E-IG0005 — Unbounded dictionary keys in random generation

Error Type	Instance Generation Error
Phase	generating random instance data
Python exception	`jijmodeling.ModelingError`

Message

Cannot generate random data for a dictionary whose key type has unboundedly many values, such as `int` or `str`.

Possible fix: do one of the following:
  - use a key type with finitely many values, such as a natural-number expression or a category label
  - for `str` keys, list the keys to generate with the `keys` option of this placeholder
  - prepare the instance data for this placeholder yourself

Cause

A partial dictionary placeholder, or a placeholder whose elements are partial dictionaries, has a key type with unboundedly many values, such as int or str. To generate a dictionary, the generator picks its keys from the list of all allowed keys, and no such list exists for these key types. For str keys, the keys option of the placeholder supplies that list instead.

Fix

Declare the keys with a type that has finitely many values, such as a natural-number expression N (the keys 0 to N - 1) or a category label:

import jijmodeling as jm

problem = jm.Problem("example")
N = problem.Natural("N")
weights = problem.Float("weights", dict_keys=N, partial_dict=True)
data = problem.generate_random_dataset(options={"N": {"value": range(3, 6)}})

For str keys, list the keys to generate with the keys option:

problem = jm.Problem("example")
prices = problem.Float("prices", dict_keys=str, partial_dict=True)
data = problem.generate_random_dataset(options={"prices": {"keys": ["apple", "pear"]}})

Otherwise, prepare the instance data for this placeholder yourself instead of generating it.