Texinfo 5.0 in Perl

There was a story on Slashdot on Sunday saying that the new version of Texinfo had replaced the old, C, implementation of makeinfo with one written in Perl.

I thought it would be interesting to look at the Perl they’ve written. This is, after all, a reasonably large example of Perl code that will be getting a bit of attention in the open source[1] world. If you want to look too, you can download the tarball or browse the CVS repository. In both cases, the Perl code is in the directory called ‘tp’.

This first thing to note is that this is obviously code which has been written by programmers who know their craft. This has not been written by script kiddies. There are, however, some rather bizarre touches which imply that the authors don’t know Perl as well as they might hope.

• The code is nicely partitioned into modules. And many of the modules are really classes. But some of the modules (the ones in the init directory) have no package statement. So they are more like Perl4-style libraries than what we’d recognise as modules.
• Many (in fact it might be all) of the subroutines in the modules have prototypes. In many cases, that doesn’t do any harm (although Perl prototypes don’t really address the issues that most people assume they address). But many of these subroutines are called as methods. And prototypes have no effect on method calls at all.
• There is rather more use of package variables (instead of lexical variables) than I’d be comfortable using in my code.
• I can see no use of CPAN modules. Perhaps there are no CPAN modules that help with this code. But I’d find that surprising in a project of this size.

Then I started to wonder which version of Perl they were targetting. So I searched for “use 5.xxx” statements. And found quite a range. Many of the files insist on 5.006 and most of the rest want 5.00405 – which I consider a scarily old version of Perl to try and support. There was one file that wanted 5.007_003. Then in one file (Texinfo::Parser) I found this:

Reading the original Slashdot story, it claimed that one of the improvements in this new version was its Unicode support. And suggesting that you need Perl 5.006 for decent Unicode shows some major misunderstanding. Perl 5.006 was probably the point at which the Perl 5 Porters started to take Unicode seriously. It took until 5.12 or 5.14 before they got it right. Trying to support Unicode properly on anything earlier is almost certainly doomed to failure.

It’s great that another heavily-used project has started to use Perl. But it’s a shame that people might mistake this 5.6-era code for state of the art Perl. I can’t help wondering if this is a symptom of the “Perl 5 can’t have a new version number” problem that I’ve been reading about recently

I’ll get in touch with the Texinfo team and make some suggestions to them, of course.

Update: I emailed the Texinfo mailing list with a link to this blog post. I got a reply from Patrick Dumas who wrote most of this Perl code. His reply is available on the mailing list archives.

[1] Sorry, it’s a GNU project so obviously I mean “free”, not “open source”.