Re: hints and constraints

Lester Caine wrote on 05/02/2015 14:51:

On 05/02/15 14:24, Rowan Collins wrote:

There is nothing new about PHP's userland int type being 64-bit on 64-bit platforms. For instance, raising 2 to the power of 62 returns exactly the same thing on every version of PHP back to 4.3.0: http://3v4l.org/VBMbv

Unfortunately, that's not true for Windows, see <http://windows.php.net/download/#x64>: | The x64 builds of PHP for Windows should be considered experimental, | and do not yet provide 64-bit integer or large file support. Please | see this post for work ongoing to improve these builds.

Yes, the picture on Windows is rather different, hence my second comment that most production builds of PHP are probably on (64-bit) Linux. The point is that writing code in PHP and assuming integers will overflow after 32 bits has been a bad idea for a long time, outside of really unusual cases like COM integration [1], where there's a valid reason to assume you'll never want to run it on Linux. [1] http://php.net/manual/en/book.com.php

This is all past history ... Those of us who have had to support windows only sites in the past have often run 32 bit builds of Linux simply to maintain some level of compatibility. It is the growing switch to 64bit builds of PHP5 that are highlight many of the problems that we NOW want to fix things for which wrong decisions in the past are catching up with us. Overflowing to float had good reason for some users but not for others and now is the time to sort out the whole mess.

That sounds like a pretty good argument in favour of Andrea's RFC to me.

The simple answer here is that there is not a 'single' definition of integer ...

True. But the definition of "integer" in PHP is, and has been for many years, "as big as this build can handle". With Andrea's patch, all systems can handle really really big integers, which seems like a big win to me. Regards, -- Rowan Collins [IMSoP]