Marcel Weisgut

Even if we define the dimensions/characteristics that are relevant for the scan executions and their possible options/blocks, we also should come up with a specific coverage strategy. Especially when the...

> > When a `PartialHashIndex` is created using an empty list of chunks, this Impl cannot be created because the data type cannot be extracted from the given chunks. >...

> We have been discussing hash indexes in the context of a master's thesis. Is there a reason to choose the `robin_map` over tessil's `sparse_map`? The author writes (https://tessil.github.io/2016/08/29/benchmark-hopscotch-map.html): >...

> > Yes, but only if the PartialHashIndexImpl also gets a pointer (reference) to the referenced table. Right now, the constructor takes the following arguments: > > const std::vector& chunks_to_index,...

Currently, the descriptions of `PartialHashIndexTest, MemoryConsumption*` test cases are outdated. We should carefully check which memory consumption is expected and why.

I ran the TPC-H benchmark (SF 10, dict. enc.) (1) without index joins, (2) with index joins using chunk indexes, and (3) with index joins using (partial) table indexes both...

> Dropping this here since we had a biref discussion about thread-safe operations on tbb's concurrent maps: In our cache implementation, we decided against tbb due to some restrictions and...

Or is it actually expected behavior?

> This issue is independent of multiple clients but an issue of shuffled, isn't it? Right

### Benchmark Results **System** nemea - click to expand | property | value | | -- | -- | | Hostname | nemea | | CPU | Intel(R) Xeon(R) Platinum...