POC make fmt"{foo}\n{\bar}" do the right thing

[RedBeard0531]

This is a POC. If the general concept is approved, I'll test, document, etc.

This is an alternative to #7070

[RedBeard0531]

I had to move this above the runableExamples so the diff is huge. The only actual change was adding the makeStrLit template and calling on lines 252 and 285

[RedBeard0531]

This check was added in commit 2 to improve the error message when " is used incorrectly

[Araq]

Not sure if that is its purpose. I imagined it to use some lowest common denominator escape rules.

[RedBeard0531]

I'm not too opinionated on where this functionality belongs, but it does seem useful in stdlib. Does macros.nim sound better since it is most useful when writing macros?

[dom96]

So it's either:

• % and fmt

or

• just fmt but with modified semantics to get rid of the raw string literal semantics?

[GULPF]

@dom96 This PR stills keeps %. It fixes the case where the text outside of {} needs to be treated as a normal string, while the text inside {} needs to be treated as a raw string (note that this is not the same thing as getting rid of the raw string literal semantics).

This is not possible to solve with %, since both %"" and %r"" would give the wrong result. To make "{foo}\n{\bar}" work with % you would need to use %"{foo}\n{\\bar}" instead, which is a bit silly and might be impossible to syntax highlight in a sane way.

[dom96]

To make "{foo}\n{\bar}" work with % you would need to use %"{foo}\n{\\bar}" instead, which is a bit silly and might be impossible to syntax highlight in a sane way.

In what way is that silly? That's what I would expect. If you want a literal \ in your string you need to escape it.

[RedBeard0531]

I think it is a distinction in whether you view it as "writing a string that is passed to the fmt macro" or "writing an fmt string". I think your description makes sense if you think of it like the former, which is technically correct but less useful. I think we should make fmt work like the later, similar to how when using re"" you think about writing a regex rather than writing a string that gets passed to re. Isn't that the goal when writing a DSL?

[GULPF]

It becomes a big problem with syntax highlighting. A good highlighter should highlight text inside {} as code, but that means that there will be no visual indication that \ is actually the start of an escape sequence. Besides, fmt"{foo}\n{\bar}" uses the raw string syntax so there is no reason to assume that the string follows normal escape rules (I would assume the opposite even - that it follows the escape rules of a raw string).

[Araq]

Thank you for this PR but we decided for & or .fmt which doesn't require this additional complexity.