- User Since
- Mar 20 2019, 8:25 PM (29 w, 5 d)
black n lint
now with less infinite loops
I think i am ok with the overall idea, especially once we get strict_mode or something akin in place. I will leave this open for now to give @schrockn another chance to take a look
more useful sanity check endpoint
ill file an issue
we should probably go through and gate all the wipe commands with a user prompt for a string like DELETE
its kinda gross but nothing objectionable
schedule metadata files for schedules in the current repository
all dagster-created cron tabs, regardless of which repository they are from
This is great but since it doubles the number of "runs w/ filters" methods I am going to push for the long term sustainable approach up front. This will also fix the current short coming that we can only filter one dimension at a time.
Given that we currently don't catch custom types having serialization problems til runtime - this feature could make it very easy to accidentally introduce a problematic output to a solid in a big pipeline and not catch it til N hours in to the job when it fails. Am I mistaken and we catch these errors earlier? If not I think it would be better to add that safeguard in place before we auto up new types this aggressively.
Fri, Oct 11
is_or_descends_from on SolidHandle
parse string in to SolidHandle
shift the results APIs around to better handle composites
remove flag, update docs and tests
works for me
Thu, Oct 10
@bengotow you have thoughts here?
more reliable way to get directory / path for previous run
Wed, Oct 9
move impl up in class heirarchy
round 2 lets go
Tue, Oct 8
Mon, Oct 7
how does this work with multiple tabs open?