has_many relationship from Rails for Pandas

has_many.py

import pandas as pd

@pd.api.extensions.register_dataframe_accessor("rel")
class RelationshipAccessor:
    '''
    Add a relationship accessor to dataframe objects allowing Rails-like
    access to related dataframes. e.g.

    >>> authors = pd.DataFrame({'name': ['C. S. Lewis', 'Lewis Carroll']})
    >>> books = pd.DataFrame({'title': ["Alice's Adventures in Wonderland",
    ...                                 'Through the looking glass',
    ...                                 'The Lion, the Witch, and the Wardrobe'],
    ...                       'author_id': [1, 1, 0]})
    >>> authors.rel.has_many('books', books, 'author_id')
    >>> len(authors[authors.name == 'Lewis Carroll'].rel.books)
    2
    '''
    def __init__(self, pandas_obj):
        self.obj = pandas_obj
        self.relations = {}
        # HACK: recurse up the tree of dataframes that this dataframe was
        # copied from (if any)
        # This has to be tracked manually because Pandas creates a fresh
        # RelationshipAccessor object every time it is accessed on a new
        # dataframe.
        if self.obj._is_copy:
            self.relations.update(self.obj._is_copy().rel.relations)

    def _get_related(self, df, colname):
        return df[df[colname].isin(self.obj.index)]

    def has_many(self, relname, df, colname):
        self.relations[relname] = (df, colname)

    def __getattr__(self, name):
        if name in self.relations:
            df, colname = self.relations[name]
            return self._get_related(df, colname)
        raise AttributeError('undefined attribute', name)

if __name__ == '__main__':
    import doctest
    doctest.testmod()