Improve perf of fromCharCode(), fixes #71 #72
Conversation
|
@nolanlawson any reason you can't apply this optimization to |
|
OK, done |
| function decodeCodePointsArray (buf) { | ||
| var len = buf.length | ||
|
|
||
| if (len <= 0x10000) { |
There was a problem hiding this comment.
I might pull this constant out and put the explanation comment on it, that way its not magically here.
There was a problem hiding this comment.
thanks, that's a good idea!
There was a problem hiding this comment.
The issue isn't manual string concatenation; it's doing an unnecessary Array.slice() when the buffer.length is < 0x10000. I think it's pretty clear that doing slice() is slower than not doing slice(), although a JSPerf would tell us exactly how slow.
There was a problem hiding this comment.
I can indeed verify that slice is slower than no-slice: http://jsperf.com/string-fromcharcode-apply-vs-string-fromcharcode-using-/4
Improve perf of fromCharCode(), fixes #71
|
Thanks @nolanlawson |
|
No prob! Any chance of a 3.4.3 release soon? Would unbreak some of our tests in PouchDB. |
| @@ -723,13 +723,17 @@ function asciiSlice (buf, start, end) { | |||
| } | |||
|
|
|||
| function binarySlice (buf, start, end) { | |||
| var ret = '' | |||
| end = Math.min(buf.length, end) | |||
There was a problem hiding this comment.
Will re-add this, shouldn't have been removed.
Probably need to add a test for it.
There was a problem hiding this comment.
ah, my bad. sorry!
Cool, that test of ours I mentioned in #70 (the "big png" test) went from 2.7s to 1.4s. Thanks @jessetane for pointing this out!