Request: (fortran) Add support for FORTRAN 77 style comments

[interkosmos]

The style for source code in Fortran accepts only ! as a comment identifier, not the old C as in FORTRAN 77 and earlier. Is it possible to treat C as the beginning of a comment?

[joshgoebel]

Please show us some example code.

[interkosmos]

Please see this example:

<pre><code class="fortran">
C
C  This program in FORTRAN 77 outputs "Hello, World!".
C
      PROGRAM MAIN
      WRITE (*, *) 'Hello, World!'
      END
</code></pre>

The C is not recognised as a comment, while ! for Fortran ≥ 90 is.

[joshgoebel]

Can a line in Fortran >= 90 not legitimately start with C?

[interkosmos]

Fortran 90 introduced free-format that requires to use ! for comments. Only in the old fixed source format the C comment style is still allowed. That means, on Fortran 90, it depends on the file ending/the used compiler flag.

[joshgoebel]

It sounds like we can really only support one then since we can’t know if the file is FORTRAN 90 or FORTRAN 77 - so we can’t know which rule to apply. And having two separate FORTRAN languages seems a bit much. Am I missing something?

[joshgoebel]

Fortran 90 introduced free-format that requires to use ! for comments. Only in the old fixed source format the C comment style is still allowed.

I'm not sure you understood. I meant in Fortran 90 can C meaning something else entirely? Like the name of a user defined function, etc... then it's meaning would be ambiguous if you didn't have the compiler flag or file ending, so it'd be impossible for us to know whether the source in question was Fortran 77 or Fortran 90.

[interkosmos]

I'm not sure you understood. I meant in Fortran 90 can C meaning something else entirely? Like the name of a user defined function, etc... then it's meaning would be ambiguous if you didn't have the compiler flag or file ending, so it'd be impossible for us to know whether the source in question was Fortran 77 or Fortran 90.

Fortran is quite backwards compatible. That means, C just indicates a comment and nothing else, even if modern Fortran compiler will complain about the incorrect comment. But there is some ambiguity related to variables and functions, i. e.:

C This is a comment. Now, we assign foo to variable c:
c = foo

[joshgoebel]

Ah, yes since case_insensitive is true then yes, this is one example of the problem I was referring to...

[c/C] This is a comment. Now, we assign foo to variable c:
[c/C] = foo

No solid way to tell a comment from something else.

[interkosmos]

My example is misleading. Of course, writting in FORTRAN 77 requires fixed format, that means the first six characters of each line are reserved for line number, format descriptions, carriage control, and so on (no statements allowed). But that would require to distinguish between FORTRAN 77 and Fortran 90.

FORTRAN 77 fixed format:

C This is a FORTRAN 77 example.
      PROGRAM MAIN
      PRINT *, 'Hello!'
      END

Fortran 90 free format:

! This is a Fortran 90 example.
program main
    print *, 'Hello!'
end program main

The Wikipedia linter seems to get it right.

[joshgoebel]

But that would require to distinguish between FORTRAN 77 and Fortran 90.

Right, that's the real problem.

The Wikipedia linter seems to get it right.

An interesting but not necessarily related data-point. ;-)

[egor-rogov]

Hm, this is another example of problem that can be solved by parametrized grammar. So that one using Olde Fortran can just turn on C-comments.
Can we achieve it by using callbacks (#2261) to call a function provided by plugin?

[joshgoebel]

Hm, this is another example of problem that can be solved by parametrized grammar.

When and where are you imagining you'd pass these params to the grammar? I can see that helping SOME people... but I think a lot of people just want to throw two code snippets all the wall - one of then Fortran 77 and the other Fortran 90 and it magically just works.

<pre><code data-hljs-language="fortran(version:77)">...

??? Or were you imagining GLOBALLY setting such things? That's already SORTA possible if you hacked registerLanguage a bit...

registerLanguage("fortran", fortranLangDefinition, {version: 77})

Can we achieve it by using callbacks (#2261) to call a function provided by plugin?

The conditional rules perhaps... but not passing the params themselves, that's something new (one idea above).