refactor: optimize memoize use, misc cases #10122
Conversation
SimonSchick
force-pushed the
SimonSchick:optimize/memoize
branch
3 times, most recently
from
a055458
to
057a566
SimonSchick
force-pushed the
SimonSchick:optimize/memoize
branch
from
057a566
to
f93d65b
Codecov Report
@@ Coverage Diff @@
## master #10122 +/- ##
==========================================
- Coverage 96.36% 96.34% -0.02%
==========================================
Files 63 63
Lines 9441 9419 -22
==========================================
- Hits 9098 9075 -23
- Misses 343 344 +1
Continue to review full report at Codecov.
|
|
LGTM, just one change |
| * @private | ||
| */ | ||
| function addSetToArray(array, set) { | ||
| for (const v of set) { |
sushantdhiman
Nov 12, 2018
Contributor
return array.concat(Array.from(set))
return array.concat(Array.from(set))
SimonSchick
Nov 12, 2018
Author
Member
'use strict';
function addSetToArray(array, set) {
for (const v of set) {
array.push(v);
}
}
const set = new Set([1, 2, 3, 4, 5, 6, 7, 8, 9]);
console.time('addSetToArray');
for(let i = 0;i<0xffffff;++i) {
addSetToArray([], set);
}
console.timeEnd('addSetToArray');
console.time('Array.from');
for(let i = 0;i<0xffffff;++i) {
[].concat(Array.from(set));
}
console.timeEnd('Array.from');
addSetToArray: 736.759ms
Array.from: 20121.014ms
On my machine using .concat(Array.from(set)) is about 25x slower.
The intention here is to make the code faster using sets, converting sets to arrays is a rather expensive operation, addSetToArray is the fastest way to do it pretty much, using Array.from would nullify all performance gained 😛
'use strict';
function addSetToArray(array, set) {
for (const v of set) {
array.push(v);
}
}
const set = new Set([1, 2, 3, 4, 5, 6, 7, 8, 9]);
console.time('addSetToArray');
for(let i = 0;i<0xffffff;++i) {
addSetToArray([], set);
}
console.timeEnd('addSetToArray');
console.time('Array.from');
for(let i = 0;i<0xffffff;++i) {
[].concat(Array.from(set));
}
console.timeEnd('Array.from');
addSetToArray: 736.759ms
Array.from: 20121.014ms
On my machine using .concat(Array.from(set)) is about 25x slower.
The intention here is to make the code faster using sets, converting sets to arrays is a rather expensive operation, addSetToArray is the fastest way to do it pretty much, using Array.from would nullify all performance gained
sushantdhiman
Nov 12, 2018
Contributor
I personally verified and Array.from is indeed really slow. Bummer, it was really concise. Will go with for + push approach
I personally verified and Array.from is indeed really slow. Bummer, it was really concise. Will go with for + push approach
|
I guess I enabled you to edit and there were conflicts, I would've rebased it for you :) |
|
@sushantdhiman mergy merge? |
| if (this.constructor._hasVirtualAttributes) { | ||
| keys = keys.concat(this.constructor._virtualAttributes); | ||
| Utils.addSetToArray(keys, this.constructor._virtualAttributes); |
sushantdhiman
Nov 15, 2018
Contributor
I didn't merged because this still doesn't looks right to me, I think we should keep using array for _virtualAttributes
I didn't merged because this still doesn't looks right to me, I think we should keep using array for _virtualAttributes
SimonSchick
Nov 16, 2018
Author
Member
Any reason why?
The overall code will be faster and it's not any less readable :P
Any reason why?
The overall code will be faster and it's not any less readable :P
sushantdhiman
Nov 17, 2018
Contributor
We are explicitly converting set to an array, this points out what we really need here is an array. Also simple concat on array is more readable and less complex
We are explicitly converting set to an array, this points out what we really need here is an array. Also simple concat on array is more readable and less complex
SimonSchick
Nov 17, 2018
Author
Member
What about the 5 other .has (previously .includes) calls? 😛
What about the 5 other .has (previously .includes) calls?
sushantdhiman
Nov 17, 2018
Contributor
has and includes got strong interchangeability , on other had for _virtualAttributes case we had to write special code to convert set to array, which indicate need of array rather than set
has and includes got strong interchangeability , on other had for _virtualAttributes case we had to write special code to convert set to array, which indicate need of array rather than set
SimonSchick
Nov 17, 2018
Author
Member
The point of this PR was to primarily to get rid of memoize + includes and replace with sets since that's slows all reads. Also keep in mind the set is only converted for iteration.
I could change the code in this location to not create arrays at all.
The point of this PR was to primarily to get rid of memoize + includes and replace with sets since that's slows all reads. Also keep in mind the set is only converted for iteration.
I could change the code in this location to not create arrays at all.
sushantdhiman
Nov 17, 2018
Contributor
I could change the code in this location to not create arrays at all.
Hmm, give it a shot
I could change the code in this location to not create arrays at all.
Hmm, give it a shot
SimonSchick
Nov 17, 2018
Author
Member
Done, waiting for build, I'm in the process of moving atm so I might take a while to reply.
Done, waiting for build, I'm in the process of moving atm so I might take a while to reply.
SimonSchick
force-pushed the
SimonSchick:optimize/memoize
branch
from
b9bc8bf
to
ac3ae90
|
Thanks @SimonSchick |
Pull Request check-list
Please make sure to review and check all of these items:
npm run testornpm run test-DIALECTpass with this change (including linting)?Description of change
Followup of #10091, re-implements the previous changes that were dropped for being too out of scope.
This PR also optimizes/streamlines type checks eg.
_.isTypetotypeof ... === 'type', lodash is generally slower for those as it uses alternative methods for type checking that are compatible with legacy browsers and the likes.