UnboundLocalError when req_clust exactly matches a partition size

[saveli]

Bug

FINCH() raises UnboundLocalError: cannot access local variable 'requested_c' where it is not associated with a value when req_clust is not None and the requested number of clusters is already exactly present in num_clust.

Root cause

In finch.py (lines 363–373 in v0.2.2), the req_clust not in num_clust branch only handles the mismatch cases (too large, or needs sub-clustering). When req_clust is an exact match, the entire inner block is skipped and requested_c is never assigned before the return statement.

if req_clust is not None:
    if req_clust not in num_clust:          # False when exact match → block skipped
        if req_clust > num_clust[0]:
            requested_c = c[:, 0]
        else:
            requested_c = req_numclust(...)
    # no else → requested_c undefined when req_clust in num_clust
else:
    requested_c = None

return c, num_clust, requested_c           # UnboundLocalError

Fix

Add an else branch to handle the exact-match case:

if req_clust is not None:
    if req_clust not in num_clust:
        if req_clust > num_clust[0]:
            requested_c = c[:, 0]
        else:
            requested_c = req_numclust(...)
    else:
        requested_c = c[:, num_clust.index(req_clust)]
else:
    requested_c = None

Reproduction

Any dataset where FINCH produces a partition with exactly req_clust clusters will trigger this. Confirmed on v0.2.2 (latest).

[ssarfraz]

Hi @saveli , thanks for catching this, this line got deleted in the last update. have updated it.