TFW you think your last command at the CLI is the one you need to run again, but it’s actually the “exit” command from a remote session you were just running and there goes your session.
Remote
Willow, Data Witch
@willow_datawitch@hachyderm.io
mastodon 4.7.3Mom, wife, board gamer, and home cook, with a small but growing interest in solarpunk. When I have to make money, I’m a data engineer and coach. 🏳️⚧️
0 Followers
0 Following
19 Posts
Joined December 12, 2023
Pronouns:
She/Her
What in the name of rainbows is a “zero-width no-break space,” and why did powershell start jamming them into my clipboard when I pipe to clip.exe at some point around 2020? AAAAARG!
Ah, the nice feeling of processing your data model on rented hardware, so it might actually finish before I go fully grey.
Replying to
Seriously, what happened to my tabular model? I confirmed it isn’t insufficient memory, I rolled back one version of the VM image, I processed a toy-sized subset of the models tables successfully, the rest of the model processes if I drop a central fact table THAT HAS NO CURRENCY DATA TYPES IN IT.
I guess I could limit the underlying data in the fact table to a week ago and see if that helps… that would indicate weirdness in recent data OR a data type maximum being recently exceeded (everything SHOULD have decades of room under the data type maximums, in theory. Though… maybe they don’t?). I could roll back another vm image version (why doesn’t MS datestamp those!? Arg!). I could check if Tabular Editor has any recent patches and roll back to before those.
Sigh. What do you think I should do, #RubberDuckDebugging?
Replying to
@Foritus that’s just “eyeball commit hashes from gitk open next to the commandline,” right? I can get behind that. Yay!
The more I build my own tooling for exploring a database before I build a reporting query, the more I just do all my work at the command line. I don’t need #SSMS to show me my results, I just dump them to the cli output.
This post brought to you by my use of #dbatools and the Invoke-DbaQuery command specifically. That plus my data profiler query.
TFW you look at the transaction table and it’s 16 GB with one single-column index on not-the-column you need it on.
Replying to
Thank you all! The answer wound up being: launch the gitk gui and use the right-click menu to cherrypick a few commits after creating a new branch and switching to it.
Replying to
It wasn’t the aggregates! It was ID values that were not sequentially generated by the source system, but were instead arbitrarily enormous INT values stored in decimals.
Sigh.
Ugh, walking directly into CPU count caps on my Azure subscription when I’m debugging a resource constraint issue in prod… not my favorite time to be hamstrung by paperwork.
REALLY!? It was an issue of file path lengths? You know what, fine. FINE. I won’t use “long” filepaths for backup files, like anything close to 90 characters.
Really, #SqlServerAnalysisServices? You just couldn’t handle a file path that is barely longer than your default directory path for backups?
Sigh.
@Specialist_Being_677@hachyderm.io @Foritus@toot.dusepo.co.uk This is probably my primary use case for this.
“Willow, how do you debug a massive reporting query that is misbehaving?”
“Data profiling.”
Replying to
@KindlyWizard This is the kind of shit that will wind up with me wrapping all text outputs in a bunch of mostly useless cleanup code just to avoid unexpected issues with invisible characters.