INNER JOIN can join only two tables and is very slow Joins will be completely refactored and released in the December 2019. It will be faster (we will probably implement grace hash join, if we will not find a faster solution) and it will support joining multiple tables. We are currently working on it.
Multiple aggregation functions when using GROUP BY does not work We will support it in the next release.
There have to be an aggregation function used when using GROUP BY We will support it in the next release.
Casting cannot be done in JOIN clause We are aware of this but we do not plan to add it soon. It is possible to support it but it is difficult to implement in our solution.
GROUP BY fails on large datasets We will refactor GROUP BY so it uses less memory. It will be added in the next release.
Aggregations without GROUP BY do not work with NULL values We will fix this in the next release.
All data has to be in the memory (RAM), if data cannot fit in there, it will fail We will implement loading from disk on demand and release it in the December 2019.
The order of table columns after importing a database from a CSV file may be different from CSV file This would require too much work to fix and we do not consider it as an important issue, so we do not plan to fix this. However, if you consider it as a major issue, let us know and we will reconsider.
There is a huge performance drop when querying datetime columns on NVIDIA Tesla T4 GPU cards in comparison with NVIDIA Tesla V100 GPU cards We are currently working on improvements and we expect them to be done in the November 2019.