Better Parameter Validation

.github/workflows/tests.yml

@@ -13,7 +13,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: [3.8]
+        python-version: ['3.9', '3.10', '3.11']
 
     steps:
       - uses: actions/checkout@v3

requirements.txt

@@ -6,3 +6,4 @@ scipy
 speclite
 pyyaml
 matplotlib
+pydantic>=2

setup.py

@@ -19,7 +19,7 @@
 setup(
     author="DESC/SLSC",
     author_email="[email protected]",
-    python_requires=">=3.6",
+    python_requires=">=3.9",
     classifiers=[
         "Development Status :: 2 - Pre-Alpha",
         "Intended Audience :: Developers",

slsim/Deflectors/_params/elliptical_lens_galaxies.py

@@ -0,0 +1,5 @@
+from pydantic import BaseModel, PositiveFloat
+
+
+class vel_disp_from_m_star(BaseModel):
+    m_star: PositiveFloat

slsim/Deflectors/_params/velocity_dispersion.py

@@ -0,0 +1,11 @@
+from pydantic import BaseModel, PositiveFloat
+from astropy.cosmology import Cosmology
+
+
+class vel_disp_composite_model(BaseModel, arbitrary_types_allowed=True):
+    r: PositiveFloat
+    m_star: PositiveFloat
+    rs_star: PositiveFloat
+    m_halo: PositiveFloat
+    c_halo: float
+    cosmo: Cosmology

slsim/Deflectors/elliptical_lens_galaxies.py

@@ -4,6 +4,7 @@
 from slsim.Deflectors.velocity_dispersion import vel_disp_sdss
 from slsim.Util import param_util
 from slsim.Deflectors.deflector_base import DeflectorBase
+from slsim.Util import check_params
 
 
 class EllipticalLensGalaxies(DeflectorBase):
@@ -120,6 +121,7 @@ def elliptical_projected_eccentricity(ellipticity, **kwargs):
     return e1_light, e2_light, e1_mass, e2_mass
 
 
+@check_params
 def vel_disp_from_m_star(m_star):
     """Function to calculate the velocity dispersion from the staller mass using
     empirical relation for elliptical galaxies.

slsim/Deflectors/velocity_dispersion.py

@@ -2,6 +2,7 @@
 import scipy
 from skypy.galaxies.redshift import redshifts_from_comoving_density
 from skypy.utils.random import schechter
+from slsim.Util import check_params
 
 """
 This module provides functions to compute velocity dispersion using schechter function.
@@ -11,6 +12,7 @@
 #  from skypy.galaxies.velocity_dispersion import schechter_vdf
 
 
+@check_params
 def vel_disp_composite_model(r, m_star, rs_star, m_halo, c_halo, cosmo):
     """Computes the luminosity weighted velocity dispersion for a deflector with a
     stellar Hernquist profile and a NFW halo profile, assuming isotropic anisotropy.

slsim/ParamDistributions/_params/gaussian_mixture_model.py

@@ -0,0 +1,35 @@
+from pydantic import (
+    BaseModel,
+    PositiveFloat,
+    PositiveInt,
+    model_validator,
+    field_validator,
+)
+import numpy as np
+
+
+class GaussianMixtureModel(BaseModel):
+    means: list[float] = [0.00330796, -0.07635054, 0.11829008]
+    stds: list[PositiveFloat] = [
+        np.sqrt(0.00283885),
+        np.sqrt(0.01066668),
+        np.sqrt(0.0097978),
+    ]
+    weights: list[PositiveFloat] = [0.62703102, 0.23732313, 0.13564585]
+
+    @field_validator("weights")
+    @classmethod
+    def check_weights(cls, weight_values):
+        if sum(weight_values) != 1:
+            raise ValueError("The sum of the weights must be 1")
+        return weight_values
+
+    @model_validator(mode="after")
+    def check_lengths(self):
+        if len(self.means) != len(self.stds) or len(self.means) != len(self.weights):
+            raise ValueError("The lengths of means, stds and weights must be equal")
+        return self
+
+
+class GaussianMixtureModel_rvs(BaseModel):
+    size: PositiveInt

slsim/ParamDistributions/gaussian_mixture_model.py

@@ -1,4 +1,5 @@
 import numpy as np
+from slsim.Util import check_params
 
 
 class GaussianMixtureModel:
@@ -8,11 +9,13 @@ class GaussianMixtureModel:
     is defined by its mean, standard deviation and weight.
     """
 
-    def __init__(self, means=None, stds=None, weights=None):
+    @check_params
+    def __init__(self, means: list[float], stds: list[float], weights: list[float]):
         """
         The constructor for GaussianMixtureModel class. The default values are the
         means, standard deviations, and weights of the fits to the data in the table
-        2 of https://doi.org/10.1093/mnras/stac2235 and others.
+        2 of https://doi.org/10.1093/mnras/stac2235 and others. See "_params" for
+        defaults and validation logic.
 
         :param means: the mean values of the Gaussian components.
         :type means: list of float
@@ -21,20 +24,12 @@ def __init__(self, means=None, stds=None, weights=None):
         :param weights: The weights of the Gaussian components in the mixture.
         :type weights: list of float
         """
-        if means is None:
-            means = [0.00330796, -0.07635054, 0.11829008]
-        if stds is None:
-            stds = [np.sqrt(0.00283885), np.sqrt(0.01066668), np.sqrt(0.0097978)]
-        if weights is None:
-            weights = [0.62703102, 0.23732313, 0.13564585]
-        assert (
-            len(means) == len(stds) == len(weights)
-        ), "Lengths of means, standard deviations, and weights must be equal."
         self.means = means
         self.stds = stds
         self.weights = weights
 
-    def rvs(self, size):
+    @check_params
+    def rvs(self, size: int):
         """Generate random variables from the GMM distribution.
 
         :param size: The number of random variables to generate.

slsim/Util/__init__.py

@@ -0,0 +1,3 @@
+from .params import check_params
+
+__all__ = ["check_params"]

slsim/Util/params.py

@@ -0,0 +1,208 @@
+"""Utilities for managing parameter defaults and validation in the slsim package.
+
+Desgined to be unobtrusive to use.
+"""
+from functools import wraps
+from importlib import import_module
+from typing import Callable, Any, TypeVar
+from enum import Enum
+import inspect
+import pydantic
+
+"""
+Set of routines for validating inputs to functions and classes. The elements of this
+module should never be imported directly. Instead, @check_params can be imported 
+directly from the Util module.
+"""
+
+
+class SlSimParameterException(Exception):
+    pass
+
+
+_defaults = {}
+
+
+class _FnType(Enum):
+    """Enum for the different types of functions we can have. This is used to determine
+    how to parse the arguments to the function.
+
+    There are three possible cases:
+        1. The function is a standard function, defined outside a class
+        2. The function is a standard object method,
+           taking "self" as the first parameter
+        3. The funtion is a class method (or staticmethod), not taking
+           "self" as the first parameter
+    """
+
+    STANDARD = 0
+    METHOD = 1
+    CLASSMETHOD = 2
+
+
+def determine_fn_type(fn: Callable) -> _FnType:
+    """Determine which of the three possible cases a function falls into. Cases 0 and 2
+    are actually functionally identical. Things only get spicy when we have a "self"
+    argument.
+
+    However the tricky thing is that decorators operate on functions and methods when
+    they are imported, not when they are used. This means "inspect.ismethod" will always
+    return False, even if the function is a method.
+
+    We can get around this by checking if the parent of the function is a class. Then,
+    we check if the first argument of the function is "self". If both of these are true,
+    then the function is a method.
+    """
+    if not inspect.isfunction(fn):
+        raise TypeError("decorator @check_params can only be used on functions!")
+    qualified_obj_name = fn.__qualname__
+    qualified_obj_path = qualified_obj_name.split(".")
+    if len(qualified_obj_path) == 1:
+        # If the qualified name isn't split, this is a standard function not
+        # attached to a class
+        return _FnType.STANDARD
+
+    spec = inspect.getfullargspec(fn)
+    if spec.args[0] == "self":
+        return _FnType.METHOD
+    else:
+        return _FnType.CLASSMETHOD
+
+
+T_ = TypeVar("T_")
+
+
+def check_params(fn: Callable[..., T_]) -> Callable[..., T_]:
+    """A decorator for enforcing checking of params in __init__ methods. This decorator
+    will automatically load the default parameters for the class and check that the
+    passed parameters are valid. It expeects a "params.py" file in the same folder as
+    the class definition. Uses pydantic models to enforce types, sanity checks, and
+    defaults.
+
+    From and end user perspective, there is no difference between this and a normal
+    __init__ fn. Developers only need to add @check_params above their __init__ method
+    definition to enable this feature, then add their default parameters to the
+    "params.py" file.
+    """
+    fn_type = determine_fn_type(fn)
+    if fn_type == _FnType.STANDARD:
+        new_fn = standard_fn_wrapper(fn)
+    elif fn_type == _FnType.METHOD:
+        new_fn = method_fn_wrapper(fn)
+    elif fn_type == _FnType.CLASSMETHOD:
+        new_fn = standard_fn_wrapper(fn)
+
+    return new_fn
+
+
+def standard_fn_wrapper(fn: Callable[..., T_]) -> Callable[..., T_]:
+    """A wrapper for standard functions.
+
+    This is used to parse the arguments to the function and check that they are valid.
+    """
+
+    @wraps(fn)
+    def new_fn(*args, **kwargs) -> T_:
+        # Get function argument names
+        pargs = {}
+        if args:
+            largs = list(inspect.signature(fn).parameters.keys())
+            for i in range(len(args)):
+                arg_value = args[i]
+                if arg_value is not None:
+                    pargs[largs[i]] = args[i]
+        # Doing it this way ensures we still catch duplicate arguments
+        defaults = get_defaults(fn)
+        parsed_args = defaults(**pargs, **kwargs)
+        return fn(**dict(parsed_args))
+
+    return new_fn
+
+
+def method_fn_wrapper(fn: Callable[..., T_]) -> Callable[..., T_]:
+    @wraps(fn)
+    def new_fn(obj: Any, *args, **kwargs) -> T_:
+        # Get function argument names
+        parsed_args = {}
+        if args:
+            largs = list(inspect.signature(fn).parameters.keys())
+
+            for i in range(len(args)):
+                arg_value = args[i]
+                if arg_value is not None:
+                    parsed_args[largs[i + 1]] = arg_value
+        # Doing it this way ensures we still catch duplicate arguments
+        parsed_kwargs = {k: v for k, v in kwargs.items() if v is not None}
+        defaults = get_defaults(fn)
+        parsed_args = defaults(**parsed_args, **parsed_kwargs)
+        return fn(obj, **dict(parsed_args))
+
+    return new_fn
+
+
+def get_defaults(fn: Callable) -> pydantic.BaseModel:
+    module_trace = inspect.getmodule(fn).__name__.split(".")
+    file_name = module_trace[-1]
+    parent_trace = module_trace[:-1]
+    parent_path = ".".join(parent_trace)
+    param_path = ".".join([parent_path, "_params"])
+    fn_qualname = fn.__qualname__
+    cache_name = parent_path + "." + fn_qualname
+    if cache_name in _defaults:
+        return _defaults[cache_name]
+
+    try:
+        _ = import_module(param_path)
+    except ModuleNotFoundError:
+        raise SlSimParameterException(
+            f'No default parameters found in module {".".join(parent_trace)},'
+            " but something in that module is trying to use the @check_params decorator"
+        )
+    try:
+        param_model_file = import_module(f"{param_path}.{file_name}")
+    except AttributeError:
+        raise SlSimParameterException(
+            f'No default parameters found for file "{file_name}" in module '
+            f'{".".join(parent_trace)}, but something in that module is trying to use '
+            "the @check_params decorator"
+        )
+
+    if fn.__name__ == "__init__":
+        expected_model_name = "_".join(fn_qualname.split(".")[:-1])
+    else:
+        expected_model_name = "_".join(fn_qualname.split("."))
+
+    try:
+        parameter_model = getattr(param_model_file, expected_model_name)
+    except AttributeError:
+        raise SlSimParameterException(
+            "No default parameters found for function " f'"{fn_qualname}"'
+        )
+    if not issubclass(parameter_model, pydantic.BaseModel):
+        raise SlSimParameterException(
+            f'Defaults for "{fn_qualname}" are not in a pydantic model!'
+        )
+    _defaults[cache_name] = parameter_model
+    return _defaults[cache_name]
+
+
+def load_parameters(modpath: str, obj_name: str) -> pydantic.BaseModel:
+    """Loads parameters from the "params.py" file which should be in the same folder as
+    the class definition."""
+    try:
+        defaults = import_module(modpath)
+    except ModuleNotFoundError:
+        raise SlSimParameterException(
+            "No default parameters found in module " f'"{modpath[-2]}"'
+        )
+    try:
+        obj_defaults = getattr(defaults, obj_name)
+    except AttributeError:
+        raise SlSimParameterException(
+            f"No default parameters found for class " f'"{obj_name}"'
+        )
+    if not issubclass(obj_defaults, pydantic.BaseModel):
+        raise SlSimParameterException(
+            f'Defaults for "{obj_name}" are not in a ' "pydantic model!"
+        )
+    return obj_defaults