ENH: explore method for graph

[knaaptime]

(the graph is actually based on network distance so that water observation is a known snapping issue :P)

[codecov]

Codecov Report

Merging #617 (ce54246) into main (3976997) will increase coverage by 0.1%.
The diff coverage is 96.3%.

@@           Coverage Diff           @@
##            main    #617     +/-   ##
=======================================
+ Coverage   84.4%   84.5%   +0.1%     
=======================================
  Files        139     139             
  Lines      14857   14963    +106     
=======================================
+ Hits       12536   12638    +102     
- Misses      2321    2325      +4     

Files	Coverage Δ
libpysal/graph/base.py	97.6% <100.0%> (+<0.1%)	⬆️
libpysal/graph/tests/test_plotting.py	100.0% <100.0%> (ø)
libpysal/graph/tests/test_utils.py	98.8% <100.0%> (+0.1%)	⬆️
libpysal/graph/_plotting.py	92.9% <89.5%> (-2.8%)	⬇️

[martinfleis]

Shall we aim for API parity with static plots here?

[martinfleis]

@knaaptime a side note as I noticed the code in the screenshot above - for explore, CRS makes no difference. Map is always EPSG:3857 and is reprojected under the hood.

[knaaptime]

@knaaptime a side note as I noticed the code in the screenshot above - for explore, CRS makes no difference. Map is always EPSG:3857 and is reprojected under the hood.

Ah, good call. I get it mixed up with contextily where I need to set it

[knaaptime]

Shall we aim for API parity with static plots here?

Not sure where folks wanna take the api. Don’t have much of an opinion but figured this is enough to get us started

[knaaptime]

[knaaptime]

here's a simpler example. It's a touch verbose, i think that just is what it is. Should match the plot api now, i think

[martinfleis]

This is great and apart from that one suggestion works as expected. Nice! Do you want to give a go to tests (ideally mirroring those from plot) or shall I do that?

[knaaptime]

@martinfleis i'll take the first pass, but would probably be speediest if we tag team again. I'll try and do a thorough first draft. Would be interested to cut RCs for lib and esda today for kicking the tires

[knaaptime]

related, are there common patterns like this one that are becoming common enough to centralize in a testing module of lib? (a la numpy?). We now have explore in esda, pointpats, and here... its a tiny function so nbd, but there are probably other patterns used across the meta (and i know the original is from geopandas, i just mean there are probably pysal-common testing patterns we could stdize)

[martinfleis]

The original is from folium :D. I think it makes sense to centralise stuff as assert_series_equal that does multiple things but utils like this can easily live in each. But I don't mind if we move it to libpysal.testing and re-use it across pysal

[knaaptime]

looks like this will also require a minimum pin for a recent geopandas

[martinfleis]

We'll need to include a version of this as we currently plot both ij and ji. I will take back to this later today and expand tests a bit to ensure similar coverage we have in plot. I wanted to do it now but have family duties :).

libpysal/libpysal/graph/_plotting.py

Lines 106 to 107 in 3976997

	# avoid plotting both ij and ji
	edges = np.unique(np.sort(np.column_stack([codes]).T, axis=1), axis=0)

[martinfleis]

Did a bit of refactor based on the code from plot but as a result we get the focal IDs for free so it simplifies the code a bit as well (no need to create that adj dataframe.

I think it is ready. It is smooooth!

[knaaptime]

sweet!

[knaaptime]

i need one of you two to approve and merge though, :P

then lets do an RC?

[jGaboardi]

All looks great. One minor comment.

[martinfleis]

then lets do an RC?

Do we need RC? Can we just do 4.9.0?

[knaaptime]

we can just do 49 if you like

[jGaboardi]

My opinion is that it's never a bad idea to do an RC just to make sure things are working properly. On the other hand, we can delete tags & releases if something is broken.

[martinfleis]

I mean, what do you expect to test by RC in this case?

[knaaptime]

the only thing with an "unnecessary" RC is that the changelog tooling techninically fails and you need to copy over the changelog

[jGaboardi]

Making sure the actions (doc build & release) still work as they should. But I won't be stickler on that. Let's just try for the v4.9.0

[martinfleis]

Hold on with that, there's a bug in this PR

[knaaptime]

aw crap lol i just pushed the tag

[knaaptime]

looks like we're doing a 491 today 🙃

[martinfleis]

it'll be 4.9.1 then :D

[jGaboardi]

I done told yall! 🤣

[martinfleis]

Sorry @jGaboardi, we should listen to you more :D