Markus Dreseler

Results 35 issues of Markus Dreseler

During the work on my thesis, I noticed that boost_default_memory_resource.cpp inadvertently breaks the interface. We introduced it a while ago either because of issues with a boost version or because...

That code is part of a lambda. We return from the lambda, not from the entire method. I guess it is partially my fault for just wrapping the lambda around...

Refactor

@lawben contacted me with some questions about our decisions for a sort algorithm. My (wrong) gut feeling was that we would spend more time in the materialization phase than in...

Independent of the mode, warmup currently runs each query for *n* seconds. This leads to fast queries being executed more often than slow queries. Warmup should mimic the actual benchmarking...

``` (debug)> generate_tpch .01 ... (debug)> visualize SELECT * FROM lineitem WHERE l_o_id * l_i_id = 0 libc++abi.dylib: terminating with uncaught exception of type opossum::InvalidInputException: Invalid input error: Couldn't resolve...

Bug

For a condition like `x = (SELECT ...)`, the select query has to return one result and one result only. For unflattened subqueries, the ExpressionEvaluator makes sure that this is...

Bug

In yesterday's Fortnightly, we had a discussion about why it is important to be conservative when caching results of the CardinalityEstimator. Across calls of `estimate_cardinality`*, we have to assume that...

Optimizer

``` /** * Returns the time in ns since epoch. */ uint64_t MetaSystemUtilizationTable::_get_total_time() { auto time = std::chrono::steady_clock::now().time_since_epoch(); return std::chrono::duration_cast(time).count(); } ``` I don't think steady_clock has a usable epoch,...

I found some places where we have code in the hpp file even though it is not part of a template. Two things: * @mjendruk - Is there any reason...

Refactor

Currently, no execution times are printed for subqueries. When I call visualize on a PQP with a subquery, it should show me the number of executions and the avg. and...

Feature