How ebupi/astroariadne works
Project type and purpose
• astroARIADNE is an open-source Python library (package) whose primary purpose is to fit stellar spectral energy distributions (SEDs) using Bayesian model averaging over multiple atmosphere-model grids. It provides both a programmable API (importable modules) and command-line entry points via setup.py.
File structure
• Top-level files:
• setup.py, requirements.txt, MANIFEST.in – packaging and dependency specification
• README.md, citations.md, customization.md, filters.md, LICENSE – documentation & licensing
• test_bma.py – example/test script
• Core package directory astroARIADNE/:
• Data access & configuration: config.py, error.py, utils.py
• Datafiles subfolder: precomputed priors, model grids, lookup tables
• Domain modules:
– star.py (photometry retrieval, external catalog queries)
– phot_utils.py (filter definitions, photometric conversions)
– sed_library.py & librarian.py (model-grid loading and indexing)
– fitter.py (orchestration of nested sampling engines like dynesty/PyMultiNest, prior setup, likelihood evaluation)
External dependencies & services
• External services/APIs: Gaia cone‐search via astroquery, dust maps via dustmaps.
• Numerical & sampling libraries: numpy, scipy, pandas, numba, dynesty (optional PyMultiNest), isochrones.
• Visualization: matplotlib, corner, tqdm, termcolor.
• Data files: large model grids stored under astroARIADNE/Datafiles/model_grids.
Main components & their interactions
a. User/API entry point (script or REPL) → imports astroARIADNE.star.Star; sets RA/DEC/g_id → invokes photometry retrieval
b. Star object aggregates photometry, stellar priors (via Datafiles: Template_prior.dat, KDE/PPF pickles)
c. User configures fitter.Fitter with the Star instance, nested‐sampling settings, selected models, priors
d. Fitter calls librarian/sed_library to load atmosphere model grids from Datafiles/model_grids → synthesizes photometry through phot_utils
e. Likelihood evaluation and sampling: dynesty or optional PyMultiNest invoked inside fitter
f. Results I/O: Fitter writes out pickle files, best_fit.dat, prior.dat, mags.dat
Architectural patterns & design principles
• Layered modular design:
• Data‐Access Layer (Star, Librarian, Datafiles)
• Domain Logic Layer (Fitter, Isochrone)
• Presentation Layer (Plotter)
• Configuration & Utilities (config, utils, error)
• Pipeline / orchestrator pattern in fitter.py: orchestrates data retrieval, model evaluation, sampling
• Separation of concerns: photometry vs. modeling vs. plotting vs. I/O