Add support for Google Colab about localtileserver

[banesullivan]

It's as simple as:

from localtileserver import get_leaflet_tile_layer, examples, TileClient
from ipyleaflet import Map

client = TileClient(examples.get_data_path("bahamas_rgb.tif"))
client.client_host = f'localhost:{client.server.port}'

t = get_leaflet_tile_layer(client)

m = Map(center=client.center(), zoom=client.default_zoom)
m.add_layer(t)
m

But I'll make this a little easier in the next release

[banesullivan]

yeah... ran into that too... Google Colab installs GDAL and their build is wack (and missing that module). This was handled in afe669c and will be good on the next release

[banesullivan]

The reason it works is actually pretty specific to colab supporting this localhost remapping. Reference https://colab.research.google.com/notebooks/snippets/advanced_outputs.ipynb

Outputs in the browser can request resources from the kernel by requesting https://localhost:{port}. The protocol will automatically be translated from https to http and the localhost will be the kernel executing the code.

[banesullivan]

@giswqs, it's been a long time since I've used Colab. Could you post a link to a repo and session I could launch to debug this?

[giswqs]

Here it is
https://colab.research.google.com/drive/1tWU1Th_qmOLbixbW2HUCUW7eXEvf0Xgs?usp=sharing

[banesullivan]

I have no idea how to do this on google colab. It's seemingly not a Jupyter Hub/Lab instance and I have absolutely no idea how to expose a port on the runner. jupyter-server-proxy is really the only safe way to do this in cloud environments and I don't think that's possible with google colab

[banesullivan]

This is actually really easy! I figured it out while trying to get this to work for PyVista in pyvista/pyvista#3851

I'll see if I can add support directly in localtileserver

[giswqs]

@banesullivan That's awesome to hear. Can't wait to try it out.

[giswqs]

Excited to try this out. Got an error. I should probably wait for the next release.

[banesullivan]

So, this is fixed by de4d2ef

What do you think, @giswqs?

from localtileserver import get_leaflet_tile_layer, examples, TileClient
from ipyleaflet import Map

path = examples.get_data_path("bahamas_rgb.tif")

client = TileClient(path)
client.enable_colab()

t = get_leaflet_tile_layer(client)

m = Map(center=client.center(), zoom=client.default_zoom)
m.add_layer(t)
m

[giswqs]

This gives me hope that localtileserver might work with all major Jupyter notebook platforms, including AWS SageMaker.

client.client_host = f'localhost:{client.server.port}'

[giswqs]

You may not need that client.enable_colab() explicitly. You might be able to integrate this into TileClient.

import sys
if "google.colab" in sys.modules:
  self.client_host = "localhost"
  self.client_port = True

[banesullivan]

Ah, good point. This assumes google.clolab is only ever installed on colab... which is probably a safe assumption

[giswqs]

Correct. The google.colab module can only be found on Colab. That module can't be found on any other Jupyter platforms or local computers.

[banesullivan]

Check out https://colab.research.google.com/drive/1h1iTDl4KqbXxKiyox9eM211vAg6_ONKT?usp=sharing (you have to restart the runtime after installing localtileserver)

0.6.4 now autodetects google colab and all should work

[giswqs]

It worked like a charm! Thank you very much for this!