Front page | perl.perl5.porters |
Postings from January 2011
On Sat, Jan 29, 2011 at 09:46:32PM -0500, Eric Brine wrote: > On Sat, Jan 29, 2011 at 5:48 PM, Mark Overmeer <mark@overmeer.net> wrote: > > > The unsplit POSIX.pm file is 1024 lines (21kb). It does not use function > > signatures, but checks parameter list length manually... > > > > There's no way to simply check parameter list length using "function > signatures", if by that you mean prototypes, Indeed, given that prototype checking can be disabled by calling using a & If it *matters* that the parameter list length is correct, then it still needs to be checked, even with prototypes. > Attached is my compact rework of the POSIX module (434 lines, 15kb), as > > > example how small the module actually is. No regression tests run yet. > > All fancy tricks removed. But it may trigger some debate first ;-) > > > > What about the fancy tricks you added? Prototypes change parsing rules, so > by definition, it is a backwards incompatible change. For example, your > changes causes > > perl -e"use POSIX qw( ); @args=2; POSIX::exit(@args);" > > to exit with code 1 instead of 2. And also breaks any existing code which has marshaled the arguments for a 2 or 3 argument function into an array before calling it. At least it breaks *that* at compile time. I'm really not convinced that prototypes (as is) add any value in the general case. To my mind, they complicate reading the code, because they provide action-at-a-distance on how the parser is going to treat function calls. There seemed to be a fashion for using them extensively around 5.003-5.004 time (for example, a lot of the regen code in the core), and looking at that code now, it feels dated to me, and lacking some KISS. Nicholas ClarkThread Previous | Thread Next