Use zip_longest for more readable and faster code in the paired end iterator

[rhpvorderman]

Sorry for the PR spam. I just suddenly find so much stuff to optimize in this library. And even more is coming, but that depends on still pending PRs #14 and #20. (I want to make writing faster still, by avoiding creating new bytes objects).

This is an easy one to review. I know you are busy!

Ideally we would want to use zip(it1, it2, strict=True), but that is only available from python 3.10 and higher. (3.9 on my Debian 11 machine).

Since try ... except has some expense I tried using next differently. It has a default argument. So you can use next(it1, None) and it will yield None if the iterator is exhausted. This leads to a cleaner looking code when checking.

I had a look at itertools if there was something in the standard library, and yes there was: zip_longest. Keeps on yielding until the longest iterator is exhausted and yields None for the shortest iterator.

The code looks cleaner now now all the exhaustion checking is in zip_longest instead of in try except loops. As an added bonus, itertools is fully implemented in C so it is faster as well:

before:

Benchmark #1: python dnaio_read_paired.py ~/test/big2.fastq
  Time (mean ± σ):      4.884 s ±  0.031 s    [User: 4.667 s, System: 0.216 s]
  Range (min … max):    4.826 s …  4.919 s    10 runs

after

Benchmark #1: python dnaio_read_paired.py ~/test/big2.fastq
  Time (mean ± σ):      4.746 s ±  0.049 s    [User: 4.540 s, System: 0.206 s]
  Range (min … max):    4.691 s …  4.835 s    10 runs

It is 2-3% faster. Not bad for such a small change.

[marcelm]

Good improvement, thanks once again! I think the comment at the top of __iter__ is a bit verbose, but let’s leave it as it is for now.