Blackbear is an organic (standard library) library for key-based data manipulation and math using only built-in python dicts and sets (without numpy, pandas, polars).

Justification: there are a lot of great python-based data tools that make working with relational data much easier. pandas is such a tool, and is well suited to working with large(ish) datasets. numpy and polars are also great tools, though not as helpful if you rely on relational/labeled data. Unfortunately, pandas can be excruciatingly slow when used repeatedly on smaller data sets (see benchmarks). This can be the case in simulation tools, that have "unvectorizable" step functions. blackbear can be used to replace some of the functionality used in pandas.Series, basic math on aligned index (keys()).

Example Code:

>>> import blackbear as bb >>> data = {'foo': 60.0, 'bar': 16.0, 'baz': 24.0} >>> bb.add_scalar(data, 10) {'foo': 70.0, 'bar': 26.0, 'baz': 34.0} >>> blue = {'foo': 60.0, 'bar': 16.0, 'baz': 24.0} >>> green = {'foo': 40.0, 'bar': 4.0, 'baz': 6.0} >>> bb.add(blue, green) {'foo': 100.0, 'bar': 20.0, 'baz': 40.0}

"No numpy, no pandas, not even polars, I bet this is really, really slow. Right ?"

For certain use cases, it can be faster than any of those. Here is a guide:

• Use blackbear for frequent (millions) operations on small collections (< 20 items) where matching on an index (i.e dict keys) is needed.
• Do not use blackbear for operations on larger collections (> 50000).

See benchmark details and data below.

You are responsible for passing in the correct types to blackbear functions, we didn't want the additional overhead of type checking.

Blackbear officially supports Python 3.8+.

To install Blackbear, use pipenv (or pip, of course):

$ pipenv install blackbear

Documentation beyond this readme will be available soon.

1. Check for open issues or open a fresh issue to start a discussion around a feature idea or a bug.
2. Fork the repository on GitHub to start making your changes to the master branch (or branch off of it).
3. Write a test which shows that the bug was fixed or that the feature works as expected.
4. Send a pull request. Make sure to add yourself to AUTHORS.

> Performed on an Intel x64 chipped Mac (i7) with real blas and lapack installed.

100000 X 5 Element-wise ops on collection of 10

Pandas user 0m35.212s Polars user 0m3.398s Numpy user 0m1.437s Blackbear user 0m0.601s 

1000000 X 5 Element-wise ops on collection of 10

Pandas user 5m26.803s Polars user 0m24.115s Numpy user 0m6.734s Blackbear user 0m5.574s 

1000 X 5 Element-wise ops on collection of 10000

Pandas user 0m1.406s Polars user 0m1.055s Numpy user 0m0.737s Blackbear user 0m2.703s 

1000 X 5 Element-wise ops on collection of 100000

Pandas user 0m1.725s Polars user 0m1.230s Numpy user 0m1.035s Blackbear user 0m39.090s 

500000 X 5 Element-wise ops on collection of 5

Pandas user 2m46.098s Polars user 0m12.899s Numpy user 0m3.674s Blackbear user 0m2.025s