Allow for local filters

[martin-ueding]

Our LuescherAnalysis has become so large that running the whole thing becomes a pain. On QBIG we can run it over night, but I want to have a quick edit-run cycle in order to iterate fast. I have changed the data/rho/ensembles.tsv files locally to only include one ensemble. This change must not be checked into git, making it currently cumbersome.

I plan to have some sort of local filters that can be overlaid at various points. One has to take care to ensure correct rebuilding of dependencies.

[martin-ueding]

I still think that you should use filter_paramval for that the portion of the code where the filtering shall be applied. Something like this in foo.R:

pv_load('rho', pc)
pc <- filter_paramval(pc, irrep == 'T1u')

.func <- function (param, value) { … }
foo <- pv_call(.func, pc)
pv_save('rho', foo)

We locally overwrite the pc variable with the filtered version. By changing this code in foo.R the dependencies are clear and the change will trickle down. Yet pc.Rdata will not be generated anew because it is not changed. When one is ready to do the full data set again one should remove this filter.

The dplyr syntax is powerful, if I was to cook up some filter lists in text files or command line switches, it would be unclear how to specify things like irrep %in% c('T1u', 'A1') or total_momentum_d_sq <= 3.