Reading attributes from tables in cell_paths[0] fails if the cell wasn't simulated

[ggsun]

Many of our analysis scripts use the first cell in the array cell_paths to read in tables and get table attributes, before using read_stacked_columns to read in entire table columns across multiple cells. If some simulations failed, read_stacked_columns can handle missing tables by setting the ignore_exceptions parameter to True, but if the simulation corresponding to the first cell in the cell_paths array did not run (most likely because a cell from a previous generation failed), reading in a table from cell_paths[0] results in an error. The ordering of the cells in cell_paths seems to be random so it's impossible to predict what seed will be used to read in these attributes.

[rjuenemann]

From my experience, the first entry in cell_paths will be the minimum generation number, with a random seed. For instance, if you aren't filtering by generation when you call cell_paths = ap.get_cells() the first entry should be generation zero from one of the seeds. Even if generation 0 failed partway through the simulation, the tables should still exist.

If you are filtering out generation 0, you might consider doing something like reference_cell = ap.get_cells()[0] without filtering out any generations, then doing cell_paths = ap.get_cells(generation = {whatever} to get the locations of the cells you actually want to extract data from. This way you can use the reference_cell to get the attributes without having to worry about failures.