Wednesday, September 8, 2010

New features in PHP 6

As you may be aware the core PHP group of developers all met in Paris on November the 11th and 12th 2005. The minutes from the meeting are fascinating reading, but there is a lot to go through. So I've gone through all of the points raised and chewed them over from a developers point of view. Your comments as always are welcome.
Before I get started however I'd just like to make one thing very clear: what you read here (or in the original minutes) are in no way the 'fully 100% decided' end results / changes that we'll see in PHP6. They will most likely all be discussed further (on internals and wider), but even so we can take the information presented in the minutes as being the PHP teams most 'current' way of thinking about any given subject.

Unicode

Unicode support at present can be set on a per request basis. This equates to PHP having to store both Unicode and non-Unicode variants of class, method and function names in the symbol tables. In short - it uses up more resources. Their decision is to make the Unicode setting server wide, not request wide. Turning Unicode off where not required can help performance and they quote some string functions as being up to 300% slower and whole applications 25% slower as a result. The decision to move it to the php.ini in my mind does take the control away from the user, and puts it into the hands of the Web Host.
If you compile PHP yourself or are responsible for this on your servers then you may be interested to know that PHP 6 will require the ICU libs (regardless if Unicode is turned on or off). The build system will bail out if the required ICU libs cannot be found. In a nutshell, you'll have another thing to install if you want to compile PHP.

Register Globals to go

Say goodbye folks, this one is finally going. It will no longer be an ini file setting, and if found it will raise an E_CORE_ERROR, pointing you to the documentation on why it's "bad". This means that PHP6 will finally break all PHP3 era scripts (or any script using reg globals) with no recourse at all but to re-code it. That's a bold move, but a needed one.

Magic Quotes to go

The magic quotes feature of PHP will be going, and as with register globals it's going to raise an E_CORE_ERROR if the setting is found anywhere. This will affect magic_quotes, magic_quotes_sybase and magic_quotes_gpc.

Safe Mode to go

This may please developers who have web hosts that insist upon safe mode! But it will now go totally, again raising an E_CORE_ERROR if found. The reason is that apparently they felt it gave the 'wrong signal', implying that it made PHP secure, when infact it didn't at all. open_basedir will (thankfully) be kept.

'var' to alias 'public'

PHP4 used 'var' within classes. PHP5 (in its OO move) caused this to raise a warning under E_STRICT. This warning will be removed in PHP6 and instead 'var' will mean the same thing as 'public'. This is a nice move but I if anyone has updated their scripts to work under E_STRICT in PHP5 it will be a redundant one for them.

Return by Reference will error

Both '$foo =& new StdClass()' and 'function &foo' will now raise an E_STRICT error.

zend.ze1 compatbility mode to go

ze1 always tried to retain old PHP4 behaviour, but apparently it "doesn't work 100%" anyway, so it will be removed totally and throw an E_CORE_ERROR if detected.

Freetype 1 and GD 1 support to go

Support for both of these (very very old) libs will be removed.

dl() moves to SAPI only

Each SAPI will register the use of this function as required, only the CLI and embed SAPIs will do this from now on. It will not be available elsewhere.

FastCGI always on

The FastCGI code will be cleaned up and always enabled for the CGI SAPI, it will not be able to be disabled.

Register Long Arrays to go

Remember the HTTP_*_VARS globals from yesteryear? Well if you're not already using $_GET, $_POST, etc - start doing so now, because the option to enable long arrays is going (and will throw an E_CORE_ERROR).

Extension Movements

The XMLReader and XMLWriter extensions will move into the core distribution and will be on by default.
The ereg extension will move to PECL (and thus be removed from PHP). This means that PCRE will not be allowed to be disabled. This will make way for the new regular expression extension based on ICU.
The extremely useful Fileinfo exntesion will move into the core distribution and enabled by default.

PHP Engine Additions

64 bit integers
A new 64 bit integer will be added (int64). There will be no int32 (it is assumed unless you specify int64)
Goto
No 'goto' command will be added, but the break keyword will be extended with a static label - so you could do 'break foo' and it'll jump to the label foo: in your code.
ifsetor()
It looks like we won't be seeing this one, which is a shame. But instead the ?: operator will have the 'middle parameter' requirement dropped, which means you'd be able to do something like this: "$foo = $_GET['foo'] ?: 42;" (i.e. if foo is true, $foo will equal 42). This should save some code, but I personally don't think it is as 'readable' as ifsetor would have been.
foreach multi-dim arrays
This is a nice change - you'll be able to foreach through array lists, i.e. "foreach( $a as $k => list($a, $b))".
{} vs []
You can currently use both {} and [] to access string indexes. But the {} notation will raise an E_STRICT in PHP5.1 and will be gone totally in PHP6. Also the [] version will gain substr and array_slice functionality directly - so you could do "[2,]" to access characters 2 to the end, etc. Very handy.

OO changes

Static Binding
A new keyword will be created to allow for late static binding - static::static2(), this will perform runtime evaluation of statics.
Namespaces
It looks like this one is still undecided - if they do implement namespaces it will be using their style only. My advice? Don't hold your breath!
Type-hinted Return Values
Although they decided against allowing type-hinted properties (becaue it's "not the PHP way") they will add support for type-hinted return values, but have yet to decide on a syntax for this. Even so, it will be a nice addition.
Calling dynamic functions as static will E_FATAL
At the moment you can call both static and dynamic methods, whether they are static or not. Calling a dynamic function with the static call syntax will raise an E_FATAL.

Additions to PHP

APC to be in the core distribution
The opcode cache APC will be included in the core distribution of PHP as standard, it will not however be turned on by default (but having it there saves the compilation of yet another thing on your server, and web hosts are more likely to allow it to be enabled)
Hardened PHP patch
This patch implements a bunch of extra security checks in PHP. They went over it and the following changes will now take place within PHP: Protection against HTTP Response Splitting will be included. allow_url_fopen will be split into two: allow_url_fopen and allow_url_include. allow_url_fopen will be enabled by default. allow_url_include will be disabled by default.
E_STRICT merges into E_ALL
Wow, this is quite a serious one! E_STRICT level messages will be added to E_ALL by default. This shows a marked move by the PHP team to educate developers on 'best practises' and displaying language-level warnings in a "Hey, you're doing it the wrong way".
Farewell <%
They will remove support for the ASP style tags, but the PHP short-code tag will remain (

Conclusion

PHP6 is taking an interesting move in my mind - it's as if the PHP developers want to now educate developers about the right way to code something, and remove those lingering issues with "Well you SHOULD be doing it this way, but you can still do it the old way". This will not be the case any longer. Removing totally the likes of register globals, magic quotes, long arrays, {} string indexes and call-time-pass-by-references will force developers to clean up their code.
It will also break a crapload of scripts beyond repair that doesn't involve some serious re-writing. Is this a bad thing? I don't think so myself, but I see it making the adoption of PHP6 even slower than that of PHP5, which is a real shame. However they have to leap this hurdle at some point, and once they've done it progression to future versions should be swifter.

No comments:

Post a Comment