Should LongitudeFormatter and LatitudeFormatter convert projected coordinates to dms by default?

[gepcel]

Two examples first:

Example 1: notice the longitude labels of 2nd axis:

fig, axs = pplt.subplots(ncols=2,proj=['cyl', 'merc'],share=False)
axs.format(labels='all')
axs[0].format(lonlocator=(0.6, 113.25))
axs[1].format(lonlocator=(0.6, 113.25))

Example 2:

fig=plt.figure(figsize=(8,3))
ax0 = fig.add_subplot(1,2,1,projection=ccrs.PlateCarree())
ax0.gridlines(draw_labels=True,xlocs=(0.6, 113.25))
extent = ax0.get_extent(crs=ccrs.PlateCarree())
ax1 = fig.add_subplot(1,2,2, projection=ccrs.Mercator())
ax1.gridlines(draw_labels=True,xlocs=(0.6, 113.25))

Seems that cartopy converts coordinates of mercator to platecarree by default (like 66791.6944759644 to 0.6°E), but ultraplot does not.
And also, the ax.set_xitcks() method of cartopy also has a crs parameter (doc).
But lonlocator_kw of ultraplot doesn't seem to accept crs, am I missing something?

[cvanelteren]

Could be an issue introduced by the sharing; Proplot had no implementation for the sharing. Can put it on the docket.

[cvanelteren]

Can confirm that dms is not properly handled after sharing is added. Need to figure out where this goes wrong exactly

[cvanelteren]

Found the issue (line 1710)

[cvanelteren]

import ultraplot as uplt, numpy as np, pandas as pd
l = (0.6, 113.25)
fig, ax = uplt.subplots(ncols=2, proj=("cyl", "merc"), share=0)
ax.format(
    land=True,
    labels="both",
    lonlocator=l,
)
uplt.show(block=1)

[cvanelteren]

Fix is on #293 will need to wait for review. Thanks for reporting. I am not a geo scientist so this area is a bit of blind spot to me.

[gepcel]

The default longitude (and latitute) formatter cartopy used is decimal degrees. But ultraplot used dms (which I personally prefer), so it should be 0°36′ for 0.6, and 113°15′ for 113.25, like the formatter of axs[0] in example1.

[cvanelteren]

For my understanding the mercator projection should yield the same labels or are they correct? I am currently getting:

[cvanelteren]

My understanding is that the plot above shows both in DMS coordinate system. I misunderstood the original question, and how the DMS is propagated.

[cvanelteren]

@gepcel If I understand the issue correctly when adding a subplot iteratively you get the wrong locator when accessing through gridlines? That is you would expect the 0.36 for the cylindrical projection but not for the mercator?

fig = uplt.figure(figsize=(8, 3), share=0)
ax0 = fig.add_subplot(1, 2, 1, proj="cyl")
ax1 = fig.add_subplot(1, 2, 2, proj="merc")

ax1.format(labels=True, lonlocator=(0.6, 113.25))
ax0.format(labels=True, lonlocator=(0.6, 113.25))

this looks fine, but the gridlines approach should also work. Did I correctly summarize the issue?

[gepcel]

My understanding is that the plot above shows both in DMS coordinate system. I misunderstood the original question, and how the DMS is propagated.

The lonlabels of the 2nd ax should also be 0°36′E、113°15′E.

A little extra explanation: 0.6°E is equal to 0°36′E, which are the decimal degree format and degree-minute-second format of geodetic coordinates respectively. It is also approximately equal to 66791.6944759644, which is the coordinate format of the Mercator projection coordinate system. So the 1st ax of example 1, and the 1st and 2nd ax of example 2 all give the correct lon labels, I think (maybe someone else will give different answers). But the 2nd ax of example 1 is wrong.

I can give an example of cartopy, showing the same xticks in different lon/lat format and different crs. Besides the lon labels, I also wonder, if it's possible to add a crs kw arg to lonlocator_kw like ax.set_xticks(ticks, crs=...)

fig, (ax0, ax1, ax2) = plt.subplots(1, 3, subplot_kw={'projection': ccrs.PlateCarree()}, figsize=(9,2))
a
x0 = fig.add_subplot(1,3,1, projection=ccrs.PlateCarree())
ax0.set_xticks([0.6, 113.25])
ax0.xaxis.set_major_formatter(LongitudeFormatter(zero_direction_label=True, dms=False))

ax1 = fig.add_subplot(1,3,2, projection=ccrs.PlateCarree())
ax1.set_xticks([0.6, 113.25])
ax1.xaxis.set_major_formatter(LongitudeFormatter(zero_direction_label=True, dms=True))

ax2.set_xticks([66791.6944759644, 12606932.332338281], crs=ccrs.Mercator())
ax2.xaxis.set_major_formatter(LongitudeFormatter(zero_direction_label=True, dms=True))

plt.tight_layout()

[cvanelteren]

Thanks for the additional context. From your last reply, the dms should be formatted in the same manner for the projections listed above. What I can confirm is that we are passing dms to LongitudeFormatter (and similarly for latitude). These formatters are under ticker.py and they are a mere reference to the cartopy ones.

For example, the _CartopyFormatter is a reference (if it exists) to _PlateCarreeFormatter from cartopy if cartopy can be imported. It overrides __call__ only by calling it within a context but still uses parent class to make the call.

So I am stuck with this puzzle: Clearly the formatting is not the same but I cannot see where the formatting goes wrong. Putting down some observations

• Setting the longitude range correctly moves the view to the set range for all projections
• dms is passed to both projections by default and set by looking at the property _dms on the xformatter
• I can see that most formatters are thin wrapers of cartopy's formatters/tickers.

[gepcel]

import ultraplot as uplt, numpy as np, pandas as pd
l = (0.6, 113.25)
fig, ax = uplt.subplots(ncols=2, proj=("cyl", "merc"), share=0)
ax.format(
land=True,
labels="both",
lonlocator=l,
)
uplt.show(block=1)

Using your previous example, by commenting the following 2 lines of geo.py

1700        # gl.xformatter = self._lonaxis.get_major_formatter()
1711       # gl.yformatter = self._lataxis.get_major_formatter()

And adding the LongitudeFormatter may give the correct result. I didn't find out why those two lines would cause the lonformatter fall back to decimal degree.

import ultraplot as uplt, numpy as np, pandas as pd
from cartopy.mpl.ticker import LatitudeFormatter, LongitudeFormatter
l = (0.6, 113.25)
fig, ax = uplt.subplots(ncols=2, proj=("cyl", "merc"), share=0)
ax.format(
    labels="both",
    lonlocator=l,
    lonformatter=LongitudeFormatter(zero_direction_label=True, dms=True),
)
uplt.show(block=1)