#53405 bugfix: current in foreach (by ptrofimov,dmkuznetsov)

[ptrofimov]

#53405 bugfix: current in foreach (by ptrofimov,dmkuznetsov)

[lstrojny]

Could you please add a test for this bugfix?

[lstrojny]

ping @ptrofimov

[lstrojny]

ping again @ptrofimov

[ptrofimov]

@lstrojny Oh, sorry. I thought everybody forgot about this pull request)) I'll add tests in few days.

[lstrojny]

Nope. We never forget (alright, sometimes).

[nikic]

I don't think that this change is right. I'm pretty sure that the function was intentionally marked ZEND_SEND_PREFER_REF, just like key() was also.

This will change the behavior and it's unclear to me what behavior is the right one. At least this needs a bit more explanation.

If this is about making current()/key() independent of foreach() then this will only help in a limited way (esp when foreach-by-ref or foreach-over-ref is used). I have a patch that properly decouples foreach from the internal array pointer, but that's a rather more heavy change.

[lstrojny]

@nikic thanks for the explanation. Do you mind closing it and open a different issue for your refactoring?

[ptrofimov]

Well, the problem was that the code

$a = array(1, 2, 3);
foreach ($a as $i) {
    var_dump(current($a));
}

should give 1,1,1. But it gives 2,2,2. After this patch it gave 1,1,1.
But just the other day I downloaded latest source and wrote the test for this, and, unfortunately, it gives 2,3,false.
Why it is so, I don't know, may be something changed in new version.

Well, perhaps, we should close the issue((

[olekukonko]

I don't think it should be closed until its fixed. Has the same issue once it was a mess because it gives different results http://3v4l.org/sBDjl#v430. Can you also look at http://stackoverflow.com/a/14849560/1226894 and confirm how a function affect current output when working with foreach

[nikic]

@lstrojny What I have was basically a try to improve performance by avoiding the array copy in foreach; the decoupling from the internal array pointer was more of a sideeffect. And as I didn't yet see performance benefits for "real" code (only in artificial benchmarks) I'm not sure we want to do that change.

[jpauli]

Closing as this subject is debatted against PHP7 actually, with a more detailed patch.