Support more compiler directives

[nberth]

Adds support for >>DEFINE and simple (cascading) conditional compilation directives. >>SET directives are parsed but not interpreted yet. >>EVALUATE/>>WHEN shall be added in a subsequent PR.

[emilienlemaire]

Would this handle #153 ?

[nberth]

Would this handle #153 ?

No, the >>D is a floating indicator, not a compiler directive per se. Introducing that one may need quite some care (and we don't have that many examples of its usage I think), so it's left for later.

[emilienlemaire]

We would like to be able to parse the whole https://github.com/OCamlPro/gnucobol-contrib/tree/1618cb70d3855fd56a17bb2522c421e38993fc6d/samples/worldcities folder, (with the fixes from OCamlPro/gnucobol-contrib#1), and this is used in some files

[nberth]

We would like to be able to parse the whole https://github.com/OCamlPro/gnucobol-contrib/tree/1618cb70d3855fd56a17bb2522c421e38993fc6d/samples/worldcities folder, (with the fixes from OCamlPro/gnucobol-contrib#1), and this is used in some files

I see in there it's not used in floating position… so maybe Boris's suggestion could work (as a temporary measure).

[GitMensch]

No, the >>D is a floating indicator, not a compiler directive per se. Introducing that one may need quite some care (and we don't have that many examples of its usage I think), so it's left for later.

I think only GnuCOBOL supports that and the features is removed in COBOL 2014, so handling this specific one is likely not important.

[emilienlemaire]

Seems all good to me, but one question concerning a test failure

[emilienlemaire]

All good for me 👌

[GitMensch]

>>EVALUATE/>>WHEN shall be added in a subsequent PR.

I'd love a PR for GnuCOBOL on that as well :-)

Concerning this PR: Do you want to do a follow-up for the MF conditional compilation?

    78  my-define VALUE 5.
      $IF my-define
          DISPLAY 'something'
      $END

[nberth]

Concerning this PR: Do you want to do a follow-up for the MF conditional compilation?

    78  my-define VALUE 5.
      $IF my-define
          DISPLAY 'something'
      $END

The $-variants could be handled in this PR.
The handling of 78-level entries on the pre-processor side will come in another PR, though.

[GitMensch]

There are some issues on the MF side (see comments), but I'm mostly concerned about ISO.
I don't know if there's a preparser applied or if the compdir_grammar.mly keywords are taken "verbatim".

ISO says:

A compiler directive is composed of the compiler directive indicator >>, optionally followed by the COBOL character space, followed by compiler-instruction. The compiler directive indicator shall be treated as though it were followed by a space if no space is specified after the indicator.

Also note that "the COBOL character space" can consist of multiple character spaces.

I do wonder: does

>> IF something DEFINED
    PERFORM something
  >>  ELSE
    PERFORM other-stuff
   >>END-IF

works.

[nberth]

There are some issues on the MF side (see comments), but I'm mostly concerned about ISO. I don't know if there's a preparser applied or if the compdir_grammar.mly keywords are taken "verbatim".

Those >> and $-prefixed keywords are recognized during a lexing phase that occurs before, and interveaning spaces are removed at that stage.

ISO says:

A compiler directive is composed of the compiler directive indicator >>, optionally followed by the COBOL character space, followed by compiler-instruction. The compiler directive indicator shall be treated as though it were followed by a space if no space is specified after the indicator.

Also note that "the COBOL character space" can consist of multiple character spaces.

I do wonder: does

>> IF something DEFINED
    PERFORM something
  >>  ELSE
    PERFORM other-stuff
   >>END-IF

works.

For that, I originally followed GnuCOBOL. The version I used did not allow more than one character, so >> ELSE might not be recognized. I think that's an easy PR — but maybe that was fixed recently.

[GitMensch]

For that, I originally followed GnuCOBOL. The version I used did not allow more than one character, so >> ELSE might not be recognized. I think that's an easy PR — but maybe that was fixed recently.

Oh, please check that then.

Concerning this PR it looks good to me now.