r"""Parser for .prior specification files in bilby-style Python format."""
from dataclasses import dataclass
from pathlib import Path
from typing import Union, Any, Dict
from jesterTOV.inference.base import (
CombinePrior,
Prior,
UniformPrior,
MultivariateGaussianPrior,
)
from jesterTOV.inference.base.prior import Fixed
@dataclass
class ParsedPrior:
"""Result of parsing a ``.prior`` file.
Parameters
----------
prior : CombinePrior
Combined prior containing only the sampled (non-fixed) parameters.
fixed_params : dict[str, float]
Mapping of parameter name to fixed value for parameters declared with
``Fixed(...)`` in the prior file.
"""
prior: CombinePrior
fixed_params: dict[str, float]
[docs]
def parse_prior_file(
prior_file: Union[str, Path],
nb_CSE: int = 0,
) -> ParsedPrior:
"""Parse .prior file (Python format) and return a :class:`ParsedPrior`.
The prior file should contain Python variable assignments in bilby-style format:
.. code-block:: python
K_sat = UniformPrior(150.0, 300.0, parameter_names=["K_sat"])
Q_sat = UniformPrior(-500.0, 1100.0, parameter_names=["Q_sat"])
nbreak = UniformPrior(0.16, 0.32, parameter_names=["nbreak"])
# Pin a parameter to a fixed value (not sampled):
lambda_BL = Fixed(0.0, parameter_names=["lambda_BL"])
Parameters declared with :class:`~jesterTOV.inference.base.prior.Fixed` are
collected into :attr:`ParsedPrior.fixed_params` and excluded from the
sampling prior.
The parser will automatically:
- Include all NEP parameters (``_sat`` and ``_sym`` parameters)
- Include ``nbreak`` only if ``nb_CSE > 0``
- Add CSE grid parameters (``n_CSE_i_u``, ``cs2_CSE_i``) if ``nb_CSE > 0``
Parameters
----------
prior_file : str or Path
Path to .prior specification file (Python format)
nb_CSE : int, optional
Number of CSE parameters (0 for MetaModel only)
Returns
-------
ParsedPrior
Parsed prior with sampled :class:`CombinePrior` and ``fixed_params`` dict.
Raises
------
FileNotFoundError
If prior file does not exist
ValueError
If prior file format is invalid or no priors found
Examples
--------
>>> result = parse_prior_file("nep_standard.prior", nb_CSE=8)
>>> print(result.prior.n_dim) # Number of sampled dimensions
25 # 8 NEP + 1 nbreak + 8*2 CSE grid params
>>> print(result.fixed_params) # Any fixed parameters
{}
"""
prior_file = Path(prior_file)
if not prior_file.exists():
raise FileNotFoundError(f"Prior specification file not found: {prior_file}")
# Read the prior file
with open(prior_file, "r") as f:
prior_code = f.read()
# Create execution namespace with required imports only
namespace: dict[str, Any] = {
"UniformPrior": UniformPrior,
"MultivariateGaussianPrior": MultivariateGaussianPrior,
"Fixed": Fixed,
}
# Execute the prior file to populate the namespace
try:
exec(prior_code, namespace)
except Exception as e:
raise ValueError(f"Error executing prior file {prior_file}: {e}") from e
# Extract all Prior objects from the namespace
excluded_keys = {
"__builtins__",
"UniformPrior",
"MultivariateGaussianPrior",
"Fixed",
}
all_priors: Dict[str, Prior] = {}
for key, value in namespace.items():
if key not in excluded_keys and isinstance(value, Prior):
all_priors[key] = value
# Separate Fixed parameters from sampled priors
fixed_params: dict[str, float] = {}
sampled_priors: Dict[str, Prior] = {}
for param_name, prior in all_priors.items():
if isinstance(prior, Fixed):
fixed_params[prior.parameter_names[0]] = prior.value
else:
sampled_priors[param_name] = prior
# Filter sampled priors based on configuration
prior_list = []
for param_name, prior in sampled_priors.items():
# Always include NEP parameters (_sat and _sym)
if param_name.endswith("_sat") or param_name.endswith("_sym"):
prior_list.append(prior)
# Include nbreak only if nb_CSE > 0
elif param_name == "nbreak":
if nb_CSE > 0:
prior_list.append(prior)
else:
# Include any other parameters not handled by special cases
prior_list.append(prior)
# Add CSE grid parameters programmatically if nb_CSE > 0
if nb_CSE > 0:
for i in range(nb_CSE):
# Add n_CSE_i_u parameters (uniform [0, 1])
prior_list.append(UniformPrior(0.0, 1.0, parameter_names=[f"n_CSE_{i}_u"]))
# Add cs2_CSE_i parameters (uniform [0, 1])
prior_list.append(UniformPrior(0.0, 1.0, parameter_names=[f"cs2_CSE_{i}"]))
# Add final cs2 parameter for the grid point at nmax
# This gives us nb_CSE grid points + 1 final point at nmax
prior_list.append(UniformPrior(0.0, 1.0, parameter_names=[f"cs2_CSE_{nb_CSE}"]))
if len(prior_list) == 0:
raise ValueError(
f"No sampled priors found in {prior_file}. "
"Check file format and ensure at least one Prior object is defined. "
"Note: Fixed parameters do not count as sampled priors."
)
return ParsedPrior(prior=CombinePrior(prior_list), fixed_params=fixed_params)
