On Tue, Jul 2, 2013 at 8:21 AM, Simon Schick <simonsimc...@gmail.com> wrote:
> On Fri, Jun 28, 2013 at 11:20 PM, Derick Rethans <der...@php.net> wrote: > > > > Hey Simon, > > > > PS, please don't top-reply as per mailinglist guidelines: > > http://us2.php.net/reST/README.MAILINGLIST_RULES > > > > On Fri, 28 Jun 2013, Simon Schick wrote: > > > On Fri, Jun 28, 2013 at 7:10 PM, Derick Rethans <der...@php.net> > wrote: > > > > > > > On Fri, 28 Jun 2013, Simon Schick wrote: > > > > > > > > > I'd like to extend the API of php by a method. > > > > > > > > Great, let me know if you need any help with the implementation, or > > > > when you have any questions on how it currently works. > > > > > > Sorry - I that was easy to missunderstand ;) I thought of providing > > > the idea. But I would happily also try to implement it by my own - > > > just to also have contributed to the php-core. If you could give me a > > > hint where to start - I feel quite lost in the code right now ... even > > > so I read the great tutorials about php-internals provided by Anthony > > > Ferrara. > > > > > > And this may take some time! I've just started with C/C++. > > > > > > Do you think this is an easy task to start? Depending on knowledge of > > > php-internals ;) > > > > It's not that difficult, but not the easiest thing either. Basically, > > you have to do the following steps: > > > > 1. in ext/date/lib/interval.c add a function that has two arguments of > > the timelib_rel_time* type. It should return a *new* > > timelib_rel_time* structure that has the two intervals added up > > together. It shouldn't change either of the original > > timelib_rel_time* arguments. Add the function definition to > > ext/date/lib/timelib.h too. > > 2. You can only add two timelib_rel_time* arguments if neither of > > first_last_day_of, special, have_weekday_relative, and > > have_special_relative are set. You can'd add a "first day of next > > month" to a "+5 days" interval - it only works for simple y:m:d h:i:s > > intervals. > > 3. In ext/date/php_date.c add: > > > > a. a new ARG_INFO struct after the one for > > arginfo_date_interval_construct, > > b. Add a new method, "add" after > > PHP_ME_MAPPING(createFromDateString, > date_interval_create_from_date_string,... > > c. Add a new PHP_FUNCTION(date_interval_add) after the function > > PHP_FUNCTION(date_interval_format) that takes two DateInterval > > objects, extract the timelib_rel_time information, call the new > > function that you've added in step 1, and replace the > > timelib_rel_time* that is part of the DateInterval object with the > > returned value. > > d. Add a forwards declaration to php_date.h after > > PHP_FUNCTION(date_interval_create_from_date_string); > > > > 4. Make sure that the function can be called both as a procedural way > > (date_interval_add) and an object oriented way (DateInterval->add()). > > > > 5. Write test cases and put them in ext/date/tests. > > > > That's what I can think off right now. > > > > cheers, > > Derick > > > > > > > > -- > > http://derickrethans.nl | http://xdebug.org > > Like Xdebug? Consider a donation: http://xdebug.org/donate.php > > twitter: @derickr and @xdebug > > Posted with an email client that doesn't mangle email: alpine > > Hi, Derick > > I thought a bit about the changes here and came up with the following > feedback: > > If I can't add intervals where either one is relative, I'd also add a > method isRelative() to the class, where the user can check if it's a > relative interval. > But anyways - what happens if the interval is relative and you try to > access one of the public properties? They're all designed for static > intervals (let's say +5days), right? > > What, if the user tries to add a relative to a fixed interval? Should > I trigger a E_WARNING, because it's no breaking error, but the value > can't be calculated? > > > Bye, > Simon > > -- > PHP Internals - PHP Runtime Development Mailing List > To unsubscribe, visit: http://www.php.net/unsub.php > > Hi, here is the patch: https://github.com/php/php-src/pull/390 It's basically what Derick suggested. However there is one issue that needs to be resolved - interval normalization (timelib_do_rel_normalize). The problem is that currently there is no way how to get a base timelib_time from the interval. It means that it's not possible to find out how many days use for normalization. For example if interval1 is 20 days and interval2 is 21 days, then the resulted normalized interval will be 1 month and ?? days... :) There are few solution that come to my mind: 1. no normalization (currently implemented) - this is the easiest solution but the result from the example will be 0 months and 41 days can give interval that has over 31 days... 2. warning if the resulted interval has more than 28 days (February :) ) 3. save base timelib_time with the interval when created and then use it as a base class - it will probably require some sort of copying or ref counting but it seems like the most logical solution What do you think? Jakub
