Jupysql overrides Databricks `%sql` magic despite alias

[danielsparing]

What happens?

In Databricks, once jupysql is loaded, the original Databricks %sql magics are unusable, even though the docs say "you can use both".

Probably in a Databricks enviromnent only the %jupysql alias should be added.

To Reproduce

In a Databricks notebook:

%pip install jupysql --quiet

%load_ext sql

%sql
select 1
# Returns UsageError: No active connection.

The above error hints that %sql was trying to call jupysql.

OS:

Linux

JupySQL Version:

0.10.17

Full Name:

Daniel Sparing

Affiliation:

Goudappel

[edublancas]

shouldn't be too hard to do it. perhaps we can look at some env var to detect the databricks environment. worst case we can add a manual flag

either we can dynamically add the magic names, or add both and remove sql here

the latter seems to work:

In [1]: from sql.magic import SqlMagic

In [2]: del SqlMagic.magics["cell"]["sql"]

In [3]: del SqlMagic.magics["line"]["sql"]

In [4]: %load_ext sql

In [5]: %sql
UsageError: Line magic function `%sql` not found.

In [6]: %jupysql
UsageError: No active connection.

To fix it:

Pass a valid connection string:
    Example: %sql postgresql://username:password@hostname/dbname

OR

Set the environment variable $DATABASE_URL

For more details, see: https://jupysql.ploomber.io/en/latest/connecting.html
If you need help solving this issue, send us a message: https://ploomber.io/community

In [7]:

I just had to modify set_configs to loadjupysql instead of sql, since `jupysql will always be present.

PRs welcome!

[danielsparing]

Thanks!

I did put it in a PR, successfully tested it in Colab and Databricks, but haven't actually written any tests as there I wouldn't even know where to start.