_format_assertmsg: use safeformat instead of saferepr

[blueyed]

Custom assertion messages should not get shortened via saferepr.

Via blueyed#125.
Fixes #5199.

[nicoddemus]

Hi @blueyed,

Thanks for the patch!

The underlying reasoning is sound, but the test doesn't seem to be harden enough.

Here's the output of test_AssertionError_message on features:

___________________________________________ test_bytes ____________________________________________

    def test_bytes():
>       assert 0, b'b' * 80
E       AssertionError: b'bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb'
E       assert 0

test_AssertionError_message.py:6: AssertionError

And here's the output from your branch:

___________________________________________ test_bytes ____________________________________________

    def test_bytes():
>       assert 0, b'b' * 80
E       AssertionError: (b'bbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbbb'
E          b'bbbb')
E       assert 0

test_AssertionError_message.py:6: AssertionError

Both are not truncating the message, and in fact the message from features looks better to me. Thoughts?

[blueyed]

@nicoddemus
saferepr truncates at 240 characters.

I've thought it would be enough to show that safeformat gets used, which here wraps at 80 characters.

I've initially done this to fix:

>>> saferepr(list(range(0, 10)))
'[0, 1, 2, 3, 4, 5, ...]'

vs

>>> safeformat(list(range(0, 10)))
'[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]'

(Note that this patch would also look better with wider columns, not using 80, of course.
That's something to do as a follow-up though.)

I've also thought to use just str in the first place, but then thought that formatting is good in general.

Also note that safeformat does not wrap strings.

[blueyed]

Should maybe use _pformat_dispatch from cc503c1?

[nicoddemus]

I've thought it would be enough to show that safeformat gets used, which here wraps at 80 characters.

I see, thanks for the explanation. Do you know why it is showing the byte around () though?

Should maybe use _pformat_dispatch from cc503c1?

Hmm sounds reasonable.

[blueyed]

I've thought it would be enough to show that safeformat gets used, which here wraps at 80 characters.

I see, thanks for the explanation. Do you know why it is showing the byte around () though?

This is because it gets split into multiple lines.

Also happens for strings, but requires spaces there it seems:

>>> pprint.pprint(('b'*10 + ' ') * 8)
('bbbbbbbbbb bbbbbbbbbb bbbbbbbbbb bbbbbbbbbb bbbbbbbbbb bbbbbbbbbb bbbbbbbbbb '
 'bbbbbbbbbb ')

Should maybe use _pformat_dispatch from cc503c1?

Hmm sounds reasonable.

Not so sure myself anymore. Anything fitting 80 chars (to be the terminal width in the long run) should be ok to not wrap probably.

So I'd rather leave it as is, maybe adding the extra test for not truncating the tuple/list?