Range queries on the tai index require clipping alignment blocks. There was a bug where the block clipping would adjust the alignment's column_number field by the number of clipped bases, but not accounting for gaps.
This never came up before because the only tool using the index was taffy view, which ignores column_number completely and just prints out the block sequences and coordinates which are correct (and the tests all use taffy view too).
But it leads to a crash when trying to iterate columns with the python API, ex (thanks Konstantinos):
Then run this python script will fail because block.column_number() returns the wrong value
import pathlib
from taffy.lib import AlignmentReader
from taffy.lib import TafIndex
with AlignmentReader('evolverMammals.taf', taf_index='evolverMammals.taf.tai', sequence_name="Anc0.Anc0refChr0", start=502, length=83) as mp:
for block in mp:
for j in range(block.column_number()):
print(block.get_column(j))
Range queries on the tai index require clipping alignment blocks. There was a bug where the block clipping would adjust the alignment's
column_numberfield by the number of clipped bases, but not accounting for gaps.This never came up before because the only tool using the index was
taffy view, which ignorescolumn_numbercompletely and just prints out the block sequences and coordinates which are correct (and the tests all usetaffy viewtoo).But it leads to a crash when trying to iterate columns with the python API, ex (thanks Konstantinos):
Then run this python script will fail because
block.column_number()returns the wrong value