left_join is usually the right join: https://milesmcbain.micro.blog/2026/06/27/leftjoin-is-usually-the-right.html
Remote
MilesMcBain
@milesmcbain@fosstodon.org
mastodon 4.7.3Code hacker, number cruncher, #rstats user, board gamer, road racer, plant eater, bass slapper. Coming to you from AUS / BNE.
http://micro.blog/milesmcbain
0 Followers
0 Following
10 Posts
Joined November 04, 2022
Website:
Micro Blog:
Github:
Replying to
@defuneste @lwpembleton @drdcarpenter @mschmidty to the point where the script behind it is a labyrinth of edge case handling.
Better to express as a collection of smaller functions with a little bit of glue code here and there. Wrap in targets plan if you want to save some time.
Replying to
@defuneste @lwpembleton @drdcarpenter @mschmidty more complex schemes become riddled with issues like that.
Dive()ing into the hunt #rstats: https://milesmcbain.micro.blog/2025/07/04/diveing-into-the-hunt-rstats.html
Replying to
@mac@jawns.club very excellent points. The other thing I was thinking about recently was let’s not pretend we’re in the same economy we were prepandemic. Working class families that want to stay afloat, let alone get ahead need to be doing as many hours of work as they can get. That means both parents working. Everyone else has the same idea so spots in care for kiddies of all ages are very rare right now. WFH some of the time is the only way to make it work.
Replying to
@cutterkom@mastodon.social @djnavarro@fosstodon.org Excellent question. I had some related thoughts: https://www.milesmcbain.com/posts/assertive-programming-for-pipelines/
The gist being you're better off doing assertive programming because that ensures every pipeline run adheres to yours assertions.
But there's still a related question which is: Do I want to fail fast as soon as any assertion fails? or do I want to check all my assertions and record data about each failure?
I use both strategies at different times.
Replying to
@cutterkom@mastodon.social @djnavarro@fosstodon.org The difference is mainly:
If the object is just an intermediate object then it tends to be fail fast. If the object is a kind of output in itself - I.e. it may get into the hands of a client, or it represents some kind of important checkpoint - in that future work may branch off from it - then I tend to run something like a 'check suite' that performs a bunch of validation steps and records the results of all.