Quick Install

Very basic but critical functionality can be obtained by following these steps:

• git clone this repo into your directory of choice $DIR (Note: this means enlib will be in $DIR/enlib. For e.g. the path to this file will be $DIR/enlib/README.md)
• add $DIR to your $PYTHONPATH

This should allow you to import and use many enlib modules assuming that basic dependencies like numpy are present on your system.

Dependencies

Enlib consists of a collection of relatively independend modules, so if you don't need all of them you don't need to install all dependencies. I've split the modules into groups based on typical use cases and list the dependencies of each. Some modules require compiling via f2py. This requires a fortran and C compiler, which settings are specified in the compile_opts directory. Which settings file is used is specified via the ENLIB_COMP environment variable. You will probably want to make your own file with settings for your computer, for example foo.mk, and then export ENLIB_COMP=foo in your .bashrc or similar.

Misc. utils

• Primary modules: utils, cg, colors, bins, errors
• Internal dependencies: None
• External dependencies: python2, numpy

Coordinate transformation

• Primary modules: coordinates
• Internal dependencies: iers + [misc utils]
• External dependencies: pyephem, astropy
• To build: make coordinates

Flat-sky maps

• Primary modules: enmap, powspec
• Internal dependencies: fft, array_ops, utils, wcs, slice + [misc utils]
• External dependencies: scipy, pyfftw, astropy, h5py, lapack
• To build: make array_ops

Curved-sky maps

• Primary modules: curvedsky, lensing
• Internal dependencies: sharp + [flat-sky maps]
• External dependencies: libsharp, cython
• To build: make sharp

Mapmaking

• Primary modules: zgetdata, map_equation, scan, scansim, gapfill, rangelist
• Internal dependencies: [misc. utils, coordinate transformation, flat-sky maps]
• External dependencies: mpi4py, pygetdata, zipfile, tempfile, shutil
• To build: make

Issues

• gfortran has gotten stricter/stupider about argument mismatches, which makes it difficult to apply blas to sub-sections of arrays without copying. This is used in the enlib noise matrix acceleration, so if you're trying to compile enlib with a recent gfortran you might get an error like "Type mismatch between actual argument at (1) and actual argument at (2)" or "Element of assumed-shape or pointer array as actual argument at (1) cannot correspond to actual argument at (2)". Pass -fallow-argument-mismatch to fix this. This can be set in your compile_opts.