msfluid-bot
commented
1 month ago ⯅ @fluid-example/bundle-size-tests: +345 Bytes
| Metric Name | Baseline Size | Compare Size | Size Diff |
|---|---|---|---|
| aqueduct.js | 459.93 KB | 459.99 KB | ⯅ +60 Bytes |
| azureClient.js | 557.86 KB | 557.93 KB | ⯅ +74 Bytes |
| connectionState.js | 680 Bytes | 680 Bytes | ■ No change |
| containerRuntime.js | 260.69 KB | 260.73 KB | ⯅ +39 Bytes |
| fluidFramework.js | 401.54 KB | 401.55 KB | ⯅ +14 Bytes |
| loader.js | 134.19 KB | 134.2 KB | ⯅ +14 Bytes |
| map.js | 42.43 KB | 42.44 KB | ⯅ +7 Bytes |
| matrix.js | 146.58 KB | 146.58 KB | ⯅ +7 Bytes |
| odspClient.js | 525.21 KB | 525.28 KB | ⯅ +74 Bytes |
| odspDriver.js | 97.8 KB | 97.82 KB | ⯅ +21 Bytes |
| odspPrefetchSnapshot.js | 42.76 KB | 42.78 KB | ⯅ +14 Bytes |
| sharedString.js | 163.3 KB | 163.31 KB | ⯅ +7 Bytes |
| sharedTree.js | 392 KB | 392.01 KB | ⯅ +7 Bytes |
| Total Size | 3.3 MB | 3.3 MB | ⯅ +345 Bytes |
Baseline commit: 819cafad04cfc737d6267743dd590715045aa326
Generated by :no_entry_sign: dangerJS against ea554f8796cdf395f7e4d63f38ac47deddec9cd0
Description
Related to AB#8885
DuplicateBatchDetection maintains a datastructure of all batchIds/sequenceNumbers within the collab window (seqNum > latest MSN), and checks all incoming batchIds against that set. However, until Offline Load is enabled, there will be no duplicate batchIds. So until then, don't bother with this tracking.
Notes about performance
For whenever it is enabled
A simple benchmark shows that the time spent doing DuplicateBatchDetection is not much - around 30ms for data structure size of 1000, which is 99.99%ile collab window in the production data I looked at.
To me this seems acceptable - and more importantly, predictable (in terms of perf characteristics, based on the data structures/algorithms used) - so I don't see benefit in instrumenting the code (which would only slow it down). I tried using
SampledTelemetryHelperin this branch (3bf60452ec0bc5a7645bc4f483c6b9b234a3918d) but decided it wasn't worth it (per this analysis)