SQL Pivot: Converting Rows to Columns(databricks.com) |
SQL Pivot: Converting Rows to Columns(databricks.com) |
(by "everyday", I mean working with datasets small enough to fit within a spreadsheet's limits, and aren't yet important enough to need the reliability/regularity that comes with importing it to a database)
It's been working surprisingly well so far, with no reported issues
Looks like each DB has a proprietary way of pivoting, and I couldn't find anything on pivot in Calcite. Does anyone know if the SQL impl of pivot was created specifically for Spark, or whether it'd be upstreamed to Calcite? The benefit would be that other projects that rely on Calcite for SQL support would also have pivot capabilities. I'm thinking of Apache Beam.
I wonder if we'll ever get an ANSI-2019 SQL version that tries to merge what various DB vendors have branched off and done. Maybe it already exists? https://www.whoishostingthis.com/resources/ansi-sql-standard...
Edit: I thought we could discuss about dark patterns on HN. I was obviously wrong. Sorry for that...
I usually play safe and reserve side-topic comments to a post-script after something more immediately relevant (and if I don't have something to say on the thread topic, keep them to myself).
The downvotes on this one are going to be because of complaining about downvotes!
In Access, when you enter times, you have to use a specific format, (the datetime group specified by the Windows default) rather than just "23:40" or w/e. I've seen a hack where the hour and minutes are separate boxes, but it wasn't pretty. Just this issue was enough to make Access unappealing.
Access's interface is a bit unintuitive, and given that most people I worked with weren't particularly technically capable, I preferred to present them with the familiar Excel interface (and a few extra buttons), rather than teach them an application they've never used before.
Plus, the existing workflow already used Excel, so I hijacked their workflow and their routine wouldn't be modified much at all (introducing a completely new system would cause a lot of friction and pushback, but one or two extra buttons was fine).
But I understand wanting to not disrupt an existing workflow too much.