Let jinja throw errors on undefined variables

[Elrond]

(this is to discuss this and then possibly implement it. Implementing is very easy).

Currently jinja2 replaces {{ not_known }} by the empty string. The question is, wether it would be better for it to throw an error? Rationale being: Either it's a typo, or it's an error for the variable to be referenced but not being there.

Does that make sense?

Enabling this:

mediagoblin/tools/template.py:
@@ -43,6 +43,7 @@ def get_jinja_env(template_loader, locale):

     template_env = jinja2.Environment(
         loader=template_loader, autoescape=True,
+        undefined=jinja2.StrictUndefined,
         extensions=['jinja2.ext.i18n', 'jinja2.ext.autoescape'])

[Christopher Allan Webber]

Hm. There might be some places where this would be nicer.

I remember when working with ZPT that it would complain on these things, it was nice in some places {{ foo }} but less in other places {{ objyoudohave.fooyoudont }}... being able to call attributes that you don't actually have and it going silent is nice. That said I remember a few years ago saying that ZPT complaining about such things was a significant feature in comparison to Django's templates.

How's that for a non-answer ;)

[Elrond]

jinja2 has a defined test. That should help if you have an attribute that purposefully might or might not exist.

Edit: We might also conider jinja2.DebugUndefined

[Elrond]

Okay, I have enabled this. And fixed two breaking places (simple).

I leave this bug open for the next 14 days, if this all turns out to be a bad idea. Let's see!

Commit: f1cdd278e7cf195e485567ed0d0d8a90cad81e48 and following

[Elrond]

13 days over, okay. We had a small bit of pain here and there, but it generally was acceptable.

[Will Kahn-Greene]

The original url for this bug was ​http://bugs.foocorp.net/issues/691 .