Hacker News Reader

The atomicity in the bash version can be "simply" achieved by copying the file to a temporary, modyfing it, then using `sync; mv; sync`, right?

View on HN

Also, while you're copying you can inverse grep filter to avoid duplicates.

And since you're copying, you could maybe ensure sorting, but I don't think doing that with "bash" (plus off the shelf utils) makes a lot of sense.

That's what DJBs CDB (cdbget, cdbmake, etc) is for: https://cr.yp.to/cdb.html

Oh you're right I'll add that in later.

I would go : 1. count records 2. make copy 3. insert 4. ensure records = records + 1 5 if records != records +1 or file can be grepped, stat, filesize, then assume is corrupted and rollback to copy

Use `look` to get `O(log(n))` lookups (writes are still slow, but you could use a multi-level chain I guess). `join` does not use binary search even though it could have.

no waaay, I have never seen look before, thanks!

Reply on news.ycombinator.com