Channels
#stdlib
- v
voddan
09/28/2016, 6:56 AMfun boo(x: Any): List = x.cast() - v
voddan
09/28/2016, 6:57 AMthis is even more implicit - s
stepango
09/28/2016, 7:14 AMI thought it should be obvious that
happened insidetype cast
function, but i got your point, in some cases implementation i using have no guaranty you could actually make this castcast() - v
vmironov
09/28/2016, 7:41 AM+1 for adding
to the stdlib. It's one of the most popular extension functions in our project.cast - m
michaelrocks
09/28/2016, 8:09 AMDespite working on the same project with @vmironov I’m -1 for adding
as it’s an extremely dangerous function 🙂cast - k
kirillrakhman
09/28/2016, 9:08 AMI like that the extension method allows for a fluent style which works well with the safe-call operator. e.g.list.firstOrNull()?.cast<Foo>().bar - k
kirillrakhman
09/28/2016, 9:08 AMyou can also put it in method references:list.map(Any::cast) - k
kirillrakhman
09/28/2016, 9:09 AMbut I get why it's dangerous, e.g. it allows passing generic types and won't emit a warning for an unchecked cast - k
kirillrakhman
09/28/2016, 9:10 AMI was thinking about whether
could be made an intrinsic similar tocast()
andarrayOfNulls()
would be desugared to itas - k
kirillrakhman
09/28/2016, 9:11 AMthis would make
andfoo as Bar
equalfoo.cast<Bar>() - k
kirillrakhman
09/28/2016, 9:11 AMsame as with
and1 + 21.plus(2) - k
kirillrakhman
09/28/2016, 9:11 AMIMO it would fit nicely into the Kotlin philosophy😕 1👍 2 - s
stepango
09/28/2016, 9:27 AM@kirillrakhman or even
wich would generate warnings<http://foo.as|foo.as>(Bar::class) - k
kevinmost
09/28/2016, 2:57 PMI think I'd prefer it if it acted like
and notas?
. I'd rather have to use the safe-call operator than potentially hide an exception.asCopy code
Usage:inline fun <reified R : Any> Any?.cast(): R? = this as? RCopy codelist.firstOrNull()?.cast<Foo>()?.bar - u
uhe
09/28/2016, 2:59 PMthat should be calledcastOrNull - u
uhe
09/28/2016, 3:00 PMbecause
is also unsafe in that regard, which is why we havefirst()firstOrNull() - k
kevinmost
09/28/2016, 3:02 PMI guess so. I find myself rarely using
because it's unsafe, but I guess it's more consistent to name them that wayfirst() - c
cedric
09/28/2016, 4:27 PMWhile I like the
syntax, I’d vote against it as well because it makes it harder to locate such casts in a general search in the code base. If I want to find out how many times I cast in my code, I can just do a search forcast
. Adding synonyms to this makes it harder to conduct this kind of search.” as “ - c
cedric
09/28/2016, 4:28 PM(flow typing is already making such a search somewhat difficult) - v
voddan
09/28/2016, 4:30 PMFlow typing eliminates only save casts. Normally you want to find / refactor the others - c
cedric
09/28/2016, 4:30 PMYes, I’m talking about searching for all casts, safe and unsafe - m
michaelrocks
09/28/2016, 4:31 PMActually it’s much easier to find all usages of a function, at least when using an IDE - c
cedric
09/28/2016, 4:32 PMBut now you have two ways to cast, so you have to do two searches. - c
cedric
09/28/2016, 4:33 PMIt the language was still pre-1.0, I’d recommend
over.cast()
(one less keyword), but not today.as - m
michaelrocks
09/28/2016, 4:34 PMAgree, but the later you change something, the harder it becomes - m
michaelrocks
09/28/2016, 4:35 PMSo changing it in 1.0 is better than in 1.1 or 2.0 - k
kirillrakhman
09/28/2016, 4:35 PM@cedric if
were sugar for a method, you could just use "find usages"as - c
cedric
09/28/2016, 4:35 PMbut maybe it’s even better not to change it at all... - c
cedric
09/28/2016, 4:36 PM@kirillrakhman Yes, is it possible to look for
as a keyword with IDEA instead of a text search?as - v
voddan
09/28/2016, 4:37 PMstructural search