This describes how to install the python package. To setup an OmniSci (formerly MapD) server, see here.

pymapd can be installed with conda using conda-forge or pip.

# conda
conda install -c conda-forge pymapd

# pip
pip install pymapd

This actually installs two packages, pymapd, the pythonic interface to OmniSci, and mapd, the Apache thrift bindings for OmniSci.

There are several optional dependencies that may be useful. To return results sets into GPU memory as a GpuDataFrame, you’ll need cudf. To return results into CPU shared memory, using the Apache Arrow memory layout, you’ll need pyarrow and its dependencies.