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
We're currently lacking guidelines on the testing of plots using pre-generated baseline graphics.
There's limited documentation as part of the testing files in pyfar, which should be moved to the docs (best case the joint contribution guidelines on the gallery docs in, see #51)
The text was updated successfully, but these errors were encountered:
We found that using best practice for testing differences between baseline and newly generated plots as recommended by Matplotlib does not work. Even if using the same backend, plots differ by a few pixels between operating systems and automatic testing of differences does not work.
We will use pyfar as our best practive and document that in the guidelines. @f-brinkmann will do a draft.
We're currently lacking guidelines on the testing of plots using pre-generated baseline graphics.
There's limited documentation as part of the testing files in pyfar, which should be moved to the docs (best case the joint contribution guidelines on the gallery docs in, see #51)
The text was updated successfully, but these errors were encountered: