requests: fine tune timetout                                         2026-10-14
-------------------------------------------------------------------------------

requests.get() also accepts a tuple as timeout. Example:

  requests.get(url, timeout=5, 45))

means

  - 5  seconds connect timeout
  - 45 seconds read    timeout
                       (e.g. time between received bytes in stream=True)


pandas: Safe memory when working with categorical data               2026-04-18
-------------------------------------------------------------------------------

When working with DataFrame columns with text, use the dtype `cateogry` to safe
memory. This is increasingly important, when:

 - working with DatFrames with millions of rows
 - when same string is stored in multiple cells

The default would store the following as `object`, even if it's duplicate data:

  df['source_system'] = 'Public API, version 1.0.0'

Instead, store the data a categorical to save memory:

  df['source_system'] = 'Public API, version 1.0.0'
  df['source_system'] = df['source_system'].astype('category')

This will consume more memory during initialization of the column. To store
cateogires directly, use:

  df['source_system'] = pd.Categorical.from_codes(
      np.zeros(len(df), dtype='int8'),
      categories=['Public API, version 1.0.0']
  )