How to test plot recipes?

juliohm · March 14, 2017, 1:55am

Plot recipes can reduce the test coverage considerably. For example, my coverage is currently at 76% and a big part of the remaining 24% is due to a plot recipe: Codecov

How should I go about testing plot recipes? Is there any way to design meaningful tests without human interaction?

tbreloff · March 14, 2017, 2:39am

https://github.com/JuliaPlots/VisualRegressionTests.jl

Or you could just call RecipesBase.apply_recipe with arguments that will
dispatch to your recipe.

juliohm · March 24, 2017, 2:05am

@tbreloff I really like the idea of VisualRegressionTests.jl, but apparently I am missing something about how it works. As you can see, the recipe itself is not covered by the tests that I wrote: Codecov

giordano · October 16, 2017, 12:33am

It’s not clear to me how to use this way of testing. Should the first argument to RecipesBase.apply_recipe be an empty dictionary Dict{Symbol,Any}()?

juliohm · October 16, 2017, 12:39am

As a side comment, this coverage issue doesn’t exist anymore. It was fixed somewhere in Plots.jl or VisualRegressionTests.jl.

giordano · April 24, 2020, 3:29pm

For the record, this is what I ended up doing for testing plot recipes Measurements.jl:

github.com

JuliaPhysics/Measurements.jl/blob/2ef7fadf7c2644408cff555781b9c10adfca26f6/test/plots.jl

using RecipesBase

x = [x ± x/10 for x in 0:0.1:10];
y = sin.(x);

rec = RecipesBase.apply_recipe(Dict{Symbol, Any}(), y)
@test getfield(rec[1], 1) == Dict{Symbol, Any}(:yerror => uncertainty.(y))
@test rec[1].args == (value.(y),)

rec = RecipesBase.apply_recipe(Dict{Symbol, Any}(), sin, x)
@test getfield(rec[1], 1) == Dict{Symbol, Any}(:xerror => uncertainty.(x),
                                    :yerror => uncertainty.(y))
@test rec[1].args == (value.(x), value.(y))


rec = RecipesBase.apply_recipe(Dict{Symbol, Any}(), x, y)
@test getfield(rec[1], 1) == Dict{Symbol, Any}(:xerror => uncertainty.(x),
                                    :yerror => uncertainty.(y))
@test rec[1].args == (value.(x), value.(y))

This file has been truncated. show original

Note that this is using some internals of RecipesBase.jl, so use it at your own risk

Sorry for necroposting, but I see that this thread comes is among first results when searching for plotting tests in Julia, so I thought it could be useful to point out this option.