Best ideas from H2 2021
Back in July 2021 “Best ideas from the first 25 posts” was posted. Things have been a bit hectic since then, but now we can look back at the best ideas from the second half of 2021. Some of them feel elegant; some of them are somewhat counterintuitive or go against tradition; some of them may change the way you look at certain problems.
In no particular order:
1/ Arguably, there is no data duplication in the double-row implementation of mutual friendship, even though we literally insert the same data twice; we propose “constant-bounded updates vs unbounded updates” distinction;
2/ Anchors, attributes, and links approach to database schema migrations;
3/ Attributes are independent in the logical model, but physical model introduces dependencies that may sometimes be unwanted;
4/ Names of attributes should arguably be globally unique, for easy and reliable grepping;
5/ Sometimes the most efficient way to document some legacy code is to annihilate it;
6/ For some use cases we must store local time and local date, without any timezone information;
7/ Efficient and readable integer-based representation of quarters, months and weeks;
8/ It makes sense to explicitly care about peace of mind during migrations;
9/ Schemaless table designs are intended to solve organizational scalability problem;
10/ All derived datasets are created ad hoc;
11/ People seem to prefer treating database tables as pets while they could and should be treated like cattle.
P.S.: I’m going to organize a Zoom course on how to make sense of typical databases.