Add a "last" template filter

[Chris H.]

Hi,

It seemed like an oversight that there wasn't a "last" template filter similar to the "first" filter. I wrote it up, along with tests, and added in an nth filter that takes an arg and accessing that index of the value.

>>> last([0, 1, 2, 3, 4])
4
>>> last(u'Hello')
u'o'
>>> nth(['a', 'b', 'c', 'd'], 1)
'b'
>>> nth(u'abcd', 1)
u'b'

[James Bennett]

(cleaning up formatting in description)

[Jacob]

nth isn't needed: you can do {{ list.14 }} (or whatever) instead. last, though, sounds like a good idea.

[Aaron C. de Bruyn]

Changed first.is_safe = False to patch as well.

[Chris Beaven]

Thanks darkpixel.

(We discussed changing is_safe to false in IRC, first and last don't guarantee a safe string out, eg first(mark_safe('&')))

[Malcolm Tredinnick]

I can't think of any example where last() is going to return an unsafe result for a safe string. The first() change is correct, though.

[Chris Beaven]

perhaps last(mark_safe('<a>') -- just because '<a>' is safe doesn't mean that '>' should be considered safe, does it? (but this is probably misuse of the filter anyway, it's obviously meant primarily for lists)

[Malcolm Tredinnick]

(In [6997]) The "first" filter can return an unsafe string for safe input ( {{"<"|first}} ), so change is_safe to False. Refs #5567.

[Malcolm Tredinnick]

(In [6998]) Fixed #5567 -- Added a "last" filter. Based on a patch from darkpixel.

[Malcolm Tredinnick]

If "<a>" is marked as safe, then the user intended for the ">" to be displayed literally. There's no problem there. The ">" was already marked appropriately and its semantics don't change just because it doesn't have the "<a" prefix any longer (c.f. "&lt;" where the semantics change when you strip off the trailing "lt;", as you noted).

[Antti Kaihola]

Replying to jacob:

nth isn't needed: you can do {{ list.14 }} (or whatever) instead. last, though, sounds like a good idea.

What about {{ list|nth:offset }} where offset is an integer?