Welcome to cosmopipe’s documentation!¶
Developer documentation
Introduction¶
cosmopipe is a framework based on pypescript to extract cosmological constraints from galaxy redshift surveys. This is not a cosmological inference code, such a Cobaya 2, MontePython 3, CosmoMC 4 at al., which are designed to sample pre-coded or user-provided likelihoods. Rather, cosmopipe provides coherent interfaces (called modules) to different codes (e.g. power spectrum estimator, primordial perturbations, perturbation theory code, geometry effects, samplers), that handle common data types (data vectors, samples, etc.). pypescript 1 is then used to articulate these modules into analysis pipelines provided a configuration file. Running a clustering analysis is then as simple as:
pypescript myconfig.yaml
(probably using MPI.)
Structure¶
As described above, cosmopipe implements modules. These are documented in Modules.
In addition, cosmopipe relies on an internal library comopipe.lib, implementing common data types (data vectors, samples, etc.).
This library is documented in Library.
The rationale¶
As we are entering the era of precision cosmology with spectroscopic surveys, it becomes mandatory to provide an easy way for our collaborators to reproduce our analyses, from (at least) the clustering catalogs down to the cosmological constraints. Our collaborators should not have to clone our codes (e.g. theory models) from all over github; these codes should be linked in one place and run with the correct versions. This would be possible by running the codes inside the safe, controlled environment of a Docker 5 container.
Our collaborators should not have to worry about having the correct parameter files for each code, nor how to connect the output of one code to the input of another. Ideally, they should only have to deal with one (provided) parameter file and one output, and should not need to know anything about cosmology in general to rerun our analysis. This would be possible by linking all our codes together to form a pipeline.
This pipeline could be made flexible enough to cover the clustering analyses of the collaboration. For this, we should not have one pipeline, but a framework to script pipelines. For example, once codes (hereafter modules) to model 2 and 3-pt correlation functions (as well as their covariance) are available, it should be straightforward for anyone to run a joint 2 and 3-pt analysis, without coding anything else.
Eventually, it should be made very easy to code and include new modules, without the need for a global view of the pipeline. This would be possible by providing a simple module architecture to copy and fill, with either Python, C, C++ or Fortran code.
These should be the key features of cosmopipe, quite similar to CosmoSIS 6 used by the DES collaboration.
Acknowledgements¶
No acknowledgements to be made yet!