3 ms·
I like this. I've often had a similar issue in computational physics when handling lots of different calculation input/output files with variations in paramete
by tfgg 12y ago
I like this. I've often had a similar issue in computational physics when handling lots of different calculation input/output files with variations in parameters. I've tended to automatically generate directory hierarchies, e.g.
rc=1.0/ecut=60/kp=4,4,4/
rc=1.0/ecut=80/kp=4,4,4/
rc=2.0/ecut=60/kp=4,4,4/
rc=2.0/ecut=80/kp=4,4,4/
but some standard way of expressing it in the file name would be great, and nicer to parse out later. There are filesystems that allow metadata, but I've never seen one really being used for that purpose. You could make some associated command line tools that are equivalent to 'ls' that allow splitting/slicing the files down different parameters.
- yummyfajitas 12y agoTake a look at hdf5.
- tfgg 12y agoThanks, I'll take a look, but it seems like I'd have to change the format of the files, right? Unfortunately that's not really feasible, since their format is determined by the scientific codes I use -- that's why I was attracted to the idea of using the file name. I can, however, have a go at HDF5 for code that I write myself.