Re: [plt-scheme] eval question

[John Clements <clements@brinckerhoff.org>]

Title: Re: [plt-scheme] eval question

> > >(with-input-from-file path
> > >   (lambda ()
> > >     (eval (read))))
>
> Of course, I'll do this with load as was promised and is much
> simpler,
> but:
>
> > Not only does this strike me as a bad idea from a
> > thinking-about-your-program-and-documenting-it-clearly
> standpoint,
>
> what, exactly, should be documented? What this little bit of code
> does
> is perfectly clear from an R5RS point of
> view.

What's not documented is the settings that are stored in the file that you're loading. I can see this causing a bunch of problems down the line. Let me use the first person, to make this less acerbic. I've written code like this before, and I always end up forgetting what settings were in the file, and having to go back and check it.  If I have more than one such file, I'm really in the soup.

In essence, I think a solution using modules (or units, if you prefer them) is more robust and inherently better-documented.

You'll have to pardon my knee-jerk reaction; I've just spent a few weeks grading assignments for our software design class, and I've seen a lot of extremely non-portable, non-modular, non-documented, poorly designed code.

john clements