-
Notifications
You must be signed in to change notification settings - Fork 9
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request #409 from ImperialCollegeLondon/408-update-bounds-…
…checker New BoundsChecker implementation
- Loading branch information
Showing
9 changed files
with
309 additions
and
346 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,169 @@ | ||
"""Some functions in ``pyrealm`` are only well-behaved with given bounds but those | ||
bounds are often a little imprecise and real world data can contain extreme values. As a | ||
result, the bounds checking is deliberately not that intrusive: it warns when a variable | ||
contains out of value issues but leaves it up to the user to assess whether there is | ||
real problem and to adjust input data if needed. | ||
The ``bounds`` module: | ||
* Defines a {class}`~pyrealm.core.bounds.Bounds` dataclass used to define bounds for a | ||
particular variable. | ||
* Defines a {class}`~pyrealm.core.bounds.BoundsChecker` class with default bounds for | ||
core variables that acts as a library for bounds checking. | ||
* The main use case is e.g. ``BoundsChecker().check("tc", np.array([10, 1000])``, which | ||
will check that the alleged temperature data in °C fall within the configured bounds. | ||
A ``BoundsChecker`` class instance is created with a predefined internal dictionary of | ||
default variables and appropriate bounds. However, users can use the | ||
{meth}`~pyrealm.core.bounds.BoundsChecker.update` method to overide defaults or add new | ||
variables by providing a new ``Bounds`` instance. | ||
The {meth}`~pyrealm.core.bounds.BoundsChecker.check` method can then be used to validate | ||
a set of values against the configured bounds for a given variable name. The ``check`` | ||
method returns the input variables, to allow values to be checked while being assigned | ||
to an attribute. | ||
""" # noqa: D205 | ||
|
||
from dataclasses import dataclass | ||
from typing import Any, ClassVar | ||
from warnings import warn | ||
|
||
import numpy as np | ||
from numpy.typing import NDArray | ||
|
||
|
||
@dataclass | ||
class Bounds: | ||
"""Bounds checking dataclass for variables.""" | ||
|
||
var_name: str | ||
"""A variable name, typically the form used in function arguments.""" | ||
lower: float | ||
"""A lower bound on sensible values.""" | ||
upper: float | ||
"""An upper bound on sensible values.""" | ||
interval_type: str | ||
"""The interval type of the constraint ('[]', '()', '[)', '(]').""" | ||
unit: str | ||
"""A string giving the expected units.""" | ||
|
||
def __post_init__(self) -> None: | ||
"""Bounds data validation.""" | ||
if self.interval_type not in BoundsChecker._interval_types: | ||
raise ValueError(f"Unknown interval type: {self.interval_type}") | ||
|
||
if self.lower >= self.upper: | ||
raise ValueError(f"Bounds equal or reversed: {self.lower}, {self.upper}") | ||
|
||
|
||
class BoundsChecker: | ||
"""A bounds checker for input variables. | ||
The class provides a library of {class}`~pyrealm.core.bounds.Bounds` instances for | ||
core variables, keyed by the | ||
{attr}`Bounds.var_name<pyrealm.core.bounds.Bounds.var_name>` attribute. The table is | ||
populated from default values when a ``BoundsChecker`` instance is created but can | ||
be updated and extended by assigning new ``Bounds`` instances to existing or new | ||
variable name keys using the ``update`` method. | ||
""" | ||
|
||
# TODO - think about these argument names - some unnecessarily terse. | ||
_defaults: tuple[tuple[str, float, float, str, str], ...] = ( | ||
("tc", -25, 80, "[]", "°C"), | ||
("vpd", 0, 10000, "[]", "Pa"), | ||
("co2", 0, 1000, "[]", "ppm"), | ||
("patm", 30000, 110000, "[]", "Pa"), | ||
("fapar", 0, 1, "[]", "-"), | ||
("ppfd", 0, 3000, "[]", "-"), | ||
("theta", 0, 0.8, "[]", "m3 m-3"), | ||
("rootzonestress", 0, 1, "[]", "-"), | ||
("aridity_index", 0, 50, "[]", "-"), | ||
("mean_growth_temperature", 0, 50, "[]", "-"), | ||
("rh", 0, 1, "[]", "-"), | ||
("lat", -90, 90, "[]", "°"), | ||
("sf", 0, 1, "[]", "-"), | ||
("pn", 0, 1000, "[]", "mm day-1"), | ||
("kWm", 0, 1e4, "[]", "mm"), | ||
) | ||
"""Default bounds data for core forcing variables.""" | ||
|
||
_interval_types: ClassVar[dict[str, tuple[np.ufunc, np.ufunc]]] = { | ||
"()": (np.greater, np.less), | ||
"[]": (np.greater_equal, np.less_equal), | ||
"(]": (np.greater, np.less_equal), | ||
"[)": (np.greater_equal, np.less), | ||
} | ||
"""Dictionary of numpy function pairs for testing interval types.""" | ||
|
||
def __init__(self, *args: Any, **kwargs: Any) -> None: | ||
super().__init__(*args, **kwargs) | ||
|
||
self._data: dict[str, Bounds] = {} | ||
|
||
for var in self._defaults: | ||
var_bounds = Bounds(*var) | ||
self._data[var_bounds.var_name] = var_bounds | ||
|
||
def update(self, bounds: Bounds) -> None: | ||
"""Update or add bounds data. | ||
The {attr}`Bounds.var_name<pyrealm.core.bounds.Bounds.var_name>` attribute of | ||
the provided ``Bounds`` instance is used to update an existing entry for the | ||
name or add checking for a new name. | ||
Args: | ||
bounds: A Bounds instance. | ||
""" | ||
|
||
self._data[bounds.var_name] = bounds | ||
|
||
def check(self, var_name: str, values: NDArray) -> NDArray: | ||
r"""Check inputs fall within bounds. | ||
This method checks whether the provided values fall within the bounds specified | ||
for the given variable name and issues a warning when this is not the case. If | ||
the ``BoundsChecker`` class has not been configured the variable name then a | ||
warning will be given about lack of bounds checking. The method returns the | ||
input values, so that the method can be used as a pass through validator for | ||
assigning attributes. | ||
Args: | ||
var_name: The variable name | ||
values: An np.ndarray | ||
Returns: | ||
The input values. | ||
Examples: | ||
>>> vals = np.array([-15, 20, 30, 124], dtype=float) | ||
>>> bounds_checker = BoundsChecker() | ||
>>> bounds_checker.check("temp", vals) | ||
array([-15., 20., 30., 124.]) | ||
""" | ||
|
||
var_bounds = self._data.get(var_name) | ||
|
||
if var_bounds is None: | ||
warn( | ||
f"Variable '{var_name}' is not configured in the bounds checker. " | ||
"No bounds checking performed." | ||
) | ||
return values | ||
|
||
# Get the interval functions | ||
lower_func, upper_func = self._interval_types[var_bounds.interval_type] | ||
|
||
# Do the input values contain out of bound values? | ||
out_of_bounds = np.logical_xor( | ||
lower_func(values, var_bounds.lower), | ||
upper_func(values, var_bounds.upper), | ||
) | ||
|
||
if np.any(out_of_bounds): | ||
warn( | ||
f"Variable '{var_name}' ({var_bounds.unit}) contains values outside " | ||
f"the expected range ({var_bounds.lower},{var_bounds.upper}). " | ||
"Check units?" | ||
) | ||
|
||
return values |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.