Enhancingperf documentation updates

[smason]

I was going through the enhancingperf document recently and realised that the saying:

Note: Loops like this would be extremely slow in Python, but in Cython looping over NumPy arrays is fast.

doesn't seem to be true any more. Specifically I can implement apply_integrate_f as:

def apply_integrate_pyf(df):
    return np.fromiter((
        integrate_f_typed(*x) for x in zip(df['a'], df['b'], df['N'])
    ), float, len(df))

and get basically the same performance:

• apply_integrate_f takes 1.27 ms
• apply_integrate_f_wrap checks disabled takes 856 µs
• my apply_integrate_pyf version takes 1.13 ms

(all the above run on my computer using Jupyter %timeit, i.e. mean of 7 runs)

This feels like a much nicer way of elding the creation of all those Series objects.

I could submit a pull-request if it seems worthwhile updating this document — git blame says it's mostly been receiving cosmetic changes for 6 years or so.

Output of pd.show_versions()

INSTALLED VERSIONS
------------------
commit: None
python: 3.7.1.final.0
python-bits: 64
OS: Darwin
OS-release: 18.2.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_GB.UTF-8
LOCALE: en_GB.UTF-8

pandas: 0.23.4
pytest: None
pip: 18.1
setuptools: 40.6.3
Cython: 0.29.2
numpy: 1.15.4
scipy: 1.2.0
pyarrow: None
xarray: None
IPython: 7.2.0
sphinx: None
patsy: None
dateutil: 2.7.5
pytz: 2018.9
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 3.0.2
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: 2.10
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[WillAyd]

There are certainly some updates that can be made in the doc, so sure feel free to submit a PR. At the very least the references to .values could be replaced with .to_numpy()

With that said I'm not sure the benchmark you've provided is all that relevant. Cython looping will definitely be much faster than iteration in Python. IIUC your benchmark is only on a very small DataFrame (1,000 records) so differences may not be as apparent there but certainly on larger datasets Cython performance will be much better

[priyankaatgithub]

@WillAyd - I would like to work on this, please. It will be my first contribution to open source.

[WillAyd]

Closed via #26313