Introduction¶

This package lets you work with efficient, extendable semantic data types on tabular data in Python. The right abstraction of your data, through these types, is a powerful tool to effectively reduce complexity and increase performance, helping solve problems you are working on. Type systems are extremely flexible and applications include dataset creation, exploratory data analysis and data cleaning.

What can it do?¶

You can use the package in the following way:

• You pick the types and typeset that are relevant to your problem. The types and typeset can be chosen from the default library, or you can create custom types. (application: modeling the problem)

• visions can detect types of your data.

• It lets you define mappings from one type to another. This is helpful when working with real-world data. Provided these mappings, visions can infer and convert types. (application: data cleaning)

• You can then develop applications using the abstraction provided by the types.

How does it do it?¶

Essentially, visions does the following:

• It builds on Pandas, Numpy and Python.

• It extends the data types in Pandas to solve storage and implementation issues.

• On top of that it models typesets as graphs (networkx) and provides graph traversal algorithms for type detection, inference and conversion.