I did a quick comparison of some storage methods:
| method | read-time (s) | write-time (s) | size (MB) | size-gzipped (MB) |
|---|---|---|---|---|
| CSV | 0.3 | 6 | 43 | 7.6 |
| SQLite | 36 | 26 | 196 | 14.9 |
| JDF | 0.08 | 0.13 | 15.12 | 10 |
| Feather | 0.002 | 0.79 | 64 | 7.7 |
Source code:
I have not done any optimizations for the storage methods, only used the most simple interface for each.
Edit: the Sqlite timings look too bad for me. Maybe I am missing something obvious?