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
Define FloatingAxes boundary patch in data coordinates.
Currently, the boundary patch of FloatingAxes is drawn in "parent axes
data" (rectilinear) coordinates, using manual interpolation as defined
in get_boundary.
Instead, it can be defined in child axes data coordinates, and use
`helper_trf + transData` as transform, relying on interpolation
mechanisms already present in the transform machinery (for example, this
means that polar FloatingAxes benefit from the more-accurate
transforming of circular arcs).
Likewise, adjust_axes_lim can use normal path methods to get its
extents, rather than relying on manual interpolation. This is the cause
of the baseline image tolerance change: in test_curvelinear3, for
example, we now have exactly `bbox.xmin = bbox.ymin = -10` and
`bbox.xmax = bbox.ymax = +10` rather than slightly different values
arising from interpolation and inaccurate transform composition. (But
one can manually check that the changes are all just tiny shifts.)
(On the other hand, I'll likely further change these baselines in the
future, so let's not update the files right now.)
0 commit comments