MAGPIE: Monte cArlo weiGhted PIxel rEmapping

Author	Krishna Naidoo
Version	0.2.2
Repository	https://github.com/knaidoo29/magpie
Documentation	https://magpie-doc.readthedocs.io/

WARNING: MAGPIE is currently in development. Functions and classes may change. Use with caution.

Introduction

MAGPIE is a python module for remapping bins (in 1D), pixels (in 2D) and cells (in 3D) into different coordinate systems. The package will enable data to be remapped from cartesian to polar coordinates, cartesian to spherical polar coordinates and will enable rotations of these systems. When carrying out coordinate transformations we typically take the center of the pixel or cell and apply the transform without considering the surface area of the pixel (or volume of the cell). In MAGPIE this is taken into account by applying two remapping schemes. The first weights pixels to a new coordinate grid using monte-carlo integration. The second uses a higher-resolution mesh (denser than the new coordinate grid by some integer factor along each dimension) which is rebinned to the target coordinate grid. In both cases we sample the surface area or volume of the new coordinate pixels to accurately remap the data. The monte-carlo method is more accurate but scales poorly to 3D. For 2D this scheme will work very well even for moderately large datasets. The higher-resolution mesh method, while less accurate, is very fast and should be used for large data sets and all 3D transformations. In 1D these are computed exactly without requiring the approximate schemes above.

NOTE: MPI functionality is currently unavailable but the plan will be to implement this using mpi4py and an additional library MPIutils which will handle all of the MPI enabled functions.

Dependencies

MAGPIE is being developed in Python 3.9 but should work on all versions >=3.4. MAGPIE is written mostly in python but with some heavy computation carried out in Fortran. Compiling the Fortran source code will require the availability of a fortran compiler such as gfortran or ifort.

The following Python modules are required:

• numpy
• scipy
• healpy

For testing you will require nose or pytest.

Installation

MAGPIE can be installed in a variety of ways -- using pip, conda or by directly cloning the repository. If you are having trouble installing MAGPIE or running MAGPIE we recommend using the conda install as this will setup the environment.

1. Using pip:

    pip install magpie-pkg

2. Using conda:

    conda install -c knaidoo29 magpie-pkg

3. By cloning the github repository::

    git clone https://github.com/knaidoo29/magpie.git
    cd magpie
    python setup.py build
    python setup.py install

Once this is done you should be able to call magpie from python:

    import magpie

Support

If you have any issues with the code or want to suggest ways to improve it please open a new issue (here) or (if you don't have a github account) email krishna.naidoo.11@ucl.ac.uk.

Version History

• Version 0.2:

  • Restructured layout and the incorporation of documentation and unit testing for eventual first release.
  • Randoms in a variety of coordinate systems: cartesian, polar, spherical and from an input PDF/CDF and subsampling and stochastic weights.
  • Cartesian coordinate remapping in 1D, 2D and 3D.

• Version 0.1:

  • Coordinate transformations between cartesian, polar and spherical polar coordinates.
  • Rebinning in 1D (computed exactly), in 2D and 3D via monte-carlo weighted remapping and a dense mesh.
  • Randoms in cartesian, polar and spherical polar coordinates.
  • Rotation transformations.
  • Polar coordinate utilities and integration for polar grid to radial profiles.
  • Plotting routine for orthographic projection of unit sphere data.