Arrow

From Alliance Doc
Jump to navigation Jump to search
This site replaces the former Compute Canada documentation site, and is now being managed by the Digital Research Alliance of Canada.

Ce site remplace l'ancien site de documentation de Calcul Canada et est maintenant géré par l'Alliance de recherche numérique du Canada.

Other languages:

Apache Arrow is a cross-language development platform for in-memory data. It uses a standardized language-independent columnar memory format for flat and hierarchical data, organized for efficient analytic operations. It also provides computational libraries and zero-copy streaming messaging and interprocess communication. Languages currently supported include C, C++, C#, Go, Java, JavaScript, MATLAB, Python, R, Ruby, and Rust.

CUDA

Arrow is also available with CUDA.

Question.png
[name@server ~]$ module load gcc/9.3.0 arrow/11 cuda/11.4

Python bindings

The module contains bindings for multiple Python versions. To discover which are the compatible Python versions, run

Question.png
[name@server ~]$ module spider arrow/4.0.0

Or search directly pyarrow, by running

Question.png
[name@server ~]$ module spider pyarrow

PyArrow

The Arrow Python bindings (also named PyArrow) have first-class integration with NumPy, Pandas, and built-in Python objects. They are based on the C++ implementation of Arrow.

1. Load the required modules.

Question.png
[name@server ~]$ module load gcc/9.3.0 arrow/8 python/3.8

2. Import PyArrow.

Question.png
[name@server ~]$ python -c "import pyarrow"

If the command displays nothing, the import was successful.

For more information, see the Arrow Python documentation.

Fulfilling other Python package dependency

Other Python packages depends on PyArrow in order to be installed. With the arrow module loaded, your package dependency for pyarrow will be satisfied.

Question.png
[name@server ~]$ pip list | grep pyarrow
pyarrow    8.0.0

Apache Parquet format

The Parquet file format is available.

To import the Parquet module, execute the previous steps for pyarrow, then run

Question.png
[name@server ~]$ python -c "import pyarrow.parquet"

If the command displays nothing, the import was successful.

R bindings

The Arrow package exposes an interface to the Arrow C++ library to access many of its features in R. This includes support for analyzing large, multi-file datasets (open_dataset()), working with individual Parquet files (read_parquet(), write_parquet()) and Feather files (read_feather(), write_feather()), as well as lower-level access to the Arrow memory and messages.

Installation

1. Load the required modules.

Question.png
[name@server ~]$ module load gcc/9.3.0 arrow/8 r/4.1 boost/1.72.0

2. Specify the local installation directory.

[name@server ~]$ mkdir -p ~/.local/R/$EBVERSIONR/
[name@server ~]$ export R_LIBS=~/.local/R/$EBVERSIONR/


3. Export the required variables to ensure you are using the system installation.

[name@server ~]$ export PKG_CONFIG_PATH=$EBROOTARROW/lib/pkgconfig
[name@server ~]$ export INCLUDE_DIR=$EBROOTARROW/include
[name@server ~]$ export LIB_DIR=$EBROOTARROW/lib


4. Install the bindings.

Question.png
[name@server ~]$ R -e 'install.packages("arrow", repos="https://cloud.r-project.org/")'

Usage

After the bindings are installed, they have to be loaded.

1. Load the required modules.

Question.png
[name@server ~]$ module load gcc/9.3.0 arrow/8 r/4.1

2. Load the library.

Question.png
[name@server ~]$ R -e "library(arrow)"
> library("arrow")
Attaching package: ‘arrow’

For more information, see the Arrow R documentation