Honestly, I'm skeptical that we want to develop a grammar for full algebraic type definitions in docs when the language doesn't support it. That risks encouraging some very bad practices, and if PHP itself ever adds native algebraic types it could get weird if there's any inconsistency (which there almost certainly will be).
That said, if we must support it in the doc format then follow the same precedence order as PHP itself: & binds higher than |. Having a different set of parse rules than PHP itself is the way to madness. :-) --Larry Garfield On Thursday, November 8, 2018 12:52:18 PM CST Chuck Burgess wrote: > Previous replies indicate that whitespace around operators is perfectly > acceptable, so that seems resolved. > > What about the issue of operator precedence when "|" and "&" are both > needed? > > Do we want to say one is higher order than the other, resulting in the > possibility of one Type being listed multiple times: @return A & C | B > & C > > Or should they be equal precedence, needing parentheses to enforce order: > @return (A | B) & C > > CRB > > On Tuesday, October 23, 2018 at 8:09:19 AM UTC-5, Chuck Burgess wrote: > > Having both operators at different levels would mean that combinations > > such as `@param (A|B)&C $test` would have to be written as `@param A&C|B&C > > $test`. > > > > I'm not against allowing whitespace around the operators, if the > > implementors agree it's still easy enough to parse correctly. Since the > > variable name should come between the Types and the Description, perhaps > > that's not a big deal. > > > > CRB > > *about.me/ashnazg <http://about.me/ashnazg>* > > > > On Sun, Oct 21, 2018 at 9:40 AM <[email protected]> wrote: > >> Regarding the ABNF grammar, there are few things that need to be decided. > >> > >> (1) Priority / interaction with union and array "operators". I would > >> strongly recommend disallowing union and intersection on the same "level" > >> and always require brackets to explicitly declare the intention. For > >> consistency with union types, array and intersection should be allowed on > >> the same level with array having higher priority. This matches behavior > >> used by PHPStan and can be achieved for example with the following > >> grammar > >> > >> type = atomic [union / intersection] > >> union = 1*("|" atomic) > >> intersection = 1*("&" atomic) > >> atomic = identifier [array] / "(" type ")" [array] > >> array = 1*("[]") > >> identifier = keyword / class-name > >> keyword = "array" / "bool" / ... > >> class-name = ["\"] label *("\" label) > >> label = label-head *label-tail > >> label-head = ALPHA / "_" / %x80-FF > >> label-tail = ALPHA / DIGIT / "_" / %x80-FF > >> > >> > >> (2) Allowing horizontal whitespaces around operators. With more complex > >> types it makes sense to allow horizontal whitespaces around operators. It > >> complicates the grammar a bit, but it makes complex types a lot readable. > >> It may be better the post allowing horizontal whitespaces as a standalone > >> PR independent of intersection types. > >> > >> Regards, > >> Jan Tvrdík > >> > >> > >> > >> ---------- Původní e-mail ---------- > >> > >> Od: Chuck Burgess <[email protected]> > >> > >> Komu: [email protected] > >> > >> Datum: 21. 10. 2018 3:22:06 > >> > >> Předmět: Re: [PSR-5] Intersection Types > >> > >> Yes, parentheses would be required for controlling order of precedence. > >> > >> In your example, I would expect: > >> > >> > >> > >> > >> * @param (CacheInterface&ResetableInterface)|ResetableCacheInterface > >> $cache > >> > >> > >> CRB > >> > >> > >> > >> On Sat, Oct 20, 2018, 17:43 AzJezz <[email protected]> wrote: > >> > >> Hi, > >> > >> i find it quite confusing myself, here's a use case mixing `&` and `|`. > >> > >> > >> ```php > >> > >> <?php > >> > >> namespace App\Foo; > >> > >> class Bar { > >> > >> /** > >> * in this example `Bar` constructor accepts an object that implement > >> * ( `CacheInterface` and `ResetableInterface` ) or > >> > >> `ResetableCacheInterface` > >> > >> * > >> * should the doc block be formatted this way : > >> * > >> * @param > >> > >> object&CacheInterface&ResetableInterface|ResetableCacheInterface $cache > >> > >> * > >> * or maybe like this ? > >> * > >> * @param > >> > >> object&ResetableCacheInterface|ResetableInterface&CacheInterface $cache > >> > >> * > >> * or maybe we can use parentheses ? ¯\_(ツ)_/¯ > >> * > >> * @param object & ( ResetableCacheInterface | ( ResetableInterface & > >> > >> CacheInterface ) ) $cache > >> > >> */ > >> public function __construct($cache) { > >> > >> // initialize cache property > >> > >> } > >> > >> } > >> ``` > >> > >> > >> > >> > >> On Sat, Oct 20, 2018 at 11:53 AM Johannes Schmitt <[email protected]> > >> wrote: > >> > >> Hi there, > >> > >> generally, I think the addition of `&` is a great idea. > >> > >> > >> One thing regarding the grammar specifically, right now you would support > >> mixing `|` with `&` like `A|B&C`. I'm not sure if mixing would be > >> desirable > >> (I don't have use-case for this at this point). Also the grammar > >> currently > >> requires `|` to be before `&`- so something like `A&B|C` would not be > >> supported - I'm not sure if this is intentional? Maybe it's best to only > >> either support `|` or `&`, but not allow to mix them for the moment? > >> > >> > >> Thanks, > >> Johannes > >> > >> > >> On Fri, Oct 19, 2018 at 2:50 PM Chuck Burgess <[email protected]> > >> wrote: > >> > >> Hello everyone, > >> > >> There is a new PR [1] from a contributor, asking for an Intersection Type > >> Operator. This appears to simply use `&` akin to how `|` is used for > >> Union > >> Types. > >> > >> > >> Neither Unions nor Intersections are (yet) in the language itself, but > >> `string|null` Union Typing in Tags has been in wide usage for a while > >> now. > >> In looking over RFCs on attempts to get these two Type Operators into the > >> language, it seems likely to me that the Operators chosen will be `|` and > >> `&` if they do ever get in. As such, I'm personally good with the choice > >> of `&` for Intersection Operator for Typing in Tags. > >> > >> > >> Please keep discussion on this request on this ML thread. > >> > >> > >> Chuck Burgess, Editor > >> > >> > >> [1] -- https://github.com/php-fig/fig-standards/pull/1104 -- You received this message because you are subscribed to the Google Groups "PHP Framework Interoperability Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/php-fig/12147766.9nEoe5WhWs%40vulcan. For more options, visit https://groups.google.com/d/optout.
signature.asc
Description: This is a digitally signed message part.
