Repository navigation
Persistence of index #140
Description
Activity
This is an excellent suggestion. Implementing serialization of the search index could potentially result even a 10x performance improvement during reindexing operations, such as when a user refreshes the page. The serialized index could be stored locally using
localStorageor as a file on the server side for persistence.There's a notable parallel here with how pandas can swiftly load Parquet files, which, despite containing the same data as CSV files, are loaded significantly faster due to their optimized format. I'm curious to see how much of a performance gain we could achieve in this context with JavaScript.
I'd be interested to hear if this aligns with your thoughts or if there's anything you might want to add.
Reacted by roonready, Harpo, Neil Rubens, T Tharanga and Jure Zakotnik@jzakotnik it's done. It's 3-5x faster now. You can benchmark with
npm run benchmark:snapshot. More details in readme. #147Very nice, thanks a lot! Will try it out on our open source library software.
Reacted by MateuszHope it's gonna work for you
Is there a way to persist an existing itemjs index and reload it again for search later? This would be a great enhancement.
If I see it correctly, the index is built via lunr, so the serialisation might also be possible