TextIterator changes

php.internals

Andrei Zmievski

20 years ago
I am working on implementing BreakIterator API [1]. I considered two approaches: making a separate class or merging the API into the existing TextIterator. Having a separate class would be a bit cleaner, but I can see people wanting to use it in foreach(), and since TextIterator already provides a lot of BreakIterator's functionality, I decided that merging would be the best option. However, there is an overlap between the BreakIterator API and the current TextIterator one, so there will have to be some changes. 1. TextIterator::current() signature will change from: mixed current() to: mixed current(integer &$offset) in order to support BreakIterator's functionality of returning offset in current(). 2. TextIterator::next() will return the offset of the next boundary instead of returning nothing. 3. TextIterator::rewind() will be renamed to TextIterator::first() to conform to BreakIterator's first()/last() API. So this is heads up. Let me know if you have a problem with this. -A [1] http://icu.sourceforge.net/apiref/icu4c/ubrk_8h.html

Michael Wallner

20 years ago
Andrei Zmievski wrote:
> I am working on implementing BreakIterator API [1]. I considered two > approaches: making a separate class or merging the API into the existing > TextIterator. Having a separate class would be a bit cleaner, but I can > see people wanting to use it in foreach(), and since TextIterator > already provides a lot of BreakIterator's functionality, I decided that > merging would be the best option. However, there is an overlap between > the BreakIterator API and the current TextIterator one, so there will > have to be some changes. > > 1. TextIterator::current() signature will change from: > > mixed current() > > to: > > mixed current(integer &$offset)
This will raise the same issue as we have/had with the reflection API: mike@honeybadger:~/build/php-unicode-debug$ cli -r 'interface i{function f();} class c implements i{function f($a){}}' Fatal error: Declaration of c::f() must be compatible with that of i::f() in Command line code on line 1 Nah, don't look at me--I don't like that either.
> 3. TextIterator::rewind() will be renamed to TextIterator::first() to > conform to BreakIterator's first()/last() API. > > So this is heads up. Let me know if you have a problem with this.
Huh? Rename or alias? It can't implement Iterator if there's no rewind() method. Regards,
-- Michael

Andrei Zmievski

20 years ago
TextIterator does not implement Iterator interface, only Traversable. It just happens to have functions of the same name as Iterator. -Andrei On Jun 23, 2006, at 2:16 PM, Michael Wallner wrote:

Michael Wallner

20 years ago
Andrei Zmievski wrote:
> TextIterator does not implement Iterator interface, only Traversable. It > just happens to have functions of the same name as Iterator.
Ah, okay. That leaves the picky OO strictness I personally don't like. I think we should give PHP some freedom back in this area. Regards,
-- Michael

Sönke Ruempler

20 years ago
Michael Wallner <mailto:mike@php.net> wrote on Friday, June 23, 2006 11:27 PM:
> Andrei Zmievski wrote: >> TextIterator does not implement Iterator interface, only >> Traversable. It just happens to have functions of the same name as >> Iterator. > > Ah, okay. > That leaves the picky OO strictness I personally don't like. > I think we should give PHP some freedom back in this area.
+1 :) -soenke

Andrei Zmievski

20 years ago
I think I'm done with TextIterator for now. Feel free to play with it. :) -Andrei On Jun 25, 2006, at 11:15 PM, Soenke Ruempler wrote:

Marcus Börger

20 years ago
Hello Andrei, you should reconsider giving it full iterator interface. Adding more functionality is of course quite easy and no problem at all. Sharing an interface also makes the whole thing feel much better as in no conflicts as in consistent. Actually that's what interfaces are for. Friday, June 23, 2006, 11:20:28 PM, you wrote:
> TextIterator does not implement Iterator interface, only Traversable. > It just happens to have functions of the same name as Iterator.
> -Andrei
> On Jun 23, 2006, at 2:16 PM, Michael Wallner wrote:
>> Andrei Zmievski wrote: >>> I am working on implementing BreakIterator API [1]. I considered two >>> approaches: making a separate class or merging the API into the >>> existing TextIterator. Having a separate class would be a bit >>> cleaner, but I can see people wanting to use it in foreach(), and >>> since TextIterator already provides a lot of BreakIterator's >>> functionality, I decided that merging would be the best option. >>> However, there is an overlap between the BreakIterator API and the >>> current TextIterator one, so there will have to be some changes. >>> 1. TextIterator::current() signature will change from: >>> mixed current() >>> to: >>> mixed current(integer &$offset) >> >> This will raise the same issue as we have/had with the reflection API: >> mike@honeybadger:~/build/php-unicode-debug$ cli -r 'interface >> i{function f();} class c implements i{function f($a){}}' >> >> Fatal error: Declaration of c::f() must be compatible with that of >> i::f() in Command line code on line 1 >> >> >> Nah, don't look at me--I don't like that either. >> >> >>> 3. TextIterator::rewind() will be renamed to TextIterator::first() to >>> conform to BreakIterator's first()/last() API. >>> So this is heads up. Let me know if you have a problem with this. >> >> Huh? Rename or alias? It can't implement Iterator if there's no >> rewind() method.
Best regards, Marcus

Andrei Zmievski

20 years ago
Thanks, I know what they are for. I remember discussing Iterator vs. Traversable with you on IRC, and for some reason we settled on Traversable. Anyway, I'm changing it now. -Andrei On Jun 24, 2006, at 2:35 AM, Marcus Boerger wrote:

Andrei Zmievski

20 years ago
Sean (on IRC) convinced me that something called *Iterator had better implement Iterator interface (which TextIterator currently does not). So changing method signatures is out of the question. Towards that, the current functions will stay as they are, but I'll have to add current_offset() (for getting the offset of the current element as opposed to the element itself), and alias rewind() to first(). -Andrei On Jun 23, 2006, at 2:00 PM, Andrei Zmievski wrote: