I did use the performance branch. And I had a chat with a few people in the automerge community about the performance numbers I was seeing long before I published to see if I was doing anything wrong. I tested a few different versions of automerge but in this test there wasn’t much performance difference between 0.12, 1.0.x-preview versions (which are built from the merged performance branch) and I tried the unreleased automerge-rs. When I ran my tests timing numbers for automerge ranged from about 5m with the old non performance branch down to about 4m20s or so with automerge-rs. Still far from Yjs’s 0.9 seconds.

I just checked and it looks like automerge 1.0.1-preview-4 has landed. I wrote the post benchmarking preview-2. I’ve been knee deep in diamond types lately and haven’t been watching. Fingers crossed there’s some more performance improvements in the pipeline. I’d love to do a follow up in 6 months showing much improved performance.