Series Module Overview

Modin’s pandas.Series API

Modin’s pandas.Series API is backed by a distributed object providing an identical API to pandas. After the user calls some Series function, this call is internally rewritten into a representation that can be processed in parallel by the partitions. These results can be e.g., reduced to single output, identical to the single threaded pandas Series method output.

Public API

class modin.pandas.series.Series(data=None, index=None, dtype=None, name=None, copy=False, fastpath=False, query_compiler=None)

Modin distributed representation of pandas.Series.

Internally, the data can be divided into partitions in order to parallelize computations and utilize the user’s hardware as much as possible.

Inherit common for DataFrames and Series functionality from the BasePandasDataset class.

  • data (modin.pandas.Series, array-like, Iterable, dict, or scalar value, optional) – Contains data stored in Series. If data is a dict, argument order is maintained.

  • index (array-like or Index (1d), optional) – Values must be hashable and have the same length as data.

  • dtype (str, np.dtype, or pandas.ExtensionDtype, optional) – Data type for the output Series. If not specified, this will be inferred from data.

  • name (str, optional) – The name to give to the Series.

  • copy (bool, default: False) – Copy input data.

  • fastpath (bool, default: False) – pandas internal parameter.

  • query_compiler (BaseQueryCompiler, optional) – A query compiler object to create the Series from.


See pandas API documentation for pandas.Series for more.

Usage Guide

The most efficient way to create Modin Series is to import data from external storage using the highly efficient Modin IO methods (for example using pd.read_csv, see details for Modin IO methods in the separate section), but even if the data does not originate from a file, any pandas supported data type or pandas.Series can be used. Internally, the Series data is divided into partitions, which number along an axis usually corresponds to the number of the user’s hardware CPUs. If needed, the number of partitions can be changed by setting modin.config.NPartitions.

Let’s consider simple example of creation and interacting with Modin Series:

import modin.config

# This explicitly sets the number of partitions

import modin.pandas as pd
import pandas

# Create Modin Series from the external file
pd_series = pd.read_csv("test_data.csv", header=None).squeeze()
# Create Modin Series from the python object
# pd_series = pd.Series([x for x in range(256)])
# Create Modin Series from the pandas object
# pd_series = pd.Series(pandas.Series([x for x in range(256)]))

# Show created `Series`

# List `Series` partitions. Note, that internal API is intended for
# developers needs and was used here for presentation purposes
# only.
partitions = pd_series._query_compiler._modin_frame._partitions

# Show the first `Series` partition


# created `Series`

0      100
1      101
2      102
3      103
4      104
251    351
252    352
253    353
254    354
255    355
Name: 0, Length: 256, dtype: int64

# List of `Series` partitions

[[<modin.engines.ray.pandas_on_ray.frame.partition.PandasOnRayFramePartition object at 0x000001E7CD11BD60>]
[<modin.engines.ray.pandas_on_ray.frame.partition.PandasOnRayFramePartition object at 0x000001E7CD11BE50>]
[<modin.engines.ray.pandas_on_ray.frame.partition.PandasOnRayFramePartition object at 0x000001E7CD11BF40>]
[<modin.engines.ray.pandas_on_ray.frame.partition.PandasOnRayFramePartition object at 0x000001E7CD13E070>]]

# The first `Series` partition

0   100
1   101
2   102
3   103
4   104
..  ...
60  160
61  161
62  162
63  163
64  164

[65 rows x 1 columns]

As we show in the example above, Modin Series can be easily created, and supports any input that pandas Series supports. Also note that tuning of the Series partitioning can be done by just setting a single config.