Transpose grid_finder tick representation.

[anntzer]

Instead of representing the information about tick and gridlines as

{
    "lon_lines": [line, ...],  # unused
    "lat_lines": [line, ...],  # unused
    "lon": [
        "tick_levels": [level, ...],
        "tick_locs": {"left": [(xy, angle), ...], "bottom": [(xy, angle), ...], ...},
        "tick_labels": {"left": [label, ...], "bottom": [label, ...], ...},
        "lines": [line, ...],
    ],
    "lat": ...,  # as for lon
}

where the locs and labels are implicitly associated by the iteration order, group the information for each tick, and remove the redundant gridlines info as well:

{
    "lon": {
        "lines": [line, ...],
        "ticks": {
            "left": [
                {"level": level, "loc": (xy, angle), "label": label}, ...
            ],
            "bottom": [...], ...,
        }
    }
    "lat": ...,  # as for lon
}

(This representation turns out to make the implementation quite shorter, and makes more sense to me...)

PR summary

PR checklist

• "closes #0000" is in the body of the PR description to link the related issue
• new and changed code is tested
• Plotting related features are demonstrated in an example
• New Features and API Changes are noted with a directive and release note
• Documentation complies with general and docstring guidelines

[QuLogic]

It looks good to me, though I do have an implementation question.

[QuLogic]

I wonder if it would make sense for this interior dictionary to be a NumPy record array?

[anntzer]

Conceptually that makes sense, but I'm not overly convinced by the patch, which is just

diff --git i/lib/mpl_toolkits/axisartist/grid_finder.py w/lib/mpl_toolkits/axisartist/grid_finder.py
index 897a4152b0..a01a4df53e 100644
--- i/lib/mpl_toolkits/axisartist/grid_finder.py
+++ w/lib/mpl_toolkits/axisartist/grid_finder.py
@@ -208,12 +208,12 @@ class GridFinder:
                 for side, crossings in zip(
                         ["left", "right", "bottom", "top"], all_crossings):
                     for crossing in crossings:
-                        gi["ticks"][side].append({"level": level, "loc": crossing})
+                        gi["ticks"][side].append((level, *crossing, None))
             for side in gi["ticks"]:
-                levs = [tick["level"] for tick in gi["ticks"][side]]
-                labels = self._format_ticks(idx, side, factor, levs)
-                for tick, label in zip(gi["ticks"][side], labels):
-                    tick["label"] = label
+                gi["ticks"][side] = ticks = np.array(
+                    gi["ticks"][side],
+                    [("level", float), ("loc", float, 2), ("angle", float), ("label", object)])
+                ticks["label"] = self._format_ticks(idx, side, factor, ticks["level"])

         return grid_info

diff --git i/lib/mpl_toolkits/axisartist/grid_helper_curvelinear.py w/lib/mpl_toolkits/axisartist/grid_helper_curvelinear.py
index 451e1159ed..d7ed6e8062 100644
--- i/lib/mpl_toolkits/axisartist/grid_helper_curvelinear.py
+++ w/lib/mpl_toolkits/axisartist/grid_helper_curvelinear.py
@@ -83,7 +83,7 @@ class FixedAxisArtistHelper(_FixedAxisArtistHelperBase):
                     (self.nth_coord_ticks, True), (1 - self.nth_coord_ticks, False)]:
                 gi = self.grid_helper._grid_info[["lon", "lat"][nth_coord]]
                 for tick in gi["ticks"][side]:
-                    yield (*tick["loc"], angle_tangent,
+                    yield (*tick[["loc", "angle"]], angle_tangent,
                            (tick["label"] if show_labels else ""))

         return iter_major(), iter([])
@@ -322,8 +322,7 @@ class GridHelperCurveLinear(GridHelperBase):
         lon_or_lat = ["lon", "lat"][nth_coord]
         if not minor:  # major ticks
             for tick in self._grid_info[lon_or_lat]["ticks"][axis_side]:
-                yield *tick["loc"], angle_tangent, tick["label"]
+                yield *tick[["loc", "angle"]], angle_tangent, tick["label"]
         else:
             for tick in self._grid_info[lon_or_lat]["ticks"][axis_side]:
-                yield *tick["loc"], angle_tangent, ""
+                yield *tick[["loc", "angle"]], angle_tangent, ""

Thoughts?

[QuLogic]

I think it would be better if we actually used the arrays as a whole, but since the users (in grid_helper_curvelinear.py) are just doing a yield on each row, there doesn't seem to be too much benefit here.

[story645]

Does this need an API change note or not quite public APi?

[anntzer]

It's not touching any public API.

[story645]

Don't know what to milestone this as so you're welcome to milestone then self merge.

[anntzer]

I'll let @QuLogic chime in as to whether we want to use structured arrays here. In general I like them very much (as much as I don't like pandas 🤪) but here I'm not overly convinced by their use.