The comparison is evaluated against the full underlying column instead of the slice’s valid-row mask, so where() can include rows that are outside the sliced view.
Minimal repro: create a table with ids [0, 1, 2, 3, 4, 5], take view = table.row[1:4] so the view contains [1, 2, 3], then run view.where(view.id[0:2] > 1). The expected result is [2], because the predicate only applies to the sliced subcolumn [1, 2], but the actual result includes an extra row from outside that slice. See attached script.
bool-sliced-cols.py
The comparison is evaluated against the full underlying column instead of the slice’s valid-row mask, so where() can include rows that are outside the sliced view.
Minimal repro: create a table with ids [0, 1, 2, 3, 4, 5], take view = table.row[1:4] so the view contains [1, 2, 3], then run view.where(view.id[0:2] > 1). The expected result is [2], because the predicate only applies to the sliced subcolumn [1, 2], but the actual result includes an extra row from outside that slice. See attached script.
bool-sliced-cols.py