Closed kimikage closed 2 years ago
Merging #508 (11257ec) into master (9c55e4a) will decrease coverage by
0.48%
. The diff coverage is76.92%
.
@@ Coverage Diff @@
## master #508 +/- ##
==========================================
- Coverage 95.45% 94.97% -0.49%
==========================================
Files 9 9
Lines 1188 1193 +5
==========================================
- Hits 1134 1133 -1
- Misses 54 60 +6
Impacted Files | Coverage Δ | |
---|---|---|
src/precompile.jl | 0.00% <0.00%> (ø) |
|
src/utilities.jl | 98.84% <75.00%> (+0.38%) |
:arrow_up: |
src/conversions.jl | 99.39% <100.00%> (+<0.01%) |
:arrow_up: |
src/colormaps.jl | 94.39% <0.00%> (-3.74%) |
:arrow_down: |
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 9c55e4a...11257ec. Read the comment docs.
Some problematic math functions have been removed from the major paths. The problems that still remain are trigonometric functions, especially sincos
.
Since sincos
is not that bad in terms of speed and accuracy, there is not enough motivation to reimplement it.
Edit: See PR #510
RGB
<--> LCHab
/LCHuv
conversions still have problems. However, I would like to address some of the problems, in PR #509.
This uses the
cnvt
instead of the top-levelconvert
when source and destination are both known colors. This also blocks the propagation of inference failures by adding type assertions to the result ofconvert
. ~However, these changes do not solve the essential problem (#496). So, this PR is for reference only, and I have no plans to merge this for now.~ By PR #510, we can see the practical effects, e.g.:I haven't solved the problem yet, though.