* Allow passing read name to mappy
This adds the (optional) ability to pass the read
name to the mappy `map` method. Without the
read name, the call to `map` can sometimes give
different output than the command line version
of `minimap2` because of the way minimap uses
the hash of the read name to break ties in ordering
hits. This can affect which / if certain
supplementary alignments are generated, and even
which / if non-primary alignments are generated.
* Pass name directly to mm_map_aux
Get rid of additional function, and always
accept the name parameter in the mm_map_aux
function (can be nullptr if not available).
---------
Co-authored-by: Rob Patro <rob@newton>
The Minimap2 behavior was found to handle sequences with large
deletions differently when upgraded from v2.17 to v2.26, causing
potential issues in projects mapping extensive deletions of ~1200 base
pairs. The originally suggested solution of setting `-r 500,500` was
observed to be partially non-applicable since the Python Wrapper,
`mappy`, only allowed manipulation of parameter `bw`.
In response to issue #1111, where this was originally reported,
this commit introduces a modification in the Python wrapper,
`mappy`. Until now, `mappy` only allowed manipulation of the `bw`
parameter, preventing the suggested fix of setting `-r 500,500`.
This commit introduces a modification in the Python wrapper to include
the `bw_long` option in the `Aligner` class. Consequently, both
parameters `bw` and `bw_long` can be manipulated, thereby allowing the
desired Minimap2 behavior encountered in version 2.17. As a result,
this patch ensures consistent handling of sequences containing large
deletions irrespective of the version upgrade."
Closes#1111
mappy creates a CIGAR string by default (`-c` flag) and so is
incompatible with indexes that are created using the `--idx-no-seq`
flag.
The previous implementation of mappy did not check for the `MM_I_NO_SEQ`
flag and would seg fault when attempting to map a read or retrieve a
reference sequence from the index. This patch adds a check to both
`mappy.Aligner.seq` and `mappy.Aligner.map` and returns `None` if there
is no index sequences. I've chose `None` as this is inline with the
behaviour of mappy for reads that do not align/retreiving sequences that
aren't in the index, however it might be better to raise an exception so
that this error is distinct and can be communicated to the caller.
Define MM_CIGAR_STR to the full string of CIGAR operators (including
the 'B' operator as well) and use it throughout the C code.
It would be possible to use it from the Cython code too, but it's easier
to keep that as a Cython string literal to avoid adding extra runtime
code to handle locale conversion.
