to_cassandra(table, session)

[rustyrazorblade]

Adding cassandra support should be very similar to to_sql() but is different enough where I think it needs it's own call, as it wouldn't use SQLAlchemy or behave even remotely close.

Python driver is here: https://github.com/datastax/python-driver

Willing to help on this if it'll get pulled in.

[cpcloud]

This might be more suitable for odo. @rustyrazorblade thoughts on doing a PR there?

[rustyrazorblade]

Is all future work on interacting with different Databases going to the odo project? I don't mind doing the PR there - I just want to make sure Cassandra is a first class citizen in the Pandas world.

[cpcloud]

Is all future work on interacting with different Databases going to the odo project?

Not sure.

Pros:

• Odo's entire job is converting things to other things and is built with extensibility in mind. Pandas IO system isn't extensible without tacking on a new method.
• As soon as you hook into the odo graph you get conversion to nearly every other file format without having to write any additional code

Cons:

• Pandas has a lot of users, odo doesn't
• By extension, Pandas has familiar interface and odo doesn't

[rustyrazorblade]

I suspect having that functionality in odo is a good thing no matter what. I'll take a look at it and see if I can get a PR in there for Cassandra.

[cpcloud]

@rustyrazorblade awesome! great to have more contributors there

[jreback]

@rustyrazorblade I think go ahead and contribute this to odo. If it makes sense we can then port back to pandas.

[jorisvandenbossche]

I think this is currently out of scope for core pandas, therefore closing this.
But it would certainly make for a nice standalone package if someone is interested (and we may include some mechanism to plug such extension packages into pandas, see #9378)

[kassett]

@mroeschke I see that the odo repository mentioned above has not been maintained for the last 7 or 8 years. I have personally implemented the to_cql functionality in several projects and think it would be tremendously useful to have it in Pandas, especially given that Cassandra has become very popular in the last few years.
I have two questions

1. Can I implement this in Pandas?
2. Writing to Cassandra is incredibly slow if you don't implement multi-processing. Is this something that we are okay with doing in Pandas?

[mroeschke]

It's probably still too niche to belong in pandas, but if you create a Python package with it we can add it to the ecosystem docs