Limit allowed values

[IzaakWN]

In some cases, one might want to limit the allowed values of an input variable to some discrete list or range. If the input is not within the predefined range, an error is thrown. This would be an optional key in the input variables, e.g.

  'inputs': [
      {'name': "eta",      'type': "real", range: [-2.5,2.5] },
      {'name': "pt",       'type': "real" },
      {'name': "genmatch", 'type': "int", allow: [0,1,2,3,4,5,6] },
      {'name': "wp",       'type': "string", 'allow': ["Loose","Medium","Tight"]},
  ],

Of course in some cases you might capture these cases with overflow, underflow, or default values elsewhere down the line.

[nsmith-]

I think the nodes encode this information already: if a value is out of range or a key is not in the map then an error is thrown. Each node should probably get a flag to declare alternative handling of out-of-range, like clamp or default.

[IzaakWN]

Hi @nsmith-, could we reopen this issue?

There's a use case in the TauPOG for limiting the values of string keys for systematic variations of the tau ID SF: We would like to decorrelate the statistical uncertainties between pt bins. To keep the file size down, we are using the default field in the Category node to assign the nominal SF to systematic variation that do not affect a given pt bin. As it is now in schema v2, any misspelling of the systematic key by the user would not throw an error.

Implementing an optional field of allowed values in the Variable node (or perhaps in each Category) would like to protect against this at the top of evaluation.