Channels
#stdlib
- y
yole
08/08/2016, 3:11 PMthis won’t update the property - d
dimsuz
08/08/2016, 3:12 PMoh, but still nice to know, didn't know aboutorEmpty() - d
dimsuz
08/08/2016, 3:13 PMso I can use
🙂map = map.orEmpty() - d
dimsuz
08/08/2016, 3:14 PMah, no, orEmpty() returns not mutable map - d
dimsuz
08/08/2016, 3:14 PMbut I can do similar to how
is implemented:orEmpty()map = map ?: mutableMapOf() - d
dimsuz
08/08/2016, 3:14 PMlooks nicer than if - d
dmitry.petrov
08/08/2016, 3:42 PMYou can implement lazy property manually. This will have an advantage of not creating a separate object for a delegate. Still, you'll have to take care about synchronization and so on. - d
dimsuz
08/08/2016, 4:00 PMin this case I don't care about synchronization. It just so happens that this code path where I need this map is quite rare and this code is for library. So I'd better use null checks than instantiate a lazy property because otherwise this object will be created in 90% cases completely unnecessary - v
voddan
08/08/2016, 5:28 PMwont JIT eliminate the object if it is not used? - d
damian
08/08/2016, 5:43 PMif you have JIT on your platform, and it's good 🙂 - k
kirillrakhman
08/09/2016, 7:51 AMis there a way to build a map from a list of keys and a value selector function? I can only findkeys.associate { it to selector(it) } - k
kirillrakhman
08/09/2016, 7:54 AMI'd like to have
(optimally without creating intermediatekeys.toMap(selector)
instancesPair - v
voddan
08/09/2016, 9:58 AMfun <T, K, V> Iterable<T>.associateBy(keySelector: (T) -> K, valueTransform: (T) -> V): Map<K, V> - v
voddan
08/09/2016, 9:58 AMit has no overhead - a
Andreas Sinz
08/11/2016, 1:32 PMwhy does
returnSet.filter
?List - k
kirillrakhman
08/11/2016, 1:57 PMThere is no
, onlySet.filterIterable.filter - k
kirillrakhman
08/11/2016, 1:57 PMyou could probably writeset.filterTo(HashSet()) { ... }👍 1 - a
Andreas Sinz
08/11/2016, 2:06 PMI've created my own
^^ let me rephrase my question: why is there noSet.filter(): Set
?Set.filter(): Set
seems like wasted cpu timeSet.filter { }.toSet() - m
mikehearn
08/11/2016, 2:20 PMis
really so commonly needed that it deserves to be in the standard library?buildString - m
mikehearn
08/11/2016, 2:21 PMit saves only a small number of keystrokes, but if you came across it you'd have to know what it did or look it up - m
mikehearn
08/11/2016, 2:21 PMit seems like a small saving especially as once java 9 comes out, kotlinc can use the same invokedynamic approach javac will to make string concatenation much faster - k
kirillrakhman
08/11/2016, 2:35 PMI've used it a couple of times, I like it - m
mikehearn
08/11/2016, 2:59 PMthe java 9 support can do quite advanced pre-sizing of stringbuilders - m
mikehearn
08/11/2016, 3:00 PMthat's partly how it can go so fast. so i worry that this part of the stdlib, if launched in 1.1, would have quite a short shelf life. at least for people who can keep up with the jvm. i suppose it will as always be useful on android where the VM is quite weak - i
ilya.gorbunov
08/11/2016, 3:00 PM@mikehearn: I suppose invokedynamic approach in java 9 is for better string concatenation from a fixed number of parts, it doesn't provide a replacement for the code operating on string builders. - m
mikehearn
08/11/2016, 3:04 PM - m
mikehearn
08/11/2016, 3:04 PMit is more sophisticated than that - m
mikehearn
08/11/2016, 3:06 PMkotlin should use this new bytecode form for its own string interpolation needs anyway - m
mikehearn
08/11/2016, 3:06 PMsometimes the new code can correctly guess the right stringbuilder size and then there is a big win - m
mikehearn
08/11/2016, 3:07 PMadditionally, the use of indy means it could potentially inline the expected size into the call-site, thus optimising future runs. not sure it currently does though