Set nbsp instead of space as thousand separator

[fadeyev]

Space was replaced with nbsp only for Polish locale: https://code.djangoproject.com/ticket/13577
I wonder why it hasn't been done for all other locales that use a space as a thousand separator.

[Aymeric Augustin]

I'd be nice to check the typographical conventions of the 8 countries involved in this patch, just to be sure that NBSP is the proper character — there are other variants of SP.

That said, I can't imagine a reason why a SP would be better than a NBSP as thousand separator. The only difference is that the latter prevents line wrapping, and line wrapping in the middle of a number can't be correct. So, even if NBSP isn't the sanctioned character, I'd still be better than SP.

If we wanted to follow this logic entirely, we should also patch docs/topics/i18n/formatting.txt.

[Claude Paroz]

Also update docs/topics/i18n/formatting.txt

[Jannis Leidel]

In [17478]:

Fixed #17217 -- Use non breaking spaces for format localization in which spaces are used. Thanks, Claude Paroz.

[Alexey Boriskin]

One thing that had not been considered when that patch was applied is that thousand separator is used not only for formatting, but in forms input too. It is rather hard to force users to use non-breaking space when they input numbers.
Probably we should have one setting for formatting and another for forms validation.

[Claude Paroz]

Yeah, this striked me too, some days ago. Wouldn't non-breaking space common enough to justify some special-casing when validating inputs?

[Alexey Boriskin]

I think about using unicodedata module. Non-breaking space normalizes to normal space:

>>> import unicodedata
>>> unicodedata.normalize('NFKD', u'\xa0')
u' '

This way we can determine that non-breaking and usual spaces are roughly the same. So we can special-case non-breaking space without actually hardcoding it -)

[Alexey Boriskin]

I've created a pull request, please review: ​https://github.com/django/django/pull/726

[Claude Paroz <claude@…>]

In b19d83fc12ebbabbaa9d72286f2e4bfa071ff784:

Improved input sanitizing with thousand separators

For languages with non-breaking space as thousand separator,
standard space input should also be allowed, as few people know
how to enter non-breaking space on keyboards. Refs #17217.
Thanks Alexey Boriskin for the report and initial patch.