On Fri, Sep 5, 2014 at 9:59 PM, Ben Kloosterman <[email protected]> wrote:

> Interesting comment on this very issue and some of the benefits of 2 phase
> .
>
> http://compilers.iecc.com/comparch/article/94-10-101
>

Nice legacy reference!

This morning in the shower another reason occurred to me: legacy languages.
Early languages like FORTRAN and COBOL have positional constraints
governing where tokens may appear. The most obvious of these is that "C" in
column 7 in legacy FORTRAN means that the rest of the row is a comment.

I think it's now generally agreed that we're done with punch cards. While
I'm not inclined to feel overly constrained by these languages 45+ years
later, there is *some* merit to being able to parse them for analysis and
transition purposes.

What this mainly tells me is that we really want it to be *possible* for
the tokenizer to be a separate phase.

It's not that big a deal, in the end. It actually saves me some work if we
declare that the tokenizer should be seperable.


shap
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to