A minimal fasta/q parser in python using kseq.h

fastx.py

/* The MIT License

   Permission is hereby granted, free of charge, to any person obtaining
   a copy of this software and associated documentation files (the
   "Software"), to deal in the Software without restriction, including
   without limitation the rights to use, copy, modify, merge, publish,
   distribute, sublicense, and/or sell copies of the Software, and to
   permit persons to whom the Software is furnished to do so, subject to
   the following conditions:

   The above copyright notice and this permission notice shall be
   included in all copies or substantial portions of the Software.

   THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
   EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
   MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
   NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
   BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
   ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
   CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
   SOFTWARE.
*/

from cffi import FFI

def build():
    ffibuilder = FFI()
    ffibuilder.set_source(
        "_fastx",
        r"""
        #include <zlib.h>
        #include <stdio.h>
        #include "kseq.h"

        KSEQ_INIT(gzFile, gzread)

        kseq_t *open_fastx (const char *fn) {
            gzFile fp = gzopen(fn, "r");
            return kseq_init(fp);
        }

        void close_fastx(kseq_t *kseq) {
            gzclose(kseq->f->f);
            kseq_destroy(kseq);
        }

        """,
        libraries=['z']
    )

    ffibuilder.cdef(
        r"""
        typedef struct kstring_t {
            size_t l, m; char *s;
        } kstring_t;

        typedef struct kseq_t {
            kstring_t name, comment, seq, qual;
            ...;
        } kseq_t;

        kseq_t *open_fastx(const char *fn);
        void close_fastx(kseq_t *kseq);
        int kseq_read(kseq_t *kseq);
        """
    )
    ffibuilder.compile(verbose=True)

    
def run():
    import _fastx
    import sys

    class Fastx:

        def __init__(self, fname):
            self.kseq = _fastx.lib.open_fastx(fname.encode())

        def __del__(self):
            _fastx.lib.close_fastx(self.kseq)

        def __next__(self):
            if _fastx.lib.kseq_read(self.kseq) >= 0:
                return (
                    _fastx.ffi.string(self.kseq.name.s),
                    _fastx.ffi.string(self.kseq.seq.s),
                    _fastx.ffi.string(self.kseq.qual.s) if self.kseq.qual.l > 0 else "")
            else:
                raise StopIteration()

        def __iter__(self):
            return self


    n, slen, qlen = 0, 0, 0
    for name, seq, qual in Fastx(sys.argv[1]):
        n += 1
        slen += len(seq)
        qlen += qual and len(qual) or 0
    print('{}\t{}\t{}'.format(n, slen, qlen))


if __name__ == "__main__":
    import sys
    if len(sys.argv) < 2:
        build()
    else:
        run()

This was created because I found the results for fastq parsing with pyfastx presented here to be a little odd: since pyfastx uses kseq.h one would expect near the baseline C speed (which also uses kseq.h).

If we line profile the code then the original program spends as much time adding python objects together as it does in the pyfastx package:

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
    28                                           @profile
    29                                           def main():
    30         1         21.0     21.0      0.0      n, slen, qlen = 0, 0, 0
    31   5682011    7366975.0      1.3     47.9      for name, seq, qual in pyfastx.Fastq(sys.argv[1], build_index=False):
    32   5682010    2324629.0      0.4     15.1          n += 1
    33   5682010    2783887.0      0.5     18.1          slen += len(seq)
    34   5682010    2909447.0      0.5     18.9          qlen += qual and len(qual) or 0
    35         1        158.0    158.0      0.0      print('{}\t{}\t{}'.format(n, slen, qlen))

The code in this gist works under both CPython and PyPy, unlike pyfastx which is strictly a CPython extension. When using PyPy, the difference between the Python and C implementation is narrowed dramatically:

5682010	568201000	568201000

real	0m11.444s
user	0m10.944s
sys	0m0.284s
Running pypy
5682010	568201000	568201000

real	0m1.973s
user	0m1.555s
sys	0m0.258s
Running C
5682010	568201000	568201000

real	0m1.764s
user	0m1.508s
sys	0m0.217s