问题
Apache Superset has some kind of quirk, where any temporal query is sent with datetimes, even when the respective columns are mere dates.
To work around it, I wanted to add some preprocessing to any passing query, so that datetimes are swapped by dates if the respective columns require it.
Any insight I gained about writing the dialect came by comparing ready-made code, which is often repeatable, as other writers seem to encounter the same problem. I haven't so far found anything tangible to shed light on this particular issue.
Has someone perhaps managed to overcome a similar obstacle, and can point me to the relevant SQLalcemy construct / function that I should implement or add to to allow this preprocessing?
Much appreciation.
来源：https://stackoverflow.com/questions/46027152/which-part-of-sqlalchemys-dialect-is-responsible-for-pre-processing-queries