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
This is mainly for mpld3, exporting figure-level text objects (such as
fig.suptitle) would never get exported. It looks like figure-level text
objects (and objects in general) was completely missing here. This adds
text objects, I might add more in the future.
The alternative was to put figure text objects into the first axis
object, but that's extremely hacky, so I went for the larger but proper
fix instead.
I did this together with gpt-5.1-codex, not alone. Here is what it has
to say:
- Exporter now emits figure-level text (suptitle + fig.text) via a dedicated draw_figure_text call before crawling axes; figure transforms passed directly to process_transform instead of shoving text into the first axes.
- Renderer API gains a draw_figure_text hook (no-op default in base) so non-mpld3 renderers don’t break; FakeRenderer already implements it.
- Figure JSON now carries a texts array and MPLD3Renderer serializes figure-level text entries with proper coordinates/attrs; tests cover presence/positions of exported figure texts.
0 commit comments