I've fixed the _boxed formatting to strip ANSI codes when calculating string lengths (so you can use ANSI colours properly)

changed the colours for the "failed systems" warning so it's absolutely obvious

wrapped the JSON ensemble output in a write_ensemble() function which automatically gzips or bzips depending on the filename extension (or this can be explictly set in grid_options['compress_ensemble'] ).

make gridcode filename blue

cleaned up some of the verbosity to be more like binary_grid and less annoying

implemented binary_grid-style statistics over last n_logging_stats verbose outputs (50 by default)

added ANSI colour to output if in a terminal (i.e. with a tty)

fixed the cmd-line parser to work in the form x=y as intended (downside: no help output, but that just said how to use the old parser which doesn't exist now)

updated the internal API for how grids are constructed to simplify the code

add option such that when resolution==0, the resolution is calculated automatically from the sample list.

add delete_grid_variable function

In my CEMP simulations, I use the above with something like this to zoom in around 0.8Msun in M1:

emp_population.update_grid_variable(
    name="lnm1",
    spacingfunc="gaussian_zoom(np.log(0.4),np.log(6.0),np.log(0.8),0.05,0.99,100)",
    resolution=0
)

shows:
current grid status (#/# and modulo)
estimate of time remaining and time to finish
memory use (of all processes)
time per run

Integrated the new functions, population_id will now be renewed upon running a grid, prevented evolve_single from running if there are no arguments passed, added some check in the export_all_info function that raises an error if the user tries to export without outfile OR data_dir

add methods to grid.py:

rename_grid_variable
update_grid_variable
clean

add functions to functions.py:

convfloat
datalinedict
pad_output_distribution