View operations are extremely slow

[vitawasalreadytaken]

This is the first time I'm using zero_buffer so I'm not sure if this is a regression or if I'm doing something wrong. The following is run on head -n10000 /usr/share/dict/words:

$ time python3 bench_line_reading.py py
real    0m0.353s
user    0m0.334s
sys 0m0.016s

$ time python3 bench_line_reading.py zero_buffer
real    0m9.250s
user    0m9.226s
sys 0m0.019s

Profiling indicates that BufferView.__getitem__ and BufferView._split_char are the biggest culprits.

I tried to replace simple bytes-copying code with zero_buffer in a real-world parser I'm building and performance is also worse by orders of magnitude.

Python 3.4.1
cffi==0.8.6
zero-buffer==0.5.1

[Ivoz]

@ze-phyr-us any chance you could provide a simple example program?

[vitawasalreadytaken]

I rewrote my parser a long time ago but the zero_buffer benchmark seems pretty simple to me.

$ git clone https://github.com/alex/zero_buffer.git && cd zero_buffer
[...]

$ virtualenv -p $(which python3) env && . env/bin/activate
[...]

$ python3 -V
Python 3.4.3

$ pip install .
[...]

$ time python3 bench/bench_line_reading.py py
real    0m8.554s
user    0m8.336s
sys 0m0.189s

$ time python3 bench/bench_line_reading.py zero_buffer
real    5m22.047s
user    5m15.736s
sys 0m1.099s

[vitawasalreadytaken]

$ wc /usr/share/dict/words
  235886  235886 2493109 /usr/share/dict/words