Tafra: a minimalist dataframe

The tafra began life as a thought experiment: how could we reduce the idea of a dataframe (as expressed in libraries like pandas or languages like R) to its useful essence, while carving away the cruft? The original proof of concept stopped at "group by".

This library expands on the proof of concept to produce a practically useful tafra, which we hope you may find to be a helpful lightweight substitute for certain uses of pandas.

A tafra is, more-or-less, a set of named columns or dimensions. Each of these is a typed numpy array of consistent length, representing the values for each column by rows.

The library provides lightweight syntax for manipulating rows and columns, support for managing data types, iterators for rows and sub-frames, pandas-like "transform" support and conversion from pandas Dataframes, and SQL-style "group by" and join operations.

Tafra	Tafra
Aggregations	Union, GroupBy, Transform, IterateBy, InnerJoin, LeftJoin, CrossJoin
Aggregation Helpers	union, union_inplace, group_by, transform, iterate_by, inner_join, left_join, cross_join
Constructors	as_tafra, from_dataframe, from_series, from_records
SQL Readers	read_sql, read_sql_chunks
Destructors	to_records, to_list, to_tuple, to_array, to_pandas
Properties	rows, columns, data, dtypes, size, ndim, shape
Iter Methods	iterrows, itertuples, itercols
Functional Methods	row_map, tuple_map, col_map, pipe
Dict-like Methods	keys, values, items, get, update, update_inplace, update_dtypes, update_dtypes_inplace
Other Helper Methods	select, head, copy, rename, rename_inplace, coalesce, coalesce_inplace, _coalesce_dtypes, delete, delete_inplace
Printer Methods	pprint, pformat, to_html
Indexing Methods	_slice, _index, _ndindex

Getting Started

Install the library with pip:

pip install tafra

A short example

>>> from tafra import Tafra

>>> t = Tafra({
...    'x': np.array([1, 2, 3, 4]),
...    'y': np.array(['one', 'two', 'one', 'two'], dtype='object'),
... })

>>> t.pformat()
Tafra(data = {
 'x': array([1, 2, 3, 4]),
 'y': array(['one', 'two', 'one', 'two'])},
dtypes = {
 'x': 'int', 'y': 'object'},
rows = 4)

>>> print('List:', '\n', t.to_list())
List:
 [array([1, 2, 3, 4]), array(['one', 'two', 'one', 'two'], dtype=object)]

>>> print('Records:', '\n', tuple(t.to_records()))
Records:
 ((1, 'one'), (2, 'two'), (3, 'one'), (4, 'two'))

>>> gb = t.group_by(
...     ['y'], {'x': sum}
... )

>>> print('Group By:', '\n', gb.pformat())
Group By:
Tafra(data = {
 'x': array([4, 6]), 'y': array(['one', 'two'])},
dtypes = {
 'x': 'int', 'y': 'object'},
rows = 2)

Flexibility

Have some code that works with pandas, or just a way of doing things that you prefer? tafra is flexible:

>>> df = pd.DataFrame(np.c_[
...     np.array([1, 2, 3, 4]),
...     np.array(['one', 'two', 'one', 'two'])
... ], columns=['x', 'y'])

>>> t = Tafra.from_dataframe(df)

And going back is just as simple:

>>> df = pd.DataFrame(t.data)

Timings

In this case, lightweight also means performant. Beyond any additional features added to the library, tafra should provide the necessary base for organizing data structures for numerical processing. One of the most important aspects is fast access to the data itself. By minimizing abstraction to access the underlying numpy arrays, tafra provides an order of magnitude increase in performance.

Import note If you assign directly to the Tafra.data or Tafra._data attributes, you must call Tafra._coalesce_dtypes afterwards in order to ensure the typing is consistent.

Construct a Tafra and a DataFrame:

>>> tf = Tafra({
...     'x': np.array([1., 2., 3., 4., 5., 6.]),
...     'y': np.array(['one', 'two', 'one', 'two', 'one', 'two'], dtype='object'),
...     'z': np.array([0, 0, 0, 1, 1, 1])
... })

>>> df = pd.DataFrame(t.data)

Read Operations

Direct access:

>>> %timemit x = t._data['x']
55.3 ns ± 5.64 ns per loop (mean ± std. dev. of 7 runs, 10000000 loops each)

Indirect with some penalty to support Tafra slicing and numpy's advanced indexing:

>>> %timemit x = t['x']
219 ns ± 71.6 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

pandas timing:

>>> %timemit x = df['x']
1.55 µs ± 105 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)

This is the fastest methed for accessing the numpy array among alternatives of df.values(), df.to_numpy(), and df.loc[].

Assignment Operations

Direct access is not recommended as it avoids the validation steps, but it does provide fast access to the data attribute:

>>> x = np.arange(6)

>>> %timeit tf._data['x'] = x
65 ns ± 5.55 ns per loop (mean ± std. dev. of 7 runs, 10000000 loops each)

Indidrect access has a performance penalty due to the validation checks to ensure consistency of the tafra:

>>> %timeit tf['x'] = x
7.39 µs ± 950 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)

Even so, there is considerable performance improvement over pandas.

pandas timing:

>>> %timeit df['x'] = x
47.8 µs ± 3.53 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)

Grouping Operations

tafra also excels at aggregation methods, the primary of which are a SQL-like GROUP BY and the split-apply-combine equivalent to a SQL-like GROUP BY following by a LEFT JOIN back to the original table.

>>> %timeit tf.group_by(['y', 'z'], {'x': sum})
138 µs ± 4.03 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)

>>> %timeit tf.transform(['y', 'z'], {'sum_x': (sum, 'x')})
161 µs ± 2.31 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)

The equivalent pandas functions are given below. They require a chain of several object methods to perform the same role, and the transform requires a copy operation and assignment into the copied DataFrame in order to preserve immutability.

>>> %timeit df.groupby(['y','z']).agg({'x': 'sum'}).reset_index()
2.5 ms ± 177 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

>>> %%timeit
... tdf = df.copy()
... tdf['x'] = df.groupby(['y', 'z'])[['x']].transform(sum)
2.81 ms ± 143 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

petbox-dev / tafra