Reduce build time with minor reference refactor

[xinranw]

Adding a self. reference to a instance variable reduced the compile
time significantly (over 10x improvement). A single line evaluation of a
variable was further broken out into separate steps to reduce the
compile time as well.

This cuts at least 2 seconds from the build time

Compile times were analyzed with the following bash command:
xcodebuild -scheme Charts -sdk iphonesimulator PLATFORM_NAME=iphonesimulator clean build OTHER_SWIFT_FLAGS="-Xfrontend -debug-time-function-bodies" 2>&1 | grep -e [0-9][0-9].[0-9].ms | sort -nr

Compile times before the change:

Compile times after the change:

[codecov-io]

Codecov Report

Merging #2679 into master will decrease coverage by <.01%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master    #2679      +/-   ##
==========================================
- Coverage   19.65%   19.65%   -0.01%     
==========================================
  Files         112      112              
  Lines       13717    13715       -2     
==========================================
- Hits         2696     2695       -1     
+ Misses      11021    11020       -1

Impacted Files	Coverage Δ
.../Charts/Renderers/HorizontalBarChartRenderer.swift	0% <0%> (ø)	⬆️
...ource/Charts/Renderers/ChartDataRendererBase.swift	47.05% <0%> (-2.95%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e72f280...80a45c0. Read the comment docs.

[liuxuan30]

Hi there, thanks for the input. Do you have any doc or post could explain this change?

[xinranw]

The part at the bottom of this article references a similar problem, where chaining operations can sometimes cause dramatic increases in compile time (also referenced by an Apple build engineer at WWDC).

For the self. addition that actually accounted for the majority of the build time improvement...I honestly don't have a good answer. Based on articles such as this and this, it seems a lot of the optimizations that others have found can be somewhat arbitrary. I ended up just trying various suggestions and noticed that adding the self. made a 1000x improvement. I wonder if this change is somehow helping the compiler figure out the reference to viewPortHandler (perhaps since it's a inherited property from the Renderer class.

[xinranw]

Bump

[liuxuan30]

Thank you for your information. I'm fine with this PR, but also needs other peoples' review.

BTW, if it helps, do we need to change the entire code base?

[xinranw]

There are certainly other parts of the code that can be optimized but these two lines were taking the longest amount of time (over 2 seconds if being used as a cocoapod). So I only tackled these two as the easiest way to cut down a somewhat significant part of the build time.

[liuxuan30]

I got self is improving the speed. But can we just uses self, without using two instants? like

return data.entryCount < Int(CGFloat(dataProvider?.maxVisibleCount ?? 0) * (self.viewPortHandler?.scaleX ?? 1.0))

I assume it has the same effect?

[xinranw]

The self change does account for most of the improvements. Chained operations do tack on some build time increase but it's probably not significant for this case. Happy to go ahead and update with that change.

[danielgindi]

Cool! Thanks!