1edd1496383e7ae0374ec5dd8b455d58077e1d62,cnvlib/_merge.py,,_merge_overlapping,#Any#Any#,41

Before Change

def _merge_overlapping(table, combiners):
    Merge overlapping regions within a group.
    // Short-circuit the simple, common cases
    if len(table) == 1:
        return table
    if table["start"].nunique() == table["end"].nunique() == 1:
        return _squash_rows(table, combiners)
    // Identify & enumerate (non)overlapping groups of rows
    overlap_sizes = table.end.cummax().values[:-1] - table.start.values[1:]
    non_overlapping = np.r_[False, (overlap_sizes <= 0)]
    // Squash rows within each non-overlapping group
    return (table.groupby(non_overlapping * np.arange(len(non_overlapping)),
                          as_index=False, group_keys=False, sort=False)

After Change

    // advantage of the grouping and sorting already done, and don"t repeat
    // pandas" traversal and inferences.
    // ENH: Find & use a lower-level, 1-pass pandas function
    keyed_groups = itertools.izip(_nonoverlapping_groups(table),
                                  table.itertuples(index=False))
    merged_rows = [_squash_tuples(row_group, combiners)
                   for _key, row_group in itertools.groupby(keyed_groups,
                                                            lambda x: x[0])]
    return pd.DataFrame.from_records(merged_rows, columns=merged_rows[0]._fields)


def _nonoverlapping_groups(table):

Instances