Dan Luu

I believe the problem here is that the Recycler is never invoked. CreatePrimeFactorsIndex returns an ISimpleIndex, which should have a Recycler. But we never expire documents, which means that they...

Because we have no reasonable way to conditionally add/remove rows from a plan at query time, it may make sense to have an `if` at the top level of the...

I'm going to check in a fix that speed up the time in the loop that's not related to function calls by 3x to 4x. However, all this means is...

It looks like the mapping for private terms/rows is somehow getting confused? If we shrink down this test case to a single document with two terms, we still get a...

AbstractRowEnumerator has some logic for this, but it appears that the scaffolding around it isn't correct. Additionally, `Show::Execute()` directly calls the `RowIdSequence` constructor on a `Term` that's created without any...

Here are some disconnected notes on the old code: ShardedDocFreqTable is used by ~~~ ShardStatisticsStage TermTableStage ExplicitEntries (part of TermTableBuilder) QueryPipeline ~~~ In QueryPipeline, ShardedDocFreqTable is only used in ValidatingResultsCallback,...

We should probably restore/re-create IngestorTest before we do this. I'd say that there's some chance I'll get to this before StrangeLoop, but that it's relatively unlikely. I'll give it a...

The test referred to above now exists. I should probably do this soon.

One example of this is ~~~ #0 0x00007ffff7ae28bd in __cxa_throw () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6 #1 0x00000000004a9792 in BitFunnel::BlockAllocator::AllocateBlock() () #2 0x0000000000471b45 in BitFunnel::Slice::Slice(BitFunnel::Shard&) () #3 0x000000000046b679 in BitFunnel::Shard::CreateNewActiveSlice() () #4 0x000000000046b5bc...

From looking at htop while running "case 2", we don't appear to be leaking threads. We do use more memory as we continue to run longer, though. It seems plausible...