Paul, I never used and didi not even known the existence of ERXGenericRecord.valueForKey(). The problem with these powerful ERXKey is that some operator will reduce a collection and some will create a collection. There is no way to infer these results with the current way they are done (using generics). I see no way to make the ERXGenericRecord.valueForKey() type to work all the time.
For simple key this may works but with key path, things get very complex. For exemple traversing 2 toMany relationships will return an array of array of the final type that can be reduced to a single level array using atFlatten() and this one to a single object with atObjectAtIndex(index) or atMax(), ... A new method could be added to return a new no-op key with the NSArray version of the type to use with ERXGenericRecord.valueForKey() if required. As I do not know where ERXGenericRecord.valueForKey is used, I cannot comment on this. This complexity may explain the proposition to have a native equivalent in Swift, they probably want make sure it will always works by having a better knowledge of the effect of operations chain. Samuel > Le 29 mars 2017 à 19:40, Paul Hoadley <pa...@logicsquad.net> a écrit : > > On 29 Mar 2017, at 09:19, Paul Hoadley <pa...@logicsquad.net> wrote: > >> Anyone else got any thoughts on all this? > > > I see it was discussed on the old Wonder list back in 2009. Dave Avendasora > asked the exact question I asked: > >> Since this is a to-many relationship shouldn't it be >> "ERXKey<NSArray<ScheduledRoutingRelationship>>"? > > > And Mike Schrag answered: > >> no it just doesn't really work this way ... if it did this, it's easy to >> extend it to add the array, but it's impossible to deconstruct it to get the >> inner type. there are methods to extend keys, etc with array variants for >> most operations, but you rarely actually want the type of the key to be >> NSArray<T> > > > Along with Samuel’s observations, seems like case closed. It does still seem > to leave ERXGenericRecord.valueForKey(ERXKey<T> key) broken in the case of > to-many relationship ERXKeys, though, doesn’t it? (Worse, it will break at > runtime.) > > > -- > Paul Hoadley > http://logicsquad.net/ > > > > _______________________________________________ > Do not post admin requests to the list. They will be ignored. > Webobjects-dev mailing list (Webobjects-dev@lists.apple.com) > Help/Unsubscribe/Update your Subscription: > https://lists.apple.com/mailman/options/webobjects-dev/samuel%40samkar.com > > This email sent to sam...@samkar.com _______________________________________________ Do not post admin requests to the list. They will be ignored. Webobjects-dev mailing list (Webobjects-dev@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
