| commit | e0b4b73f0100dfeb6918deb490860f1df2b7f77c | |
| author | Michael Stapelberg <michael@stapelberg.de> | |
| Sun, 14 Dec 2014 22:28:35 +0000 (14 23:28 +0100) | ||
| committer | Michael Stapelberg <michael@stapelberg.de> | |
| Sun, 14 Dec 2014 22:28:35 +0000 (14 23:28 +0100) | ||
| tree | 29d1e5a485f1dc70bd16f7821922698bcbc7b6d7 | treesnapshot (tar.gz zip) |
| parent | 5d46a572a762c2e2ccf8b7b4322cb140c8aa9f0b | commitdiff |
store resultPointers instead of entire results, re-use capnproto buffer
The latter requires the former. But since we only ever use the results
to figure out whether a new result is making it to the top10, storing
resultPointers (which require only a copy of the ranking and the path
hash, i.e. one float32 and one uint64) is fine.
The latter requires the former. But since we only ever use the results
to figure out whether a new result is making it to the top10, storing
resultPointers (which require only a copy of the ranking and the path
hash, i.e. one float32 and one uint64) is fine.
| cmd/dcs-web/querymanager.go | diffblobblamehistory |