[pmwiki-users] Re: recipes on core elements

[Joachim Durchholz]

Radu wrote:
> I'll just go ahead and implement it, then put it up and let maintainers 
> comment/improve on it if they feel like.

Do as you please :-)

I predict all kinds of trouble though. Either (1) you won't detect all 
relevant changes, or (2) you'll detect changes that are irrelevant, or 
(3) you'll be in for a whole lot of dependency tracing in PmWiki.

Examples for the cases:
1) You may overlook a function that your code is calling or overriding. 
Might not be an issue if you're calling/overriding something very 
simple, of course (but then checking whether a change induces breakage 
isn't that difficult if you do it manually - IOW this case limits what 
you can gain).
OTOH after fifteen revisions of a more advanced (read: complicated) 
recipe, done by more than one recipe author, you'll have difficulties 
tracking the list of called/overridden functions. In other words, you'll 
get into a maintenance problem - in a year or two.
2) A function that you're calling/overriding is modified to fix a bug. 
Or to extend its functionality (none of which is needed by your code).
3) Your recipe is calling/overriding foo(), which in turn calls bar(). 
Or foo() may call things through a table, which means you have to 
monitor the table's contents (that latter case probably isn't very 
common, but it's definitely a boundary where watching PmWiki's source 
for changes doesn't help anymore).

The issue isn't that it isn't useful - you'll certainly get some 
warnings that will help you along. To me, it's unclear that the effort 
is worth the effect though. Well, maybe it's worth the effort if you're 
checking just very few very simple functions that are unlikely to change.
My stance is also somewhat weak because I don't have a better 
alternative. So if you feel it's necessary, go ahead :-)

One last advice: it's probably better to do that function change 
checking during installation. Checking whenever pmwiki.php is called up 
will just burden the system. (Consider others who start to use the same 
kind of checking - imagine a dozen or more recipes, all scanning large 
parts of the PmWiki sources to see whether there are changes. It's not 
*too* bad - markup table construction and application probably takes 
more time - but it's an unnecessary inefficiency.)

Regards,
Jo