Feature: escape hatch for colliding template syntax in Django templates

[David Gouldin]

As discussed in django-developers:

​http://groups.google.com/group/django-developers/browse_thread/thread/e914638d8fe859de

There is a need for a template block tag which renders template syntax characters within the block as normal text. This serves as an escape hatch to avoid syntax collisions when using a template language meant to be rendered on the client, not the server. A potential implementation is attached.

[floguy]

I don't think this patch is quite sufficient because it can actually change the contents of the text in the noparse block. Specifically, it won't preserve whitespace within things that it interprets as blocks or variables.

However, I don't think it's going to be possible to escape out of parsing using the usual templatetag constructs, because it's already tokenized and parsed by the time it's handed off to the node.

Maybe a pre-processing step that happens before everything else is in order?

[Jeremy Dunck]

Eric, to be clear, are you referring to Lexer.create_token using .strip?

If so, I see the need for that (later context[var_name] kinds of lookups), but perhaps tokens could hang on to their raw contents? Or perhaps a final step in Parser could perform those strips after nodes have been constructed? (Neither sounds ideal...)

[Aymeric Augustin]

#16318 was a duplicate. It contains an alternative implementation.

Another discussion : ​http://groups.google.com/group/django-developers/browse_thread/thread/eda0e9187adcbe36

"verbatim" is a better name than "noparse" IMO — that's how people call this feature spontaneously.

[Chris Beaven]

See ​https://github.com/django/django/pull/51

[Aymeric Augustin <aymeric.augustin@…>]

In [4464bbba15d50ed32beb5995c13d26791ca61fe4]:

Fixed #14502 -- Added a verbatim template tag.

Thanks SmileyChris for the patch.

[Jannis Leidel]

I'm not sure I understand the purpose of the alternate closing tag, can you elaborate (in the docs and here) why this was needed to be added (as it's a rather extensive departure of a core tag)?

[Aymeric Augustin]

So, I just reviewed the mailing list discussions on this topic.

Clearly there's some demand for a verbatim tag, and it should be implemented in the template engine itself (because the verbatim content shouldn't be parsed).

Personally, I'd prefer if the verbatim text could contain any text -- including {% endverbatim %} -- otherwise someone's going to notice and complain. But we could very well live without this possibility: there's always the (ugly) alternative of splitting the {% endverbatim %} to output, like this:

{% verbatim %}bla bla {% end{% endverbatim %}{% verbatim %}verbatim %} bla bla{% endverbatim %}

For the record, the alternate closing tag was suggested ​on the mailing list (and also in a few follow up posts). Chris implemented it and considered his patch ready to go, basically endorsing the idea. The implementation looked reasonable to me, so I went ahead and committed the patch.

[Aymeric Augustin]

Assigning to myself so I don't forget to resolve this, one way or another.

[Aymeric Augustin]

Following a discussion with Jannis, here's a patch that removes the overridable end tag. It's trading debatable benefits for some complexity and doesn't fit so well with the rest of the template language. Specifically, it could make refactoring the template language significantly harder.

If we absolutely wanted to keep a way to escape {% endverbatim %}, I'd suggest using {% verbatim blahblah %}...{% endverbatim blahblah %}, similar to what the {% block %} tag does.

[Chris Beaven <smileychris@…>]

In [c57ba673312cb5774d544353044e2182b6223040]:

Fixed #14502 again -- saner verbatim closing token

Previously, the closing token for the verbatim tag was specified as the
first argument of the opening token. As pointed out by Jannis, this is
a rather major departure from the core tag standard.

The new method reflects how you can give a specific closing name to
{% block %} tags.

[anonymous]

Jannis: if you're still not happy with how this looks, then I'm fine with just removing this "edge case" and never allowing for having {% endverbatim %} unrendered.

[Chris Beaven]

(says me)