Dev/sso login ux

[vaimdev]

An attempt to improve UI/UX to the databricks dashboard with SSO usage
Added a few helper functions, so that the databricks dashboard can keep a simpler flow and UX.
Example databricks notebooks that utilize the new helper functions
https://dbc-ac10efa1-77c5.cloud.databricks.com/editor/notebooks/3715908489050384?o=7408253037638600#command/1247495116400904

3 pre-steps for the SSO login using the helper functions (with different cells), followed by the actual plotting python code

graphistry.databricks_sso_login_init()

graphistry.databricks_sso_login(server="test-2-40-74-dbt.grph.xyz", org_name="louiedev")

graphistry.databricks_sso_wait_for_token()

[lmeyerov]

Can you share the code sample in the PR?

(Trouble accessing link, and discussion should be here)

[lmeyerov]

@vaimdev @DataBoyTX This looks strange from a dev UX layer, afaict:

Demo

If I understand correctly, the below is the intended setup:

cell 1:

clear_output()
graphistry.databricks_sso_login_init()

cell 2:

clear_output()
graphistry.databricks_sso_login(server="test-2-40-74-dbt.grph.xyz", org_name="louiedev")

cell 3:

clear_output()
graphistry.databricks_sso_wait_for_token()

cell 4:

g.plot()

Discussion

I understand wanting new underlying primitives, but how to expose it to users should be minimal, clean, integrated, smart defaults, etc

• Why is this 4 cells instead of 2?

• why clear_output() ? Can we make that default-on parameter for the auth step with pass-in overrides?

• I believe part of this weird path is SSO differences in how databricks headless, databricks notebooks, and databricks dashboards work; are these primitives for dashboards only, or notebook + headless too?

Examples of a two-cell: graphistry.register(sso='databricks').... g.plot():

[lmeyerov]

This PR should also include sample ipynb + docs updates. Sounds like we need to iterate first on interface definitions tho.

The current is https://pygraphistry.readthedocs.io/en/latest/demos/demos_databases_apis/databricks_pyspark/graphistry-notebook-dashboard.html

Maybe dashboards or auth should split out?

[lmeyerov]

public methods should have .rst-friendly docstrs

[lmeyerov]

also, we're trying to make all new code mypy-typed

[vaimdev]

ok, noted. I will add them in.

[vaimdev]

@vaimdev @DataBoyTX This looks strange from a dev UX layer, afaict:

Demo

If I understand correctly, the below is the intended setup:

cell 1:

clear_output()
graphistry.databricks_sso_login_init()

cell 2:

clear_output()
graphistry.databricks_sso_login(server="test-2-40-74-dbt.grph.xyz", org_name="louiedev")

cell 3:

clear_output()
graphistry.databricks_sso_wait_for_token()

cell 4:

g.plot()

Discussion

I understand wanting new underlying primitives, but how to expose it to users should be minimal, clean, integrated, smart defaults, etc

• Why is this 4 cells instead of 2?
• why clear_output() ? Can we make that default-on parameter for the auth step with pass-in overrides?
• I believe part of this weird path is SSO differences in how databricks headless, databricks notebooks, and databricks dashboards work; are these primitives for dashboards only, or notebook + headless too?

Examples of a two-cell: graphistry.register(sso='databricks').... g.plot():

4 cells instead of 2, mainly because each cell will only display the output at the end of running of 1 cell. Previously, there are a few logic written into 1 cell and the output only be displayed after the cell is run which make the messages looks strange,
in some cases, the refresh token will take place and the message shown is not correct.
breaking it into 4 cells will make the the UX consistent and able to control the logic in the case of refresh token, to display the correct message

why clear_output() ? Can we make that default-on parameter for the auth step with pass-in overrides?

clear_output() is not necessary, can be removed.

I believe part of this weird path is SSO differences in how databricks headless, databricks notebooks, and databricks dashboards work; are these primitives for dashboards only, or notebook + headless too?

Mainly for databricks dashboard but tested in both databricks notebook and databricks dashboard.

[lmeyerov]

this error case handling may need thinking?

[lmeyerov]

un-nest

[vaimdev]

suggestion from Leo:

Show a message on Graphistry server, when SSO login from pygraphistry

[lmeyerov]

@vaimdev FYI ci lint issues

[vaimdev]

@vaimdev FYI ci lint issues

Yes, noted, added some debugging code, now it is done and cleaning up code.

[vaimdev]

Loom videos to demo the changes

Databricks Notebook
Not login - https://www.loom.com/share/b48939faf5ba4578bc66218cb89cacb9?sid=c72bae54-9ce9-4128-8f2a-ebd128e70dad

Log in - https://www.loom.com/share/2313bb86ef454a9f9d55c1535c785457?sid=42f46537-f726-4a89-a98d-96f762e0cf51

Dashboard
https://www.loom.com/share/68add6bd0c2348e0b0fc0b19abbf4939?sid=2a878545-e92c-4dca-843f-2f297d4fd52f

[lmeyerov]

@sabizmil can you check all public method docstr as these show up in the readthedocs?

Spelling, grammar, docstr types, and examples, following format of the rest

[sabizmil]

Updated in the latest commit

[sabizmil]

Left comments with updated message content.

[lmeyerov]

@vaimdev @sabizmil @DataBoyTX i'm updating def plot() as well as part of some cleanup, and realized it'll conflict. Happy to land mine after this goes in -- any sense of timing here?

[vaimdev]

@vaimdev @sabizmil @DataBoyTX i'm updating def plot() as well as part of some cleanup, and realized it'll conflict. Happy to land mine after this goes in -- any sense of timing here?

@lmeyerov , my side I am still updating the code with documentation with lower priority because focus on the entitlement Saas. I'll suggest you landed your cleaned up version first