Ahh, so we would be trading a faster insertion for a slower retrieval of information. I get the idea !
Fl4m3Ph03n1x
peerreynders
Precedent: https://youtu.be/8mXqxBBvNdk?t=1082
Possible problem: Deleting entries (e.g. take) in ETS can be potentially slow (relatively speaking), it’s faster to drop the table and recreate it.
Solution tradeoff: insert into table can fail temporarily. So errors must be caught and insert must be retried later - this can be potentially delegated to a short-lived task process if inserting process cannot afford to be blocked.
- rename table with
rename/2 - recreate new table under common, known name
- do with the renamed table what needs to be done - for example, entire contents can be retrieved with
:ets.tab2list/1,:ets.match_object(tid, {'$0', '$1'}), (don’t know performance relative to:ets.lookup(tid, :failed)). - when done
delete/1renamed table
2