perlnewmod - preparing a new module for distribution
This document gives you some suggestions about how to go about writing Perl modules, preparing them for distribution, and making them available via CPAN.
One of the things that makes Perl really powerful is the fact that Perl hackers tend to want to share the solutions to problems they've faced, so you and I don't have to battle with the same problem again.
The main way they do this is by abstracting the solution into a Perl module. If you don't know what one of these is, the rest of this document isn't going to be much use to you. You're also missing out on an awful lot of useful code; consider having a look at perlmod, perlmodlib and perlmodinstall before coming back here.
When you've found that there isn't a module available for what you're trying to do, and you've had to write the code yourself, consider packaging up the solution into a module and uploading it to CPAN so that others can benefit.
We're going to primarily concentrate on Perl-only modules here, rather than XS modules. XS modules serve a rather different purpose, and you should consider different things before distributing them - the popularity of the library you are gluing, the portability to other operating systems, and so on. However, the notes on preparing the Perl side of the module and packaging and distributing it will apply equally well to an XS module as a pure-Perl one.
You should make a module out of any code that you think is going to be useful to others. Anything that's likely to fill a hole in the communal library and which someone else can slot directly into their program. Any part of your code which you can isolate and extract and plug into something else is a likely candidate.
Let's take an example. Suppose you're reading in data from a local format into a hash-of-hashes in Perl, turning that into a tree, walking the tree and then piping each node to an Acme Transmogrifier Server.
Now, quite a few people have the Acme Transmogrifier, and you've had to write something to talk the protocol from scratch - you'd almost certainly want to make that into a module. The level at which you pitch it is up to you: you might want protocol-level modules analogous to Net::SMTP which then talk to higher level modules analogous to Mail::Send. The choice is yours, but you do want to get a module out for that server protocol.
Nobody else on the planet is going to talk your local data format, so we can ignore that. But what about the thing in the middle? Building tree structures from Perl variables and then traversing them is a nice, general problem, and if nobody's already written a module that does that, you might want to modularise that code too.
So hopefully you've now got a few ideas about what's good to modularise. Let's now see how it's done.
Before we even start scraping out the code, there are a few things we'll want to do in advance.
Dig into a bunch of modules to see how they're written. I'd suggest starting with Text::Tabs, since it's in the standard library and is nice and simple, and then looking at something like Time::Zone, File::Copy and then some of the
Mail::* modules if you're planning on writing object oriented code.
These should give you an overall feel for how modules are laid out and written.
There are a lot of modules on CPAN, and it's easy to miss one that's similar to what you're planning on contributing. Have a good plough through the modules list and the by-module directories, and make sure you're not the one reinventing the wheel!
You might love it. You might feel that everyone else needs it. But there might not actually be any real demand for it out there. If you're unsure about the demand your module will have, consider sending out feelers on the
comp.lang.perl.modules newsgroup, or as a last resort, ask the modules list at
firstname.lastname@example.org. Remember that this is a closed list with a very long turn-around time - be prepared to wait a good while for a response from them.
Perl modules included on CPAN have a naming hierarchy you should try to fit in with. See perlmodlib for more details on how this works, and browse around CPAN and the modules list to get a feel of it. At the very least, remember this: modules should be title capitalised, (This::Thing) fit in with a category, and explain their purpose succinctly.
While you're doing that, make really sure you haven't missed a module similar to the one you're about to write.
When you've got your name sorted out and you're sure that your module is wanted and not currently available, it's time to start coding.
Originally a utility to convert C header files into XS modules, h2xs has become a useful utility for churning out skeletons for Perl-only modules as well. If you don't want to use the Autoloader which splits up big modules into smaller subroutine-sized chunks, you'll say something like this:
h2xs -AX -n Net::Acme
-A omits the Autoloader code,
-X omits XS elements, and
-n specifies the name of the module.
A module's code has to be warning and strict-clean, since you can't guarantee the conditions that it'll be used under. Besides, you wouldn't want to distribute code that wasn't warning or strict-clean anyway, right?
The Carp module allows you to present your error messages from the caller's perspective; this gives you a way to signal a problem with the caller and not your module. For instance, if you say this:
warn "No hostname given";
the user will see something like this:
No hostname given at /usr/local/lib/perl5/site_perl/5.6.0/Net/Acme.pm line 123.
which looks like your module is doing something wrong. Instead, you want to put the blame on the user, and say this:
No hostname given at bad_code, line 10.
You do this by using Carp and replacing your
carps. If you need to
croak instead. However, keep
die in place for your sanity checks - where it really is your module at fault.
h2xs provides stubs for Exporter, which gives you a standard way of exporting symbols and subroutines from your module into the caller's namespace. For instance, saying
use Net::Acme qw(&frob) would import the
The package variable
@EXPORT will determine which symbols will get exported when the caller simply says
use Net::Acme - you will hardly ever want to put anything in there.
@EXPORT_OK, on the other hand, specifies which symbols you're willing to export. If you do want to export a bunch of symbols, use the
%EXPORT_TAGS and define a standard export set - look at Exporter for more details.
The work isn't over until the paperwork is done, and you're going to need to put in some time writing some documentation for your module.
h2xs will provide a stub for you to fill in; if you're not sure about the format, look at perlpod for an introduction. Provide a good synopsis of how your module is used in code, a description, and then notes on the syntax and function of the individual subroutines or methods. Use Perl comments for developer notes and POD for end-user notes.
You're encouraged to create self-tests for your module to ensure it's working as intended on the myriad platforms Perl supports; if you upload your module to CPAN, a host of testers will build your module and send you the results of the tests. Again,
h2xs provides a test framework which you can extend - you should do something more than just checking your module will compile.
If you're uploading to CPAN, the automated gremlins will extract the README file and place that in your CPAN directory. It'll also appear in the main by-module and by-category directories if you make it onto the modules list. It's a good idea to put here what the module actually does in detail, and the user-visible changes since the last release.
Every developer publishing modules on CPAN needs a CPAN ID. See the instructions at
http://www.cpan.org/modules/04pause.html (or equivalent on your nearest mirror) to find out how to do this.
perl Makefile.PL; make test; make dist
h2xs has done all the work for you. It produces the standard
Makefile.PL you'll have seen when you downloaded and installs modules, and this produces a Makefile with a
Once you've ensured that your module passes its own tests - always a good thing to make sure - you can
make dist, and the Makefile will hopefully produce you a nice tarball of your module, ready for upload.
The email you got when you received your CPAN ID will tell you how to log in to PAUSE, the Perl Authors Upload SErver. From the menus there, you can upload your module to CPAN.
Once uploaded, it'll sit unnoticed in your author directory. If you want it connected to the rest of the CPAN, you'll need to tell the modules list about it. The best way to do this is to email them a line in the style of the modules list, like this:
Net::Acme bdpOP Interface to Acme Frobnicator servers FOOBAR ^ ^^^^^ ^ ^ | ||||| Module description Your ID | ||||| | ||||\-Public Licence: (p)standard Perl, (g)GPL, (b)BSD, | |||| (l)LGPL, (a)rtistic, (o)ther | |||| | |||\- Interface: (O)OP, (r)eferences, (h)ybrid, (f)unctions | ||| | ||\-- Language: (p)ure Perl, C(+)+, (h)ybrid, (C), (o)ther | || Module |\--- Support: (d)eveloper, (m)ailing list, (u)senet, (n)one Name | \---- Development: (i)dea, (c)onstructions, (a)lpha, (b)eta, (R)eleased, (M)ature, (S)tandard
plus a description of the module and why you think it should be included. If you hear nothing back, that means your module will probably appear on the modules list at the next update. Don't try subscribing to
email@example.com; it's not another mailing list. Just have patience.
If you have a burning desire to tell the world about your release, post an announcement to the moderated
Once you start accumulating users, they'll send you bug reports. If you're lucky, they'll even send you patches. Welcome to the joys of maintaining a software project...
perlmod, perlmodlib, perlmodinstall, h2xs, strict, Carp, Exporter, perlpod, Test, ExtUtils::MakeMaker, http://www.cpan.org/ , Ken Williams' tutorial on building your own module at http://mathforum.org/~ken/perl_modules.html