#fractionalindexing — Public Fediverse posts

Continuing on my series on fractional indexing, I come to the conclusion that it is a bit of a scam. You have to be very careful with the usage patterns of the indexed list to be sure the index key length does not get out of hand. While a nice concept, it somewhat reminds me of "two digits for the year are enough" or "a 32bit signed integer" is enough to store a time stamp.

More rant details in my blog: https://blog.miamao.de/blog/2026-08/15.Notes_on_Fractional_Indexing.html

#fractionalIndexing #dataStructures #algorithm #cs

Meanwhile found the technical term: fractional indexing.

Discussed here: https://observablehq.com/@dgreensp/implementing-fractional-indexing

My (luxury) problem: it is not elegant. It wriggles through a maze of (special) cases.

With arbitrary precision numbers it would be nearly trivial:
1. append to list: last key + 1
2. prepend to list: first key -1
3. insert between a and b: (aKey+bKey)/2

Only problem: 3. generates more significant bits than needed.

Hints and boosts appreciated.

#fractionalIndexing #algorithm #dataStructures