Set delimiter for triple mustache?

[aartaka]

I see test sets and documentation only referring to double mustache delimiter redefinition. Is it intended that triple mustaches are not redefinable? I would very much like to redefine them for consistency with double mustache.

Something like {{{=|| ||=}}} might be a consistent syntax.

[jgonggrijp]

Thank you for raising this issue. I converted it to a discussion, because there is no immediately clear course of action.

The triple mustache is a special case with an unclear status in the specs. There is a regular form for the unescaped variable that uses & as the sigil: {{&variablename}}. If you want consistency, you always have the option to use that sigil with your alternative delimiters. For example {{=[ ]=}}[&variablename].

My own Mustache implementation for JavaScript, Wontache (which also powers the playground), treats the third pair of Mustaches as a special sigil that needs to have a matching closing sigil. That means that in Wontache, you could do {{=[ ]=}}[{variablename}]. This is definitely not in any way justified by the spec, but is also is not in any way ruled out. You should count on it not being portable across implementations.

As for your proposal to allow for a {{{= syntax, this is something we could discuss here. Speaking for myself, I think it will cause more problems than it will solve:

1. If you use the same replacement for double and triple mustaches, for example {{=[ ]=}}{{{=[ ]=}}}, all subsequent tags become ambiguous.
2. If you do anything creative, where the replacement of the double opening mustache is not a prefix of the replacement of the triple opening mustache or the replacement of the double closing mustache is not a suffix of the replacement of the triple closing mustache, i.e. anything NOT of the form {{=A Z=}}{{{=AB YZ=}}}, then the template becomes harder to read for humans AND most existing implementations will likely not be able to parse it correctly.
3. Having separate set delimiters makes it possible to swap the nesting relationship, for example {{=[[ ]]=}}{{{=[ ]=}}}. In addition to what I wrote in point 2, this adds a whole extra level of confusion.

Theoretically, such problems could be addressed by making the compilers more complicated and more vigilant about user mistakes, but I think this would be solving the wrong problem. Our problem is not that we lack a way to override triple mustaches, but that we have triple mustaches in the first place. I would much rather see a Mustache 2.0 where the triple mustaches are dropped entirely, in favor of the regular {{& form.

[jgonggrijp]

I would fully expect any sane implementation to detect delimiters eagerly and interpret that example as << <. That being said, you are right that this is currently undefined in the spec.

Then again, note that this problem would only become worse if there was a separate set delimiter syntax for triple mustaches. With {{=<< >>=}}{{{=<<< >>>=}}}, you cannot use parent tags at all. Or no triple mustaches at all, depending on which is given priority by the parser.

[aartaka]

I would fully expect any sane implementation to detect delimiters eagerly and interpret that example as << <. That being said, you are right that this is currently undefined in the spec.

That's why I don't assume any behavior—undefined behavior might mean anything implementation decides to do.

Then again, note that this problem would only become worse if there was a separate set delimiter syntax for triple mustaches. With {{=<< >>=}}{{{=<<< >>>=}}}, you cannot use parent tags at all. Or no triple mustaches at all, depending on which is given priority by the parser.

It's exactly what I was talking about regarding discipline. One can program themselves in the corner with Mustache, both with custom triple mustache or without it. Yes, the magnitude differs, but the problem is still there.

[jgonggrijp]

My point is that the magnitude of the problem is greater if you add the option to override the triple mustache delimiter as well. If you agree with that, why do you still want to allow it?

[aartaka]

I mostly strive for consistency here—if you can redefine double mustache, why not do so for triple one? But {{& cuts it and looks much more consistent, so triple mustache redefinition seems unnecessary.

I guess we're at the 2.0 territory with whether triple mustache should even be there. Is there a discussion for 2.0 that I can link this discussion to?

[jgonggrijp]

Not a discussion, but there is the fossilized issue ticket #50. This discussion is now linked to it by virtue of me having created that backreference.

If we want to brainstorm about a 2.0 again, we should probably create a new discussion for that purpose. All the people who were involved in that ticket at the time have moved on.

Edit: apparently GitHub does not create backreference when a discussion mentions an issue, so #50 does not link to this discussion yet. Not sure how useful that would be, though, since the last comment was 13 years ago.

[jobol]

On my C implementation, {{=<< >>=}} provides the unescape tags <<{ ... }>>.

[jgonggrijp]

Interesting to know that Wontache is not the only engine doing that.