Add a Metadata.detect_from_dataframe function

[npatki]

Problem Description

In an upcoming version of SDV, we will be consolidating the Multi and Single table metadata objects into a streamlined Metadata object. Effectively, this object is the same as multi-table, so it includes functions such as detect_from_dataframes for detecting multi tables at once.

However, if you have a single table, such a function is not ideal. It forces you to input your data as a dictionary even though you only have a single dataframe. We should not be adding more friction to the single table case.

Expected behavior

The streamlined metadata object should have a function called detect_from_dataframe (singular). To improve the usage of this function, we will make it a class function that ultimately returns an instance of the metadata object.

from sdv.metadata import Metadata

metadata = Metadata.detect_from_dataframe(data=my_dataframe, table_name='users')

Parameters:

• (required) data, a singular pandas DataFrame object that represents a single table
• table_name: The name of the table that the data represents. This will be saved in the metadata.
  • (default) If no table name is provided, just default to using table

Functionality: Detect the columns from the data create metadata with the given table name.

Output: An instance of the metadata object with the detected metadata

Additional context

For multi-table usages, we still recommend using detect_from_dataframes (plural) in order to detect relationships.