[pypy-issue] Issue #2261: PyPy 5.0.0 very slow when using BioPython BLAST XML parsing under OS X (pypy/pypy)

[Terry Jones]

New issue 2261: PyPy 5.0.0 very slow when using BioPython BLAST XML parsing under OS X
https://bitbucket.org/pypy/pypy/issues/2261/pypy-500-very-slow-when-using-biopython

Terry Jones:

I just upgrade from pypy4.0.1 to 5.0.0 (OS X 10.10.5, via brew) and was surprised to find some code running much slower. I have a simple piece of code that uses BioPython (which in turrn uses expat to parse some XML). On python 2.7 my code runs in ~9 seconds, in python 3.5 it’s ~10 secs, in pypy 4 it’s 18 seconds, and in pypy 5 it’s 25 seconds.

You can reproduce it by installing BioPython (I have 1.64 in my virtualenv) and running the following minimal example:

```
#!python

from __future__ import print_function

import sys
from Bio.Blast import NCBIXML

count = 0
with open(sys.argv[1]) as fp:
    for record in NCBIXML.parse(fp):
        count += 1

print(count)
```

with the attached XML (once uncompressed) file given on the command line.