Can we also test the error handling path?
Fri, May 29
let's look at coverage through the dagster-cron tests and then merge this if those are ok. this is not ideal but i see there are reasons for trying to ship it.
what I would like to see here is 100% branch coverage on the code in scheduler.py
Thu, May 28
Wed, May 27
I think the sync APIs really would be better served by using check_output
We should memorialize our discussion from this morning & future plans in a comment in code somewhere I think
this is great, i'd like to see a more exhaustive set of test cases (incl. multiple clauses, multiple overlapping clauses, and in the presence of composites)
requesting changes for dedupe and the docstring which is a little out of sync with the actual signature
requesting changes to surface the subprocess error info
requesting changes for clarifying comments in the docstring, this is great
Tue, May 26
i'm not sure i understand what you have in mind here - are you looking for lazy evaluation of objects other than pipeline definitions?
Fri, May 22
ah, hooray this is so good
ok for now
this looks great to me, seems like a straightforward improvement. i have a slight preference for the approach you've rejected of allowing multi-selection & reflecting it in the filter query
Thu, May 21
Wed, May 20
this is so delightful
this lgtm, i will let @bengotow have the final word
Mon, May 18
Just for the avoidance of doubt, I agree with Alex and I think this is clearer and more consistent than the alternative
I think this is a huge step forwards but will defer to @schrockn for ultimate sign off
Add tooltip for Open in Playground
Sat, May 16
Fri, May 15
I *really* like the second interaction; maybe we can talk about the first on this morning's call. It does feel a little hard to clear the subset selection.
@yuhan I think this observes those semantics, will add a test to demonstrate
Thu, May 14
add comment and reference to issue
sweet, this is much better
I just expanded the discussion of alternatives in the comments, but I'll expand on it here.
Reopening this for comment, now with pseudosql in the comments as well as more color on why I still think this is the right way to do it.
Wed, May 13
still think it would be nice if @pipeline gave you this for free