A significant bug, yet no-one besides me reported it. Either no-one noticed, no-one cared, or they just didn't actually report it somewhere I might notice. Maybe everyone liked returning fewer search results.
So, what was wrong with dclib's search handling?
dclib ignored 1 and 2 character components of search queries because it couldn't handle them properly, without the substring matching. Without the substring matching it couldn't handle longer searches properly either.
Probably, dclib was optimized to handle the typical search, which consists of a single, long, complete keyword. With unfortunate problems for other searches.
This was not fixed properly until 0.3.19. In addition, handling characters other than ascii a-z case-insensitively properly was also added.
This was only fixed by completely removing all the code that did "lookup by keyword" and "keyword indexing" or whatever, and replace it with some other code to handle searches. It didn't have to be particularly fast or clever, it just had to actually return useful search results.
The "remove all the existing code and replace it with something else", something else that works, makes sense, is readable etc. should probably have been taken as a hint for what to do with the rest of dclib.