How to use the bamnostic.bai.conditional_decorator function in bamnostic

    @bai.conditional_decorator(lambda func: lru_cache(maxsize=6)(func), _PY_VERSION[0] == 2)
    def get_ref(self, ref_id=None, idx=False):
        """Iteratively unpacks all the bins, linear intervals, and chunks for a given reference

        A reference is comprised of 2 things: 1) a series of bins that reference chunks of aligned
        reads that are grouped within that bin. 2) a series of virtual offsets of the first read of a
        16384 bp window along the given reference.

        This function also serves to "index" theCSI file such that, if it is invoked by setting
        `ids=True`, will do a single pass through theCSI file and saving the start and stop
        offsets of each of the references. This is used for minimizing the memory footprint of
        storing theCSI in memory. When queried against, the appropriate reference block will be
        loaded. Because of this constant loading, `functools.lru_cache` was applied to cache recently
        used reference blocks to speed up computation. It is assumed that when querying is done, most
        users are looking and just a few references at a time.

        Args:

    @conditional_decorator(lambda func: lru_cache(maxsize=6)(func), _PY_VERSION[0] == 2)
    def get_ref(self, ref_id=None, idx=False):
        """Interatively unpacks all the bins, linear intervals, and chunks for a given reference

        A reference is comprised of 2 things: 1) a series of bins that reference chunks of aligned
        reads that are grouped within that bin. 2) a series of virtual offsets of the first read of a
        16384 bp window along the given reference.

        This function also serves to "index" the BAI file such that, if it is invoked by setting
        `ids=True`, will do a single pass through the BAI file and saving the start and stop
        offsets of each of the references. This is used for minimizing the memory footprint of
        storing the BAI in memory. When queried against, the appropriate reference block will be
        loaded. Because of this constant loading, `functools.lru_cache` was applied to cache recently
        used reference blocks to speed up computation. It is assumed that when querying is done, most
        users are looking and just a few references at a time.

        Args: