Open JohnMav opened 4 years ago
I want to add a big +1 here and add a bit more context about why this is something that would be extremely useful. Many organizations (mine included) have a combination of R and Python scripts used in their pipelines. In particular, R is really powerful for statistical modeling and has a huge number of packages that don't exist in Python. It's not a great language for writing production code, but it's hard to move entirely off of it into Python.
I love a lot of the motivations behind Dagster (typing, testing, etc.) and would like to bring that to as much of our pipelines as possible. I understand it's probably hard to do that, so even just having a simple R operator that passes dataframes between R and Python would be incredibly powerful. It seems like the current workaround involves running a shell script or Docker container to pass data via files -- which works, but reduces a lot of the power of Dagster.
I've only just started using Dagster, so I'm not familiar enough to know how easy/hard this is or where I would even begin to contribute a feature like this, but wanted to make sure this issue doesn't die!
I've collected a few mentions from Slack where other people have mentioned this as something they would like to do: (Original post that spawned this issue): https://dagster.slack.com/archives/CCCR6P2UR/p1592431695124700 https://dagster.slack.com/archives/CCCR6P2UR/p1614100503079800 https://dagster.slack.com/archives/C01U954MEER/p1638575143351700?thread_ts=1638550558.333200&cid=C01U954MEER https://dagster.slack.com/archives/C014N0PK37E/p1641375596083800
Since 0.8.0 the separation of Dagster Host and User process has allowed creating workspace environments with different versions of Python. For teams that have users that work primarily in R driven environments it would be amazing to be able to designate a workspace environment as a specific R version and be able to run dagster pipelines within that environment.
issue #1585 raises the notion of supporting polyglot notebooks but would love for this to go a step further and support tools like RMarkdown/Shiny apps in a similar fashion.