Prepare for PHP 5.3

I wonder how can you suggest to remove short tags, deprecated, and suggest error_reporting modification to avoid deprecated code. PHP 5.3 is here since ages, it is about the time to gradually upgrade codes removing silly, useless, redundant parts. deprecated should be enabled, imho, as every other error level.

@ Andrea Giammarchi: Hi Andrea thanks for your comment.

I totally agree, if you read more carefully you will find that I encourage people to change their code.

But in the case that they run other people\'s code (Like a CakePHP 1.2 Core) chances are you will be running 5.3 to test your own code before you get a chance to upgrade your Cake Core (1.3, 2, or even 3 aren\'t stable yet). So in such cases you may want to turn of deprecated warnings for the framework until you can upgrade to a new release.

Kevin, in my opinion IT means updates. If people are stuck in old Cake version, or whatever old framework it is, they can be stuck with old PHP version and slowly go down behind updated applications/developers. We all complain about IE6 and I do not know how many of \"us\" are still in .NET version 1/2, PHP 4, etc etc ... update your knowledge/skills, this is the only way to survive in IT. Just my 2 cents, good post though in any case.

We just completed PHP 5.3 compatibility support for Zend Framework, and ran into many similar issues to what you reported. In addition, there are a variety of signatures that changed, and many array functions that used to allow passing objects no longer do.

That said, it\'s not a terribly difficult undertaking. You just have to keep in mind the benefits (ability to continue upgrading your PHP version to latest stable; faster execution; better object model; etc), as it can be a pretty boring task, particularly if you have to run a large test suite over and over again. :)

@ Andrea Giammarchi: I don\'t understand your comment, \"PHP 5.3 is here since ages\" -- last I checked, the official release was just a month ago. While it\'s been on the horizon for a good six months, that\'s not the same as saying there\'s been a stable base against which to test development. It will be some time before there is widespread PHP 5.3 adoption.

Last I researched on shorttags the vote was that they they should be kept. See: http://www.php.net/~derick/...

what are your source for this claim?
Cheers

@Matt, I mean these changes are well known since ages and most of them where already possible. I personally use PHP 5.3 since \"ages\" (first alpha) specially to test old code and start to migrate it. This is what I was talking about.

@Rune, vote or not vote, short tag is a mess for modules and a bit arrogant from development point of view. XML starts with <?xml, clear enough, other languages specify who they are ... in 10 years I have never used short tags since I always thought they could have cause mess, indeed you have to be carefull with XML declaration ... is it a good way to code? I do not think so. Same is for ASP tag emulator, so you cannot put both languages in the same page ... good? BAD!

While I agree that people should not use short_open_tag, I want to clarify that:

-- They are not disabled by default in PHP 5.3. The two distributed php.ini-* files do disable them, but the default is still On. Although, of course a package maintainer might choose to enable a certain php.ini file within a package, like the one mentioned here for Ubuntu.

-- They may or may not be removed from PHP 6, a decision that may or may not depend on a decision to decouple <?= from short_open_tag. My opinion is there will be a decoupling.

Also, mysqlnd is not enabled by default although it appears the Ubuntu package maintainer here did (which is a good thing).

See Also: http://php.net/migration53

I wish PHP failed more loudly when short tags are disabled. I\'ve had them turned off on my local server for a while, but it always leads to some odd results when I download code that uses short tags.

I was using ereg_*() functions and I have to convert to preg_*() ones. Fortunetale magic_quotes_gpc and register_globals are gone from a long time.
@Matthew: when Zend_Loader will support namespaced classes?

@ Andrea Giammarchi: No sorry I still think you\'re missing my point. As we speak, CakePHP people are working on PHP 5.3 versions. I just want to enable people to testdrive their code right now even though they don\'t have a stable 5.3 framework core - yet.
So comparison with IE6 isn\'t really working as I\'m encouraging people to update everything. Please read my article more carefully, I have the feeling your IE6 concerns where triggered by false alarm.

@ Matthew Weier O\'Phinney: That\'s a nice summary of of PHP 5.3 goodness. I\'m actually the most excited about this release since... Well a very long time : )

@ Rune Kaagaard: Thanks. I read it somewhere, but apparently the jury is still out on that one, let me adjust the article

@ Philip Olson: Thanks you for clarifying on that subject, I will adjust the article accordingly

@ Cory Kaufman: I agree that would be nice, but I imagine a fix for that would be really hacky. I mean: it either recognizes shorttags or it doesn\'t. A state inbetween that detects but does not act on it may introduce bugs on it\'s own.

@ Giorgio Sironi: Thanks I will add them as well!

i really don\'t see a reason for dropping short tag support. writing php a 1000 times is a nonsense

@ xrado: Well with any modern IDE you shouldn\'t have to type that manually.

They probably should have never supported it in the first place. PHP isn\'t the only language out there. Just think of XML:

[CODE=\"XML\"]
<?xml version=\"1.0\" encoding=\"UTF-8\"?>
[/CODE]

This can cause confusion, so I understand why they want to drop it.

@Philip Olson, Thx for clarifying. Interesting to hear about the possible decoupling of the <?=$some_var?> syntax.

@Andrea Giammarchi, how is short tag a mess for modules? (and what kind of modules are you referring to?). I have always felt that having the ease of the <?=$some_var?> syntax _by far_ made the impurity of having to echo \"<?xml\" ... worth it. PHP is a super cool template language where shorttags fits in nicely. so shortags....? GOOD! :)

Hey Kevin, good post. I just wanted to let you know that 5.3 support has now been pushed to the CakePHP 1.3 branch, which will be released shortly, so core code changes shouldn\'t be be required soon.

@ Nate Abele: Thanks for the heads-up. I saw on twitter that the dev version of 1.3 was released today. Nice job! Were you able to get your hands on some Club Mate for that ? ; )

You are right about the short tags, but not completely. Check out the stream wrappers php provides. Short tags can be used within streams, even if they are disabled. This is a good way to allow them in views only. Disabling them by default is a goog call IMO.

I recently wanted to get rid of the short tags. There was really much code and I didn\'t write too much of it, so I couldn\'t rely on a regexp. I looked around and found this:
http://stackoverflow.com/qu...

Well, I put it into a script and called it like that:
[CODE=\"sh\"]
find project/dir/ -type f -iname \"*.php\" -exec php -d short_open_tag=On the_script.php {} \\;
[/CODE]
and the_script.php would contain:
[CODE=\"php\"]
<?php

$file=$argv[1];
echo \"Replacing short open tags in \\\"$file\\\"...\";

$content = file_get_contents($file);
$tokens = token_get_all($content);
$output = \'\';

foreach($tokens as $token) {
if(is_array($token)) {
list($index, $code, $line) = $token;
switch($index) {
case T_OPEN_TAG_WITH_ECHO:
$output .= \'<?php echo \';
break;
case T_OPEN_TAG:
$output .= \'<?php \';
break;
default:
$output .= $code;
break;
}

}
else {
$output .= $token;
}
}
[/CODE]

As you can see it\'s using tokenizer parser, so it could save you some headache. (Unless you have *.php files that can not be parsed etc.)
The only problem was that it stripped some newlines even after long tags, so the svn diff got really huge. I think i can live with this though.

Another problem I faced switching to PHP5.3 was rather odd:
I had a php class named SplQueue that was autoloaded if SPL didn\'t already provide it, which was always the case - until PHP5.3. It used the same methods and I planed that it would silently be replaced by the SPL one but:
Suddenly there where exceptions all over the place because SplQueue throws (undocumented) Exceptions when dequeue-ing from an empty queue. That\'s quite silly - but OK, I figured it would be easiest just not to use the SPL-Class. So be aware that the SPL docs are... erm... not so well-proven yet.

@ Bernhard H.: That\'s very useful information Bernhard, thanks a lot for this!