···11# repo-stream
2233-an AsyncRead for atproto MSTs in CAR files
33+a futures atproto record stream from CAR file
44+55+current notes
66+77+- just buffering all the blocks is 2.5x faster than interleaving optimistic walking
88+ - at least, this is true on huge CARs with the current (stream-unfriendly) pds export behaviour
99+1010+- transform function is a little tricky because we can't *know* if a block is a record or a node until we actually walk the tree to it (after they're all buffered in memory anyway).
1111+ - still might as well benchmark a test with optimistic block probing+transform on the way in
1212+1313+1414+original ideas:
415516- tries to walk and emit the MST *while streaming in the CAR*
617- drops intermediate mst blocks after reading to reduce total memory