Geo grid fixes

[etpinard]

fixes #2896 (finally) making geo.(lon|lat)axis.tick0 work by computing grid values using Axes.calcTicks.

cc @plotly/plotly_js - I'd like to slip this one in 1.46.0.

[archmoj]

Great!
Concerning the baselines changes,
@etpinard could you please elaborate why longitudes are changed on polar stereographic projection:
https://github.com/plotly/plotly.js/pull/3706/files?short_path=2b60c83#diff-2b60c83395ec0c848268b4c720e7d73b
Thanks.

[etpinard]

could you please elaborate why longitudes are changed on polar stereographic projection:

We have a slightly different tick algo than d3.geo.graticule.

[archmoj]

I think the new longitude lines displayed on the polar stereographic projection may be off by one/half pixel or a distance.
For example 20N should cross Hilo island https://www.mapsofworld.com/usa/states/hawaii/lat-long.html.

Old:

New:

[etpinard]

Sure, there's a diff. Now, is the old baseline correct, or is the new baseline better?

Oh well, I guess I hold this one out of 1.46.0

[archmoj]

Sure, there's a diff. Now, is the old baseline correct, or is the new baseline better?

Hard to tell. To my eyes the old baseline appear to be slightly more accurate if we compare against Hawaii.

[etpinard]

@archmoj please have a look at fdfca3b where by changing the tick0 dflt, we retrieve something almost identical to the baselines on master.

[archmoj]

Excellent.
@etpinard could you please increase the width and the height of this mock: test/image/baselines/geo_tick0.png

[archmoj]

Looks great!
💃 when you wanted to start 1.47.0.

[etpinard]

when you wanted to start 1.47.0.

I'll wait for Monday.

[etpinard]

Ok, no new bug reports from 1.46.0 today, so let's start merging stuff for next week's 1.47.0