Tony Marston wrote:
"Robert Cummings" <> wrote in message
Then you've got several options:
1) Don't upgrade PHP.
Not an acceptable option.

2) Pick a different hosting provider.
Not an acceptable optional.

3) Fix your scripts.
The scripts aren't broken. It's PHP 6 that's going to be broken.
I think you're missing the point of a full version increase. This is not a minor or micro version change... script breakage is *expected*.

But breakage should be kept to an absolute minimum, and developer laziness or incompetence is not an acceptable excuse.

Not quite true... major version moves are an opportunity to make a break for freedom. All there needs to be is an upgrade path... and that is clearly in play right now with the warning indicating that POSIX regex functions are being deprecated.

You don't think PHP should support legacy cruft in the core forever do you?

Widely use regex functions are not "legacy cruft". Besides, who decides what is "cruft" and should be removed from the language?

They most certainly are cruft... hence the reason they are being removed. The people who decide what is, and is not, cruft are the very same people who are writing the code. If you are not happy with this then there's the age old saying in open source... "put up or shut up".

If unicode support is slopped onto the current POSIX regex functions won't that then make them non-POSIX? Food for thought. Also, why support two libraries for which one is obviously inferior in speed and functionality?

That is why I suggested that instead of dropping the POSIX functions entirely and seriously annoying lots of users, that they should simply be rewritten as wrappers for the PCRE functions. In that way all the calls to ereg_* would still work, but all they would do is immediately call the relevant preg_* function. The small amount of effort that tghis would take would kill two birds with one stone:

(1) There would be only one regex engine to support, which would be PCRE.
(2) Lots of developers would be spared the hassle of modifying their code as all the calls to POSIX functions would still work as expected because the language would redirect to the PCRE function automatically.

This would probably be worse than removing the POSIX functions. POSIX and PCRE I daresay are not completely compatible. At least when you remove the POSIX functions then the problem space is well defined. Suddenly having POSIX regex functions that are really wrappers around PCRE functions may introduce subtle differences in output for the same horde of users but without the same explicability.

I am not suggesting that the POSIX functions be rewritten to deal with unicode as that would require a huge amount of effort, but by redirecting al POSIX calls to the equivalent PCRE function would have the same effect for far less effort.

The choice is simple - either a small amount of effort from a small number of developers, or a large amount of effort from a large number of seriously pissed-off users. Do the maths. It's not rocket science.

This isn't a mathematical problem. It's a question of correctness. I wasn't happy to hear POSIX regex functions were going either, but when I heard the reasoning I did the best thing I could... I fixed my code to prepare for the inevitable. There's no way I'd trust my code to "just work" with POSIX functions redirected through PCRE and so I'd still need to do the same legwork.

Wrapping the POSIX regex functions around PCRE will lead to more problems than it solves IMHO.

Application and Templating Framework for PHP

PHP General Mailing List (
To unsubscribe, visit:

Reply via email to