Métriques du dépôt
- Stars
- (7 000 stars)
- Métriques de merge PR
- (Merge moyen 1j 5h) (9 PRs mergées en 30 j)
Description
Howdy. I'm not sure this really merits a Issues or if it would be better as a conversation some place. In using sqlite3 for a project and I noticed that it's quite a bit slower to select a large number of rows as compared to doing the same via their CLI client. ~4-5x slower. A little profiling shows that the time is mostly spent in Next and Scan. For small row sets it's not a big deal but I'm testing up to ~500K rows.
I suppose my starting question is how much overhead is expected going back and forth between Go and the C API? If it's expected to be fairly fast/low-overhead, any suspicions on where I might start looking?
Here's a sketch of approximately what the code is doing:
https://gist.github.com/mhat/ded97fe58922b4af9d03f375b465c339
There are two variations, one uses the new ScanTypes while the other is cheating with a list of string pointers. Neither appreciably moved the needly forScan.