We are not using correspondentia anymore
Python library to map correspondence tables in different formats to data structures.
A quick example:
from correspondentia import match_fields
numbers_to_names = {
1: [{"value": "one", "type": "exact"}],
2: [{"value": "two", "weight": 0.5, "type": "disaggregation"},
{"value": "deux", "weight": 0.5, "type": "disaggregation"}],
}
my_data = [{
'count': 1,
'name': 'foo'
}, {
'count': 2,
'name': 'bar'
}]
list(match_fields(my_data, numbers_to_names, "count"))
> [{'count': 'one', 'name': 'foo'},
{'count': 'two', 'name': 'bar', 'correspondentia_allocation': 0.5},
{'count': 'deux', 'name': 'bar', 'correspondentia_allocation': 0.5}]
match_fields
return a generator.
Input data should be an iterable of objects supporting the dictionary interface.
correspondentia
currently can import the following formats:
- CSVs following the simple schema
We plan to also eventually support the following:
- RDF (Turtle) correspondence tables following the BONSAI spec
- CSVs with BONSAI ontology predicates
You can also write custom importers, or define correspondence tables manually. In either case, the correspondence table data should include at least the following fields (additional fields are also allowed):
{
"label in origin schema (usually str, but can be int or float)": {
"value": "label in destination schema (usually str, but can be int or float)",
"type": one of ["exact", "disaggregation"],
"weight": float, # optional
}
}
A CSV with two required and one optional columns.
- First column: Label in origin schema
- Second column: Label in destination schema
- Third column (optional): Weight used for disaggregation.
If matching is 1-N or N-1, just use multiple rows with redundant labels.
CSVs should follow the Open Knowledge CSV spec. Do not use column headers.
Installation via normal pathways; currently has no dependencies.
Follow standard fork/pull-request procedure.