Front page | perl.perl5.porters |
Postings from October 2009
On Wed, Oct 28, 2009 at 10:11:13AM +0100, demerphq wrote: > 2009/10/28 Nicholas Clark <nick@ccl4.org>: > > On Tue, Oct 27, 2009 at 03:31:46PM -0500, David Nicol wrote: > >> On Tue, Oct 27, 2009 at 11:58 AM, Ben Morrow <ben@morrow.me.uk> wrote: > >> > >> > The overload needs to return either a ref to a REGEXP (a qr//) or a bare > >> > REGEXP. Any further overloading on the return value will be ignored. > >> > >> Is it possible that in the future a coderef that takes the expression > >> as its $_[0] and gives the captures as its returned array might be > >> acceptable? Just asking what seems like the obvious question. > > > > The regexp engine is intertwined with at least half a dozen perl ops. > > It might be possible to write a pluggable regexp engine that accepts code > > references conforming to this sort of API, but > > > > a: I don't know how easy it is > > b: I changed the regexp API somewhat with the move to first class regexps, and > > left that job only part done, because I didn't update the pluggable API. > > Do you recall what might need to be done? It might be as "simple" as 0: See what the APIs documented in maint-5.10 are 1: Find the prototypes of those functions in maint-5.10 2: Find the prototypes of the functions with the same names in blead 3: Figure out how complex the changes are 4: Update the documentation on the obviously simple changes 5: Punt on the rest I'm not sure if there's complication because the type of structure changed from C<struct regexp> to pointer-to-SV which in turn points to a C<struct regexp> (where those are somewhat different C<struct regexp>s) But in turn, I'm not sure how much of those changes were hidden by the macros I wrote, which didn't exist back when Avar did all the work back before 5.10 Nicholas ClarkThread Previous | Thread Next