• Welcome to MindSpore Pandas
• • What Is MindSpore Pandas
  • • Overview
    • Architecture
  • Installation Methods
  • • Confirming System Environment Information
    • Installing from pip command
    • Installing from source code
  • Installation Verification
  • Quickstart
  • Docs
  • Contributing
  • Release Notes
  • License

Welcome to MindSpore Pandas

查看中文

• What Is MindSpore Pandas
  • Overview
  • Architecture
• Installation Methods
  • Confirming System Environment Information
  • Installing from pip command
  • Installing from source code
• Installation Verification
• Quickstart
• Docs
• Contributing
• Release Notes
• License

What Is MindSpore Pandas

Overview

MindSpore Pandas uses distributed computing engine to accelerate pandas operations, seamlessly integrated and compatible with existing pandas code. Using MindSpore Pandas for calculations can use all CPU cores on the computer, which makes MindSpore Pandas works especially well on larger datasets.

Architecture

MindSpore Pandas is implemented based on distribution, while native pandas is implemented based on single thread. This means that only one CPU core can be used at a time.

However, MindSpore Pandas can use more threads and cores on the machine, or all cores of the entire cluster.

For detailed architecture design, please refer to official website document.

Installation Methods

Confirming System Environment Information

The following table lists the environment required for installing, compiling and running MindSpore Pandas:

software	version
Linux-x86_64	Ubuntu >=18.04 Euler >=2.9
Python	3.8-3.9
glibc	>=2.25

• Make sure libxml2-utils is installed in your environment.
• Please refer to requirements for other third party dependencies.

Installing from pip command

If you use the pip, please download the whl package from MindSpore Pandas page and install it.

Installing whl package will download MindSpore Pandas dependencies automatically (detail of dependencies is shown in requirements.txt) in the networked state, and other dependencies should be installed manually.

Installing from source code

Download source code, then enter the mindpandas directory to run build.sh script.

git clone https://gitee.com/mindspore/mindpandas.git
cd mindpandas
bash build.sh

The package is in output directory after compiled, and you can install with pip.

pip install output/mindpandas-0.1.0-cp38-cp38-linux_x86_64.whl

Installation Verification

Execute the following command in shell. If no No module named 'mindpandas' error is reported, the installation is successful.

python -c "import mindpandas"

Quickstart

First import MindSpore Pandas with the following command.

import mindpandas as pd

Set the running mode of MindSpore Pandas with the following command, which can speed up your MindSpore Pandas workflow.

pd.set_concurrency_mode('multithread')

The complete example is as follows:

>>> import mindpandas as pd
>>> pd.set_concurrency_mode('multithread')
>>> pd.set_partition_shape((16, 2))
>>> pd_df = pd.DataFrame([[1, 2, 3], [4, 5, 6]])
>>> sum = pd_df.sum()
>>> print(sum)
0 5
1 7
2 9
Name: sum, dtype: int64

Docs

More details about installation guide, tutorials and APIs, please see the
User Documentation.

Contributing

Welcome contributions. See our Contributor Wiki for
more details.

Release Notes

The release notes, see our RELEASE.

License

Apache License 2.0