This example uses a GLORIA release 060 SUT already parsed as db. Check db.available_aggregations for your model’s supported sets and levels.

Select a hierarchy → Prepare mapping → Review gaps → Aggregate

1 · Prepare

mapping = db.get_aggregation(
    set="Activity", classification="NACE Rev. 2.1", level="division",
)
frame = mapping["Activity"]

The result is a dictionary of DataFrames, ready for aggregate(). The Aggregation column covers the whole selected native set.

2 · Review

report = frame.attrs["concordance"]
report["unresolved"]

Unresolved items keep their original labels by default. For example, wheat and rice map to division 01, while the packaged Fishing mapping spans divisions 01 and 03 and remains separate at this level.

3 · Apply

aggregated = db.aggregate(io=mapping, levels=list(mapping), inplace=False)
aggregated.get_index("Activity")

You obtain a grouped copy. Preparing a classification mapping does not override unit compatibility: a hybrid model can reject sums across different native units.

Sections, divisions and unresolved assignments

Presets cover NACE Rev. 2.1 and CPA 2.2 hierarchies for mapped variants. level="2 digits" means divisions, not NACE Rev. 2. Use unresolved="raise" when every item must have a unique group.

Missing links, links spanning multiple groups, and overly broad source codes can remain unresolved. A unique group still inherits any partial overlap in the underlying correspondence. See How classification groups are assigned.

Group both sides of a SUT

both = {
    **db.get_aggregation(set="Activity", classification="NACE Rev. 2.1", level="section"),
    **db.get_aggregation(set="Commodity", classification="CPA 2.2", level="section"),
}
reports = {name: frame.attrs["concordance"] for name, frame in both.items()}

Review both reports, then apply db.aggregate(both, levels=list(both), inplace=False).

Regional labels and saving the mapping

Region-specific classifications use a (Region, Item) index containing actual pairs. Keep that index intact. get_aggregation() prepares the whole selected set; it has no item or region filter.

Keep the mapping and its attrs["concordance"] report with your results. CSV and Excel row exports alone do not preserve those attributes. After aggregation, concordance queries see the new labels and cannot reconstruct the original membership.

Interfaces: db.get_aggregation() and db.aggregate(). For custom groups or regions, see Aggregate a database.