It now works on a snapshot stored in tests/test-data/test_id_duplicates.py. Expected duplicates are determined manually.
Note that regenerating this file may result in tests failing, as it references ids explicitly.
The tests currently pass.
I have intentionally commented out a test I think should pass, but does not (the same one which did not succeed in the previous version of this test file).
There are now meaningful tests which do pass.
In general, this isn't really testing code correctness, so much as clustering accuracy. We might want to think about presenting it differently, but I don't think 100% matching "ground truth" is the expected outcome. Rather, we want to minimize mistakes.
I updated
tests/test_id_duplicates.py
.It now works on a snapshot stored in
tests/test-data/test_id_duplicates.py
. Expected duplicates are determined manually. Note that regenerating this file may result in tests failing, as it references ids explicitly.The tests currently pass. I have intentionally commented out a test I think should pass, but does not (the same one which did not succeed in the previous version of this test file). There are now meaningful tests which do pass. In general, this isn't really testing code correctness, so much as clustering accuracy. We might want to think about presenting it differently, but I don't think 100% matching "ground truth" is the expected outcome. Rather, we want to minimize mistakes.