max_line_length\> says that setting the policy’s max_line_length attribute to None should prevent line wrapping. But this does not work:

>>> from email.policy import Compat32
>>> from email.message import Message
>>> from email.generator import Generator
>>> from sys import stdout
>>> p = Compat32(max_line_length=None)
>>> m = Message(p)
>>> m["Field"] = "x" * 100
>>> Generator(stdout).flatten(m)  # Field is split across two lines
Field: 
 xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

>>

A workaround is to specify zero instead:

>>> p = Compat32(max_line_length=0)
>>> Generator(stdout, policy=p).flatten(m)  # All on one line
Field: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Quickly looking at the code, Compat32._fold() passes max_line_length straight to Header.encode(), which is documented as using None as a placeholder for its real default value of 76. So I think the solution would be to add a special case in _fold() to call encode(maxlinelen=0) if max_line_length is None.

[bitdancer]

That sounds reasonable to me. Clearly there is a missing test :)

[bitdancer]

Thanks for the PR. However, rereading this: since compat32 is providing backward compatibility with the behavior of the python 3.2 email package, we need to check what it would do in this situation before changing the behavior. What we may need instead is a doc fix, unfortunately :(. But a test for this is needed either way.

[mircea-cosbuc]

Thanks for the prompt feedback. In Python 3.2, the closest equivalent for the illustrated issue I could find is:

>> from email.message import Message
>> from email.generator import Generator
>> from sys import stdout
>> m = Message()
>> m["Field"] = "x" * 100
>> g0 = Generator(stdout, maxheaderlen=0)
>> gn = Generator(stdout, maxheaderlen=None)

>>> g0.flatten(m)
Field: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

>>> gn.flatten(m)
Field: 
 xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

It may be the case that a documentation change is all that is needed. I'm not sure that this change would break compatibility since max_line_length=None for the policy is not the default value. Please advise further if I should just update the documentation and modify the test to guard for future changes.

[bitdancer]

So what happens when you do that same operation in 3.5/6 with your change in place? Does the behavior change? (I haven't looked back at the code to see if I think it will :)

[mircea-cosbuc]

Just to be sure, I performed the same operations with my changes in place, there's no change in behaviour. I think it's expected since I only modified how the Compat32 policy passes max_line_length to the header class.

[bitdancer]

OK. This looks good to me. I haven't figured out the new commit process, though (like how to do misc news and backports), so I'm not going to be the one to merge it, I'm afraid. At least not until I do find time to learn.

[Mariatta]

Is the PR ready for merging?
Does this need misc/news entry?

I can help with the backport.

[bitdancer]

Looks like it just needs a NEWS entry.

[Mariatta]

New changeset b459f74 by Mariatta (mircea-cosbuc) in branch 'master':
[email] bpo-29478: Fix passing max_line_length=None from Compat32 policy (GH-595)
b459f74

[Mariatta]

New changeset 820b714 by Mariatta in branch '3.5':
[email] bpo-29478: Fix passing max_line_length=None from Compat32 policy (GH-595) (GH-2234)
820b714

[Mariatta]

New changeset e9f4d8d by Mariatta in branch '3.6':
[email] bpo-29478: Fix passing max_line_length=None from Compat32 policy (GH-595) (GH-2233)
e9f4d8d