TJPCov

TJPCov is a general covariance calculator interface to be used within LSST DESC.

Installation

Quickstart

The easiest and recommended way to install TJPCov is to install it via conda:

conda install -c conda-forge tjpcov

Alternatively you may also install tjpcov via PyPi:

pip install tjpcov 

will install TJPCov with minimal dependencies, and

pip install 'tjpcov[full]'

will include all dependencies (for details, see Optional dependencies (PyPi only) section)

NOTE: If you plan to use TJPCov with MPI on NERSC, extra work must be done to get MPI running in your conda environment. See the NERSC docs for how to install MPI there.

Developer Installation

If you wish to contribute to TJPCov, follow the steps below to set up your development environment.

1. Clone the repository
2. Create the conda environment with conda env create --file environment.yml
3. Activate the environment with conda activate tjpcov
4. Run pip install -e .
5. Run pytest -vv tests/

Optional dependencies (PyPi only)

Because TJPCov relies on some packages that may not be necessary for every user, we have added different installation options to accommodate different use cases. For example, if a user has no plans to use MPI with TJPCov, they do not need mpi4py. Below we list the different installation options available on PyPi.

• pip install tjpcov will install tjpcov and the minimal dependencies.
• pip install tjpcov'[doc]' will install tjpcov, the minimal dependencies and the dependencies needed to build the documentation.
• pip install 'tjpcov[nmt]' will install tjpcov, the minimal dependencies and the dependencies needed to use NaMaster.
• pip install 'tjpcov[mpi4py]' will install, the minimal dependencies and the mpi4py library to use MPI parallelization. Does not work on NERSC (see above)
• pip install 'tjpcov[clusters]' will install tjpcov, the minimal dependencies and the dependencies needed for the cluster covariances (crow).
• pip install 'tjpcov[full]' will install tjpcov and all dependencies

Developer installation (PyPi only)

If you are using PyPi to set up your development environment (we recommend using conda instead), due to a bug in the NaMaster installation, one needs to make sure numpy is installed before trying to install NaMaster. For a fresh install, run python -m pip install . first, and then python -m pip install .\[nmt\]

Planning & development

Ask @felipeaoli or @mattkwiecien for access to the repository and join the #desc-mcp-cov channel on the LSST DESC slack to contribute.

We have adopted the following style convention (which are enforced in each PR):

• Google-style docstrings
• Black code style (with 79 characters line-width)
• PEP8 except for E203 (for better compatibility with black)

For a general idea of TJPCov's scientific scope, see also the terms of reference.

Contributing

We use black and flake8 configuration files so that code follows a unified coding style and remains PEP8 compliant.

This means before submitting your PR you must run the following in the root directory:

black .
flake8 .

Furthermore, we are following GitHub's recommendation of using Semantic Versioning in our releases.

Supported Python Versions

TJPCov currently runs on python 3.8, but python 3.9, 3.10 and 3.11 are supported.

TJPCov also has a few specific software versions hardcoded. Please check the pyproject.toml file to see version requirements.