On Apr 2, 2011, at 15:24 , Jordi Boggiano wrote: > On Sat, Apr 2, 2011 at 1:38 PM, Jacob Oettinger <ja...@oettinger.dk> wrote: >> On Mar 31, 2011, at 21:10 , Rasmus Lerdorf wrote: >> >>> On 03/31/2011 10:58 AM, Martin Scotta wrote: >>>> I think it's time to stop thinking in terms of "functions" and move >>>> forward to "abstractions" >>>> >>>> $s1 = 'string'; >>>> $s1->contains($s2); >>>> >>>> $s1->indexOf($s2) === strpos($s1, $s2); >>>> >>>> Why can't the strings be exposed as pseudo-objects ? users can choose to >>>> use them as a regular strings or by calling methods on it. >>> >>> This is actually something I have been toying with a bit. Adding the >>> ability to call methods on both strings and arrays. I still don't like >>> the idea of making them real objects as the overhead and the amount of >>> code that would need to be changed in the core and in every extension is >>> daunting. Anybody out there have a patch along these lines? >> >> >> Sounds interesting. A few thoughts: >> >> The new "methods" could be implemented as functions that accept the string >> or array as their first argument. Thereby allowing them to be called as >> functions too. >> >> If the new methods are functions. Maybe they should be defined in the >> \string and \array namespaces. This would allow a fresh start in naming >> string and array functions, and allow addressing argument ordering issues. >> It would of curse also open endless discussions on which functions to >> include and the naming of these. >> >> If the new string and array functions were defined in some namespaces. Would >> it be possible to allow extending the string and array "classes" by defining >> more functions in those namespaces in userland and/or extensions? >> >> Should the new string functions be multibyte character set aware? >> >> Should the above be generalized so that the -> operator can be used on any >> simple type, and if the called "method" exists as a function in the types >> namespace, it should be called. This would break the task into two parts. !) >> changing php to attempt calling functions in a certain namespace for each >> simple type. 2) Writing extensions that create functions for each type, >> string and/or array would be obvious starting points. > > This would be great - especially if it would provide multibyte (or > only UTF-8 for all I care) support. Of course one potential pain point > is numbers vs strings.
As Stan Vass points out the methods should probably be implemented for scalars and for arrays, and handling of types should be done in the same way as the existing string and number functions. So maybe only \scalar and \array. Alternatively scalars should look for methods in multiple namespaces to keep the function naming consistent. > > Looking at JavaScript which acts quite similarly to the proposed > approach, the Number methods [1] are not many and not often useful, > but still they exist. What happens if you do "3.3"->ceil() ? If ceil() > exists only in \numeric, you could say autoconvert, but if it exists > in both \string and \numeric we have a problem, so the only sensible > way is to drop type juggling for those autoboxed objects imo. Which is > also what JS does btw. If you have an ambiguous value on your hands, > you should cast it (x.toString() or parseInt(x)) first and then call > what you want to call on it. Another interesting to note is that JS > doesn't allow method calls directly on numbers - 3.toPrecision(2) for > example is a syntax error, you have to put it in a var first. I guess > it's not a problem since you should just inline the value you want > anyway, the Number methods don't do super fancy things. Yes as I wrote the above I did realize that it has some resemblance to javascripts prototypal inheritance, so it is a good idea to look there for inspiration. But the above is not supposed to be a suggestion for introducing a new object model into PHP. -- PHP Internals - PHP Runtime Development Mailing List To unsubscribe, visit: http://www.php.net/unsub.php
