Optimize NSRange conversion
#120
Conversation
`UTF8Index.distanceTo(_:)` cost depends on actual distance. So, reduce cost by calculating length from start`s index instead of calculating distances of stringStart and stringEnd separately. By applying this, duration of linting [KeychainAccess](https://github.com/kishikawakatsumi/KeychainAccess/) by SwiftLint is reduced from 27sec to 21sec.
The cost of `UTF8Index.advancedBy(_:)` depends on actual distance. So, reduce cost by calculating endIndex based on startIndex instead of calculating each of stringStart and stringEnd separately.
| let startUTF8Index = string.utf8.startIndex.advancedBy(start) | ||
| let endUTF8Index = startUTF8Index.advancedBy(length) | ||
|
|
||
| guard let startUTF16Index = startUTF8Index.samePositionIn(string.utf16), |
There was a problem hiding this comment.
are utf16 views cached in the standard library? If not, we should cache it here (e.g. just calculate it once).
There was a problem hiding this comment.
At first since I filed #119, I tested caching full results of offsets. But that was not efficient.
By realm/SwiftLint#263, times of calling byteRangeToNSRange() is omitted on bottle neck case, so I did not include that.
There was a problem hiding this comment.
that's not what I mean. I mean are utf16 views on Strings cached by the Swift standard library?
e.g. is it more efficient to do this:
let utf16View = string.utf16
guard let startUTF16Index = startUTF8Index.samePositionIn(utf16View),
let endUTF16Index = endUTF8Index.samePositionIn(utf16View) else {
return nil
}or is that equivalent to this (current code):
guard let startUTF16Index = startUTF8Index.samePositionIn(string.utf16),
let endUTF16Index = endUTF8Index.samePositionIn(string.utf16) else {
return nil
}There was a problem hiding this comment.
Oh, I understand.
There was a problem hiding this comment.
I tested with changes of that and got same performance.
It seems Swift optimizer is clever enough. 😃
But apply them.
|
Thanks for doing this! |
|
Thank you for reviews! |
I'm not sure if the gains there will be worth it, but if they're 10% or more, it'd be great. |
By applying this, duration of linting KeychainAccess by SwiftLint is reduced from 27sec to 21sec.
Related: #119