Refactor web application and other infra fixes#83
Refactor web application and other infra fixes#83MBkkt wants to merge 1 commit intoquickwit-oss:masterfrom
Conversation
- Removed unused testing dependencies from package.json. - Added a new Python script (serve.py) for serving JSON files with gzip compression. - Updated index.js to use initialMode prop for Benchmark component. - Deleted setupTests.js as it is no longer needed. - Renamed style.scss to style.css and added new styles for better UI.
MBkkt
changed the title
| query_idx[query.query] = {"count": -1, "duration": []} | ||
| elif query_idx[query.query]["count"] != -1: |
There was a problem hiding this comment.
here was a bug for engines which doesn't support some queries
| @echo "--- Serving results ---" | ||
| @cp results.json web/build/results.json | ||
| @cd web/build && python3 -m http.server $(PORT) | ||
| @cd web/build && python3 ../../web/serve.py $(PORT) |
There was a problem hiding this comment.
just to serve gzip file instead of full large json (I sometimes had slow internet connection between me and the server, so it was fix suggest by claude and it actually allows to download less data (500kb vs 3.5 MB) when open this web view).
There was a problem hiding this comment.
I tried to build it as is, when I fixed bug with COUNT collection type is required, but it didn't build.
So I was required to update it, our FE engineer gave me some recommendations how to do it.
(main was rewrite this entirety but I wasn't interested :)
I will ask him is it possible to extract our webview (https://serenedb.com/search-benchmark-game)
I hope so, because it's way more convenient, because 2 things:
- It has "heatmap" with all query and collection types
- It has sort by column feature (Because it's really more useful to optimize slow queries over fast/random)
| @@ -0,0 +1,115 @@ | |||
| {"query": "the", "tags": ["term"]} | |||
There was a problem hiding this comment.
We just was interested in performance of single wand for non high term.
In original (your) queries there's only single term query ("the"), so we made a separate file to measure this and don't affect benchmarks numbers
2 participants
COUNTcollection type in UI, it was fixedAlso