We would want to be able to, for some extracted repository, generate a lookup with functions and accepted arguments. This way, we can use this lookup to compare between two versions, or compare a script (with some subset of those functions) to the signatures. Then we might be able to try and predict if a script will run for a given dependency based on the overlap of the function set (I suspect it should be perfect!)
This is done! The extractor here can generate the files, and the "database" is just the metric file on the filesystem, or using load_metric to load from a Github repository.
We would want to be able to, for some extracted repository, generate a lookup with functions and accepted arguments. This way, we can use this lookup to compare between two versions, or compare a script (with some subset of those functions) to the signatures. Then we might be able to try and predict if a script will run for a given dependency based on the overlap of the function set (I suspect it should be perfect!)