Page MenuHomePhabricator

update nothing docs now that we have IO managers
AcceptedPublic

Authored by sandyryza on Wed, Jan 13, 9:58 PM.

Details

Reviewers
nate
Test Plan

bk

Diff Detail

Repository
R1 dagster
Branch
nothing-doc (branched from master)
Lint
No Linters Available
Unit
No Unit Test Coverage

Event Timeline

Harbormaster returned this revision to the author for changes because remote builds failed.Wed, Jan 13, 10:18 PM
Harbormaster failed remote builds in B24317: Diff 29593!

hmm shouldn't this:

if the body of one solid creates a particular table in a database and the body of another solid consumes that table

be modeled w/ IOManagers instead of Nothings?

Is there a better motivating example in the post-0.10.0 world? Maybe a scenario where you know a solid produces a side-effect in an external system (API call, etc.) but otherwise don't have the ability to model that in Dagster's programming model?

This revision is now accepted and ready to land.Wed, Jan 13, 11:23 PM

shouldn't this be modeled w/ IOManagers instead of Nothings?

This might be a little extreme, but I think basically any side effects "should" be modeled with IOManagers. However, there are many mundane reasons why people might choose not to: e.g. calling an external library that invokes sql_con.execute("create table as select..."). I'll try to explain this a little better.