Re: Bug with PROCINFO["identifiers"]

[arnold]

Hi.

I took a harder look at this.  It's a difficult problem since awk
is so dynamic.  For example:

        $ cat y.awk
        BEGIN {
                if (a)
                        foo(b)
                else
                        bar(b)
                print typeof(b)
        }

        function foo(b)
        {
                b[1] = 1
        }

        function bar(b)
        {
                b++
        }
        $ ./gawk -f y.awk -v a=1
        array
        $ ./gawk -f y.awk -v a=0
        unassigned

I can improve the handling for arrays, but that's about it.

Thanks,

Arnold
---------------------------------------
diff --git a/awkgram.y b/awkgram.y
index cfc12e78..72d79312 100644
--- a/awkgram.y
+++ b/awkgram.y
@@ -2147,7 +2147,7 @@ simple_variable
          {
                char *arr = $1->lextok;
 
-               $1->memory = variable($1->source_line, arr, Node_var_new);
+               $1->memory = variable($1->source_line, arr, Node_var_array);
                $1->opcode = Op_push_array;
                $$ = list_prepend($2, $1);
          }


arnold@skeeve.com wrote:

> Hello.
>
> Thanks for taking the time to send in a bug report.
>
> anoncoward--- via "Bug reports and all discussion about gawk." 
> <bug-gawk@gnu.org> wrote:
>
> > If one has a user defined variable (scalar or array), 
> > PROCINFO["identifiers"] always returns ``untyped'' 
> > even if the type is known.
> >
> > (Hushmail will wrap the lines.  My apologies.)
> >
> > $ LC_ALL=C ./gawk 'BEGIN {s="a"; i=1; ar[1]=3; print 
> > PROCINFO["identifiers"]["s"], PROCINFO["identifiers"]["i"],  
> > PROCINFO["identifiers"]["ar"] }'
> > untyped untyped untyped
> >
> > but the program obviously knows that s and i are scalars and ar is 
> > an array.  For example,
> >
> > $ LC_ALL=C ./gawk 'BEGIN {s="a"; i=1; ar[1]=3; print typeof(s), 
> > typeof(i),  typeof(ar) }'
> > string number array
> >
> > Per your request, I compiled the latest version of gawk and tested 
> > it.
>
> PROCINFO["identifiers"] is built after the program is parsed but
> before it starts to run.  typeof is called during execution, when
> the variables types have settled down, so it "knows" more.
>
> The manual is pretty clear about this.
>
> But that said, it may be possible to make gawk a little bit smarter
> about this, so that it says "untyped" in fewer instances. I will
> take a look at it.
>
> Thanks,
>
> Arnold