The `calculate_slope` function should return -np.inf for the slope

[RGarcia-1]

In the current implementation of calculate_slope()
https://github.com/fema-ffrd/overflow/blob/main/src/overflow/_flow_direction/flow_direction.py#L93C1-L94C22

    if dem[row + dy, col + dx] == nodata_value or np.isnan(dem[row + dy, col + dx]):
        return np.inf

This value is used inside flow_direction_for_tile():

                    slope = calculate_slope(dem, row, col, dy, dx, nodata_value)
                    if slope > max_slope:
                        max_slope = slope
                        max_index = i
                    if slope > 0:
                        all_non_positive = False

Because max_slope is initialised to -np.inf:

max_slope = -np.inf

Implies that any np.inf returned from calculate_slope() always becomes the maximum slope. Just checking that the calculate_slope should return -np.inf?

[barne856]

Hey thanks for raising this! To answer your question, yes, the np.inf return is an intentional design choice. This forces edge cells to drain outward from the raster domain. That said, this behavior isn't universal, and some users might prefer to treat NoData differently (e.g., ignore those neighbors unless it is the only possible direction). We haven't put a lot of effort in to making things like this configurable quite yet, but it is a reasonable enhancement and shouldn't be difficult to support. There could be an option added for force_flow similar to ESRI's spatial analyst tools https://pro.arcgis.com/en/pro-app/3.4/tool-reference/spatial-analyst/flow-direction.htm

If you're interested in adding that option, I'd welcome a PR!