floatformat templatetag not robust

[Soeren Sonnenburg <bugreports@…>]

it fails when the input is None or inf for example.

[jefurii]

it fails when the input is None or inf for example.

I think the poster means "int". They don't define "fails" either.

According to the docs (​http://www.djangoproject.com/documentation/templates_python/#writing-custom-template-filters), filters should fail silently or return the original input. floatformat looks like it's working correctly:

This code

Float: {{ 34.12345|floatformat }}<br/>
None: {{ None|floatformat }}<br/>
Int: {{ 34|floatformat }}<br/>

produces this output,

Float: 34.1
None: 
Int: 34

which is correct according to the guidelines for filters.

[Soeren Sonnenburg <bugreports@…>]

no I meant inf (infinity), nan (not a number) etc.

[Karen Tracey <kmtracey@…>]

Replying to Soeren Sonnenburg <bugreports@nn7.de>:

no I meant inf (infinity), nan (not a number) etc.

OK, I see a problem with infinity (OverflowError) but not NaN. What problem do you see with NaN? On Linux w/ Python 2.5.1 it displays as 'nan'. On Windows w/ Python 2.5.1 it displays as '-1.$'; this seems to be due to underlying Python behavior and I'm not sure what Django could do about it. (On Linux with a fix for the Overflow error infinity display as 'inf', on Windows it's '1.#INF'.) I'll attach a fix for the OverflowError in case that is useful.

Not sure what else to test to cover the 'etc' mentioned in the comment. I did also try negative infinity and that too works OK once the overflow error is fixed.

[Karen Tracey <kmtracey@…>]

Catch overflow error on attempt to floatformat infinity.

[Philippe Raoult]

added regression tests

[Karen Tracey <kmtracey@…>]

FYI the added tests aren't going to work on Windows. Constructing special values there isn't so straightforward and they don't display the same when they are turned into strings:

D:\tmp>python
Python 2.5.1 (r251:54863, Apr 18 2007, 08:51:08) [MSC v.1310 32 bit (Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> x1 = float('infinity')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: invalid literal for float(): infinity
>>> x2 = float('nan')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: invalid literal for float(): nan
>>> x1 = float(1e30000)
>>> x1
1.#INF
>>> x2 = x1/x1
>>> x2
-1.#IND
>>>

​http://www.python.org/dev/peps/pep-0754/ is what I used to get some idea of Python's cross-platform handling of special float values. Maybe there is a better reference but it was the best I found via a quick Google search. Net that I got from that PEP (which is rejected) is that constructing these values in a cross-platform-independent fashion is possible but roundabout, and their string representations differ across platforms. Makes it hard to construct a cross-platform-independent test for floatformat.

[Philippe Raoult]

I've read that and some other stuff and it looks like there's nothing we can do right now because this is an issue with python. The tests won't work but the tags will, and I guess that'll be good enough for now.

Thanks for the heads-up!

[Malcolm Tredinnick]

(In [7336]) FIxed #6513 -- Handle overflows better in the floatformat filter. It's not
possible to test this automatically everywhere due to differing representations
on different platforms. Manual testing confirms it works, though.

Thanks, Karen Tracey.