You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In #145, I just needed to update a bunch of reftest images again. We somehow need to change how we handle them. Problems:
All assets are stored in the git repo. This is not ideal since the git repo will grow over time.
Currently, the reftests are numbered based on their ordering in the containing file; therefore, if you want to introduce another reference test in between, all the following tests change their name.
There is no "interactive" way of going through the tests and accepting updates.
On the other hand, I like the simplicity of the system, and I like that I only need to mark plot commands in the docs with @reftest to test them. Some time ago, I created a experimental reftest package based on those ideas in GraphMakie, https://github.com/hexaeder/PlotReferenceTests.jl which solves 2 by changing the syntax to something like
@reftest "reference name" plot(...)
Problem 3 is also easy to solve by implementing some tooling.
However, I don't really know how to deal with 1, which blocks further improvements. Ideas?
The text was updated successfully, but these errors were encountered:
I believe that we upload tarballs as attachments to version tags in Makie, you might find some use for the ReferenceTools and ReferenceUpdater folders in the main repo!
In #145, I just needed to update a bunch of reftest images again. We somehow need to change how we handle them. Problems:
On the other hand, I like the simplicity of the system, and I like that I only need to mark plot commands in the docs with
@reftest
to test them. Some time ago, I created a experimental reftest package based on those ideas in GraphMakie, https://github.com/hexaeder/PlotReferenceTests.jl which solves 2 by changing the syntax to something likeProblem 3 is also easy to solve by implementing some tooling.
However, I don't really know how to deal with 1, which blocks further improvements. Ideas?
The text was updated successfully, but these errors were encountered: