I have a points dataset in the following format (x, y, value), is it possible to get aggregated dataset using Polars native (maybe even lazy) code as much as possible? Basically I want to create a virtual grid and then sum all the points in respective grid cells, x_step and y_step being grid cell dimensions. The output would be a dataset with columns (cell_x, cell_y, agg_value) while cell_x and cell_y columns only taking values divisible by my predefined steps (representing bottom left coordinate of the grid cell), and agg_value being a sum of all point values inside the cell: (cell_x >= x and x < cell_x + x_step) and (cell_y >= y and y < cell_y + y_step). Currently I am iterating from start, incrementing my variable cell_x by x_step and the same for Y axis in a nested loop. Then I call sum on the filtered subset (cell) of points and I output one row to the output cell. It is rather slow in Python. Here is a visual example, all points have value of 1 for simplicity: Continue reading...
