Markus Dreseler

Note: As of now, you will not see any issues with this unless you plan to use `set_default_resource`.

Regarding the CardinalityEstimator: Yes, I think that the assertion is too strict. If you print `output_table_statistics->row_count`, I'd expect it to be less than one. We should probably accept cardinalities `<...

Hmm, I would have thought that it is close to zero. How can cardinalities be negative? That sounds like there is something wrong there.

We can probably reuse code from MetaSystemUtilizationTable or add MaxRSS over there and simply drag it into the output JSON.

I guess it is because the `PredicateSplitUpRule` has not yet split up disjunctions at the time where the composition rule is executed.

Should be safe to move the BetweenCompositionRule behind the PredicateSplitUpRule as PredicatePlacementRule is executed a second time, anyway.

Wow. I never dug that deep into the iterator requirements. It will be interesting to see which components break and, more importantly, if this has any performance impact outside of...

I believe this was introduced with 0bf7d52efb7cbfc2e6b18f373bbf2dbb5b87a30a: ``` -class BaseSegmentIterator : public boost::iterator_facade, +class BaseSegmentIterator : public boost::iterator_facade, ``` I'll add a test in a second that uses `std::is_heap`, which...

According to @Bouncner, this is still the case. Here is a way to test it (using gcc trunk): https://gcc.godbolt.org/z/TzjVhB For any given iterator, if `check(iter)` compiles, you should have a...

Here is an example on how to use ranges to create random access compatible iterators that do not return a true reference: https://ericniebler.github.io/range-v3/ (Section: Create Custom Iterators with basic_iterator)