Sometimes you'd like to get around the restrictions of designing a data grid from xPath, while keeping the advantage of letting Mendix generate filters. Also, generating reports with soem specific requirements for granularity / aggregations would be much easier and more performant using OQL with common SQL operators such as partition by, cursors, distinct, etc.
OQL is a very useful query tool next to xpath, and we plan to improve on it in the future. First steps taken last year is to provide java apis to use SQL and OQL in java actions, but we also plan to add more support in the Modeler.
I agree with this opinion. Without SQL support, it is an architectural defect of Mendix infrastructure.
Suppose that Database server and Application server are separate (eg. On-Premise system). The current Mendix system tries to do everything on Application server but does not take account of Database server's performance. As a consequence, it does not care about each performance balance of the servers.
This IS a critical point for Japanese developers to make a decision to buy Mendix.
Thus, the current Mendix is only for a very small system like an experimental desktop-server.