Zach P

09/19/2022, 8:47 PM
Question: What’s the role of graphs, and ops going to be in the future? While they’re obviously baked into assets and how they’re made, I’ve found that it’s sometimes its hard to figure out how to do things with assets where they’re quite well documented with ops or graphs. (EG: Setting up an asset to use output_config_schema from io managers). Moving forward, is best practice to be improving asset api and moving things to that, or is it expected that assets are best suited to more “simple” datasets? thanks!


09/19/2022, 8:50 PM
We plan to support both, for advanced use cases. Assets + I/O managers is a known documentation gap that we want to fix soon. Are there other areas where you've encountered gaps with asset docs? cc @erin
👍 2

Zach P

09/19/2022, 8:58 PM
Thanks Sandy! Nothing as concrete as the previous example I gave, but I guess a few points would be helpful: • More examples would be very helpful. Lots of the existing stuff are either quite simple, or written with ops/graphs/etc. in mind. • how partitions, jobs, and assets all work together. There’s some docs on this already and I’m still wrapping my head around it, but it feels a little bit spread out. I find myself jumping around a lot to try and understand what I should be doing and such.
☝️ 1
👍 2


09/20/2022, 12:27 AM
Hi Zach, we’re actively working on more examples. Meanwhile, here’s a guide for a fully feature project which includes multiple dagster concepts like assets, jobs, partitions, schedules, etc. you can find the example code here:
👍 1