Re^2: Small Hash a Gateway to Large Hash?

[lsherwood]

Rolf-

Your solution is intriguing. We don't see evidence of swapping to disk, but certainly is something to keep in mind.

[LanX]

• No swapping means your hash is kept in RAM.

• Noticeable collisions are very unlikely so accessing this hash can't be done faster.

• You said the hash is only build once and start-up time is no problem, so writing and rehashing doesn't byte you.

• You said you are reading and processing many files... thats the most likely place for optimization.

But without profiling this is only shooting in the dark

Could it be that you are parsing the files and checking if entries match against the hash?

Then trie-optimized regexes could be very performant... (you might need runs with several regexes to avoid buffer overruns, but 10000 short strings should fit)

But w/o much info that's again shooting in the dark.

[BrowserUk]

trie-optimized regexes could be very performant

No, they couldn't. Follow your own advice and profile. The results will correct your mis-impression.