Currently, it appears that if the source of two cells is not at least 70% the same, they will always be treated as separate cells and represented by a delete-cell, then add-cell operation. It feels like this threshold should be much lower, at least if whitespace-only lines are ignored.
There appears to be a relevant TODO:
|
# TODO: Add configuration framework |
|
# TODO: Tune threshold with realistic sources |
Currently, it appears that if the source of two cells is not at least 70% the same, they will always be treated as separate cells and represented by a delete-cell, then add-cell operation. It feels like this threshold should be much lower, at least if whitespace-only lines are ignored.
There appears to be a relevant TODO:
nbdime/nbdime/diffing/generic.py
Lines 30 to 31 in 458beeb