C Level Iterators and Userland Classes

php.internals

l0t3k

20 years ago
A quick question about using iterators in C code: If it sufficient to use the interface gotten from zend_class_entry->get_iterator to handle userland classes which implement Iterator ? The attachment is part of code i'm using to do Unicode based collation. Will this work for userland iterators as well ? l0t3k

Marcus Börger

20 years ago
Hello l0t3k, first of all your question does not make any sense to me whatsoever. Anyway looking at the code you seem to be using a c-level iterator manually instead of using it inside foreach() construct. The first thing i can see is that you use the iterator index while the source code (zend_interface.h) says: ulong index; /* private to fe_reset/fe_fetch opcodes */ Unil now noone tried to emulate this stuff so i am not sure whether this is correct or not. But one thing i can say is that your usage revealed an error in the engien code. Index is incremented to early. Which is fixed by now. The second thing is that you did not use the dtor() handler so you have a memory leak. Otherwise i cannot see anything wrong with your iterator usage. And the more i think over it you are probably right in using that index. To try to answer what i think you want to know: What ever object you use. If it implements Traversable or derived its zend_class_entry has a get_iterator function that gets you access to an iteator struct and the handler table. You should never touch the data member of that iterator struct and the handlers are documented. In your case there is no need to use handler invalidate_current(). hope this helps marcus Wednesday, February 15, 2006, 6:23:44 PM, you wrote:
> A quick question about using iterators in C code:
> If it sufficient to use the interface gotten from > zend_class_entry->get_iterator to handle userland classes which implement > Iterator ? > The attachment is part of code i'm using to do Unicode based collation. Will > this work for userland iterators as well ?
> l0t3k
Best regards, Marcus

l0t3k

20 years ago
"Marcus Boerger" <helly@php.net> wrote in message news:1355657626.20060215225041@marcus-boerger.de...
> Hello l0t3k, > > first of all your question does not make any sense to me whatsoever.
What im trying to allow for is sorting by iterator as well as standard arrays, for example : // assume en_US as default locale $coll = new Collator("en_US"); $iter = new TextIterator("The rain in Spain drop gently on the Plains", TextIterator::WORD); var_dump( $coll->sort($iter)); // returns the sorted list of words $expr = new Regex($patttern, $input, $flags); var_dump($coll->sort($expr, SORT_DESC)); // Sorts matches in descending order
> Anyway looking at the code you seem to be using a c-level iterator > manually instead of using it inside foreach() construct.
see above...
> The first thing i can see is that you use the iterator index while the > source code (zend_interface.h) says: > ulong index; /* private to fe_reset/fe_fetch opcodes */ > Unil now noone tried to emulate this stuff so i am not sure whether this > is correct or not.
i wrote this more than ayear ago and i remember copying it from engine code.
> The second thing is that you did not use the dtor() handler so you have > a memory leak.
which dtor handler ?
> To try to answer what i think you want to know: What ever object you use. > If it implements Traversable or derived its zend_class_entry has a > get_iterator function that gets you access to an iteator struct and the > handler table. You should never touch the data member of that iterator > struct and the handlers are documented. In your case there is no need to > use handler invalidate_current().
That helps. Thanks a lot.... clayton

Marcus Börger

20 years ago
Hello l0t3k, Anrei, Wednesday, February 15, 2006, 11:28:15 PM, you wrote:
> "Marcus Boerger" <helly@php.net> wrote in message > news:1355657626.20060215225041@marcus-boerger.de... >> Hello l0t3k, >> >> first of all your question does not make any sense to me whatsoever. > What im trying to allow for is sorting by iterator as well as standard > arrays, for example : > // assume en_US as default locale
> $coll = new Collator("en_US");
> $iter = new TextIterator("The rain in Spain drop gently on the Plains", > TextIterator::WORD);
> var_dump( $coll->sort($iter)); // returns the sorted list of words
> $expr = new Regex($patttern, $input, $flags); > var_dump($coll->sort($expr, SORT_DESC)); // Sorts matches in descending > order
>> Anyway looking at the code you seem to be using a c-level iterator >> manually instead of using it inside foreach() construct. > see above...
ic, sounds interesting. Actually Andrei doesn't it look like something for ext/unicode?
>> The first thing i can see is that you use the iterator index while the >> source code (zend_interface.h) says: >> ulong index; /* private to fe_reset/fe_fetch opcodes */ >> Unil now noone tried to emulate this stuff so i am not sure whether this >> is correct or not. > i wrote this more than ayear ago and i remember copying it from engine code.
>> The second thing is that you did not use the dtor() handler so you have >> a memory leak. > which dtor handler ?
Zend/zend_iterators.h: typedef struct _zend_object_iterator_funcs { /* release all resources associated with this iterator instance */ void (*dtor)(zend_object_iterator *iter TSRMLS_DC); ^^^that one :-)
>> To try to answer what i think you want to know: What ever object you use. >> If it implements Traversable or derived its zend_class_entry has a >> get_iterator function that gets you access to an iteator struct and the >> handler table. You should never touch the data member of that iterator >> struct and the handlers are documented. In your case there is no need to >> use handler invalidate_current().
> That helps. Thanks a lot....
> clayton
>> hope this helps >> marcus >> >> Wednesday, February 15, 2006, 6:23:44 PM, you wrote: >> >>> A quick question about using iterators in C code: >> >>> If it sufficient to use the interface gotten from >>> zend_class_entry->get_iterator to handle userland classes which >>> implement >>> Iterator ? >>> The attachment is part of code i'm using to do Unicode based collation. >>> Will >>> this work for userland iterators as well ? >> >>> l0t3k >> >> Best regards, >> Marcus
Best regards, Marcus

l0t3k

20 years ago
> ic, sounds interesting. Actually Andrei doesn't it look like something for > ext/unicode?
that's what it's for, actually...
>>> The second thing is that you did not use the dtor() handler so you have >>> a memory leak. >> which dtor handler ? > > Zend/zend_iterators.h: > typedef struct _zend_object_iterator_funcs { > /* release all resources associated with this iterator instance */ > void (*dtor)(zend_object_iterator *iter TSRMLS_DC); > > ^^^that one :-)
Solid ! Thanks again.. clayton

l0t3k

20 years ago
Marcus, one last question ...
> You should never touch the data member of that iterator
does this mean that i should make a copy of whatever i get from the get_current_data() handler ? There seems to be a bit of inconsistency in dealing with the current element in the iterators For example, the TextIterator reuses the same zval value, whereas the classes in SPL call zval_ptr_dtor() between iterations. clayton

Marcus Börger

20 years ago
Hello l0t3k, in SPL the iterator classes can be overloaded in TextIterator not. Foreach does what is necccessary and especially if inside foreach() someone uses the value it is being copied automatically. Since you do it all yourself you also need to do the copying yourself. So there is no inconsistency, just diiferent usage. regards marcus Friday, February 17, 2006, 1:09:37 AM, you wrote:
> Marcus, > one last question ...
>> You should never touch the data member of that iterator
> does this mean that i should make a copy of whatever i get from > the get_current_data() handler ? There seems to be a bit of inconsistency > in dealing with the current element in the iterators For example, the > TextIterator > reuses the same zval value, whereas the classes in SPL call zval_ptr_dtor() > between iterations.
> clayton
Best regards, Marcus