Lower-case-ing the heredoc language

[mauricioszabo]

Two small fixes here:

1. Usually, on Ruby, "heredocs" are always uppercase. So I'm lower-casing so that it tries to get the right grammar
2. When a grammar is not found, heredocs are highlighted as strings

[savetheclocktower]

Typically the injectionRegex has been written to capture either uppercase or lowercase, but I don't have a problem with this. Seems like a more elegant way of handling it.

[mauricioszabo]

@savetheclocktower just wanted to be sure what do you think about the (heredoc_content) being highlighted as string, but if you don't have any problem with it, I'll just write a test case and be done with this :)

[savetheclocktower]

Doesn't heredoc_body take care of that?

[mauricioszabo]

Not really, it was only highlighting around the doc, not the contents

[savetheclocktower]

So heredoc_content is a child of heredoc_body. We should choose to apply string.unquoted.ruby to one of them, not both, or else the scope will get repeated.

[savetheclocktower]

Hang on. This feedback still needs to be addressed:

So heredoc_content is a child of heredoc_body. We should choose to apply string.unquoted.ruby to one of them, not both, or else the scope will get repeated

[savetheclocktower]

Need to apply scope to heredoc_content or heredoc_body, not both.

Didn't realize further changes needed to be addressed

[mauricioszabo]

Ok, @savetheclocktower, it seems that it's better to apply only to heredoc_body because heredoc_content only applies to the inside of the string - so the highlighting could become really weird...

[savetheclocktower]

The reason I had for doing test.final here — and it's more of a feeling than anything — is that I felt we didn't really gain anything from tagging one range as both support.other.class.ruby and constant.ruby.

There are places where it makes sense to do stuff like that, but if we apply both scopes, we end up leaving it to the syntax theme to decide which styles win out, and often that's not an intentional decision by the theme author as much as it's just an accident of CSS specificity and ordering of selectors. It also brings ordering in the SCM file into play — since constant.ruby is declared later in the file, that scope name will be in the inner span, and that also has an effect on which styles win out.

Technically, lots of things are marked as constant in tree-sitter-ruby, just as lots of things are marked as identifier, and I'm not certain that we give the user much useful information by scoping each one as constant even when it serves a more specific function.

I've tended to assign a constant scope only for (a) number literals, (b) built-in language constants like booleans and null, and (c) declarations that LOOK_LIKE_CONSTANTS and don't serve other functions. Lots of TM-style themes go even more conservative than that — using constant for A and B but using the variable scope for C.

I'm not arguing that any of this is necessarily wrong, and I'm happy to see how it shakes out, but that was the thinking behind the original decision.

[savetheclocktower]

Was this section not needed?

[mauricioszabo]

There were some misbehaviors on the string matching - some interpolated strings were being detected as non-interpolated, and vice-versa. When I fixed these, I extracted most of the changes into specific matches.

In this specific case, I did not remove the match of double-quoted strings - I just removed the restriction on the previous rule (https://github.com/pulsar-edit/pulsar/pull/554/files#diff-82574a2099439cbee59d5f974dbe593145214402078d948cf9f49378d9637e40R264-L266) and that basically became the exact same match that we have here :D

[Spiker985]

Perhaps this PR can be updated since we've had quite a lot of Tree-Sitter related fixes implemented since this was originally opened?