Dmitry Petrov
Dmitry Petrov
Let's work on diff() and data upload separately
Good point. We need to keep information about the order and use it in the following statements when it makes sense. We should use the underlying IDs for that.
It looks like there is some misunderstanding in the ordering. First, let's clarify basic assumptions: 1. Rows are not ordering if the order is not specified. However, **unique** `sys.id` is...
DataChain should become a pipeline. Something like Parameterized Pipeline in DVC but in pythonic way. While DatasetQuery (name is still not good) is a lower level structure for running DB...
@ilongin it looks like legacy logic. at some point it worked this way but not with the latest changes.
yes, that's the problem 🙂
@ilongin looks good. please close this issue once #1020 is closed
@EdwardLi-coder agree, it seems a cleaner API. In general, I like the idea of separating DB/CPU compute from application/GPU compute. Like mutate() and map().
@rlamy we should change it in a way that pre-caching helps.
> this can be implemented inside `udf.run()` It looks like the right way of solving this. Thank you!