Mazdak Farrokhzad
Mazdak Farrokhzad
Rather than having each row serialize to its own little `Vec` of BSATN, fuse all rows into a single `Vec` allocation. https://github.com/clockworklabs/SpacetimeDB/pull/1200 is an experiment that demonstrates what this issue...
A recent profile of a large scale workload showed that brotli compression accounts for nearly 15% of total work This is due to the fact that we compress all outgoing...
We want to find the failing test case in https://github.com/clockworklabs/SpacetimeDB/actions/runs/8841180124/job/24277859858?pr=1107.
See https://github.com/clockworklabs/SpacetimeDB/pull/869/files#r1497940875.
See title.
Like how we extracted out the sources with `SourceSet`, we should extract e.g., `42` in a query `SELECT * FROM bar WHERE bar.id = 42` and instead insert a symbolic...
It's showing up in flamegraphs taking up most of that call...
Basically use `RowRef`s instead of `ProductValue`s.
Many cryptographic hashes result in 32 byte (256 bit) outputs, and storing this as a `Box` is about as cheap as we can make it. This new variant/type would be...
Basically, the idea is that we go back to only storing `TableId`s in metrics and have the UX later translate the ids to table names, so that the BitCraft team...