baseline fix for bar plots using log scales

[t-bltg]

Fix #4502.

At least values within the same decade are unified to the same value (decade start).
Would that be better @giordano ?

bar([1 2 3], [11 25 85]; yscale=:log10)

pr

master

The problem lies in the way we process series.
Can we preprocess a common minimal value before mapping to bar series ?

• check performance

[BeastyBlacksmith]

Can't we just set fillrange --> 1 in the bar recipe if the yscale is a logscale?

[t-bltg]

Could be a cheap solution too.

[giordano]

For what is worth, the original data in my example span multiple decades (10 and 100 mainly), reason why I was trying to use logscale

[codecov]

Codecov Report

Base: 90.65% // Head: 91.05% // Increases project coverage by +0.39% 🎉

Coverage data is based on head (378d4c1) compared to base (dbb33b8).
Patch coverage: 94.00% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4503      +/-   ##
==========================================
+ Coverage   90.65%   91.05%   +0.39%     
==========================================
  Files          40       40              
  Lines        7353     7688     +335     
==========================================
+ Hits         6666     7000     +334     
- Misses        687      688       +1     

Impacted Files	Coverage Δ
src/args.jl	84.66% <ø> (+0.69%)	⬆️
src/shorthands.jl	96.15% <ø> (ø)
src/recipes.jl	84.21% <90.90%> (+0.72%)	⬆️
src/pipeline.jl	96.18% <100.00%> (+0.36%)	⬆️
src/utils.jl	93.60% <100.00%> (+0.37%)	⬆️
src/plot.jl	97.77% <0.00%> (-0.61%)	⬇️
src/consts.jl	100.00% <0.00%> (ø)
src/backends/plotlybase.jl	100.00% <0.00%> (ø)
src/animation.jl	99.04% <0.00%> (+<0.01%)	⬆️
src/axes.jl	90.15% <0.00%> (+0.06%)	⬆️
... and 25 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[t-bltg]

We could compute global series extrema somewhere here https://github.com/JuliaPlots/Plots.jl/blob/master/RecipesPipeline/src/series_recipe.jl#L10-L29 ...

If we choose an arbitrary default value of 1, we'll get the same issue for e.g. bar([1 2 3], [0.01 10 100]; yscale=:log10).

[t-bltg]

Computing extrema along all series now allows the following:

bar([1 2 3], [0.02 125 10_000]; yscale=:log10)

pr

master

Better ?

[giordano]

Looks great! And I can still tweak it with fillto if I need, right?

[t-bltg]

Yes, as long as fillto (alias) fillrange is positive (for log scales). Can also be a vector of positive values.