modelbase2 0.1.79__py3-none-any.whl → 0.2.0__py3-none-any.whl

modelbase2/sbml/_import.py

@@ -0,0 +1,585 @@
+from __future__ import annotations
+
+import math  # noqa: F401  # models might need it
+import re
+import warnings
+from collections import defaultdict
+from dataclasses import dataclass, field
+from pathlib import Path
+from typing import TYPE_CHECKING, Self
+
+import libsbml
+import numpy as np  # noqa: F401  # models might need it
+import sympy
+
+from modelbase2.model import Model, _sort_dependencies
+from modelbase2.paths import default_tmp_dir
+from modelbase2.sbml._data import (
+    AtomicUnit,
+    Compartment,
+    CompositeUnit,
+    Compound,
+    Derived,
+    Function,
+    Parameter,
+    Reaction,
+)
+from modelbase2.sbml._mathml import parse_sbml_math
+from modelbase2.sbml._name_conversion import _name_to_py
+from modelbase2.sbml._unit_conversion import get_operator_mappings, get_unit_conversion
+from modelbase2.types import unwrap
+
+if TYPE_CHECKING:
+    from collections.abc import Callable
+
+__all__ = [
+    "INDENT",
+    "OPERATOR_MAPPINGS",
+    "Parser",
+    "UNIT_CONVERSION",
+    "import_from_path",
+    "read",
+    "valid_filename",
+]
+
+UNIT_CONVERSION = get_unit_conversion()
+OPERATOR_MAPPINGS = get_operator_mappings()
+INDENT = "    "
+
+
+def _nan_to_zero(value: float) -> float:
+    return 0 if str(value) == "nan" else value
+
+
+@dataclass(slots=True)
+class Parser:
+    # Collections
+    boundary_species: set[str] = field(default_factory=set)
+    # Parsed stuff
+    atomic_units: dict[str, AtomicUnit] = field(default_factory=dict)
+    composite_units: dict[str, CompositeUnit] = field(default_factory=dict)
+    compartments: dict[str, Compartment] = field(default_factory=dict)
+    parameters: dict[str, Parameter] = field(default_factory=dict)
+    variables: dict[str, Compound] = field(default_factory=dict)
+    derived: dict[str, Derived] = field(default_factory=dict)
+    initial_assignment: dict[str, Derived] = field(default_factory=dict)
+    functions: dict[str, Function] = field(default_factory=dict)
+    reactions: dict[str, Reaction] = field(default_factory=dict)
+
+    def parse(self, file: str | Path) -> Self:
+        if not Path(file).exists():
+            msg = "Model file does not exist"
+            raise OSError(msg)
+        doc = libsbml.readSBMLFromFile(str(file))
+
+        # Check for unsupported packages
+        for i in range(doc.num_plugins):
+            if doc.getPlugin(i).getPackageName() == "comp":
+                msg = "No support for comp package"
+                raise NotImplementedError(msg)
+
+        sbml_model = doc.getModel()
+        if sbml_model is None:
+            return self
+
+        if bool(sbml_model.getConversionFactor()):
+            msg = "Conversion factors are currently not supported"
+            raise NotImplementedError(msg)
+
+        self.parse_functions(sbml_model)
+        self.parse_units(sbml_model)
+        self.parse_compartments(sbml_model)
+        self.parse_variables(sbml_model)
+        self.parse_parameters(sbml_model)
+        self.parse_initial_assignments(sbml_model)
+        self.parse_rules(sbml_model)
+        self.parse_constraints(sbml_model)
+        self.parse_reactions(sbml_model)
+        self.parse_events(sbml_model)
+
+        # Modifications
+        self._convert_substance_amount_to_concentration()
+        # self._rename_species_references()
+        return self
+
+    ###############################################################################
+    # PARSING STAGE
+    ###############################################################################
+
+    def parse_constraints(self, sbml_model: libsbml.Model) -> None:
+        if len(sbml_model.getListOfConstraints()) > 0:
+            msg = "modelbase does not support model constraints. "
+            raise NotImplementedError(msg)
+
+    def parse_events(self, sbml_model: libsbml.Model) -> None:
+        if len(sbml_model.getListOfEvents()) > 0:
+            msg = (
+                "modelbase does not current support events. "
+                "Check the file for how to integrate properly."
+            )
+            raise NotImplementedError(msg)
+
+    def parse_units(self, sbml_model: libsbml.Model) -> None:
+        for unit_definition in sbml_model.getListOfUnitDefinitions():
+            composite_id = unit_definition.getId()
+            local_units = []
+            for unit in unit_definition.getListOfUnits():
+                atomic_unit = AtomicUnit(
+                    kind=UNIT_CONVERSION[unit.getKind()],
+                    scale=unit.getScale(),
+                    exponent=unit.getExponent(),
+                    multiplier=unit.getMultiplier(),
+                )
+                local_units.append(atomic_unit.kind)
+                self.atomic_units[atomic_unit.kind] = atomic_unit
+            self.composite_units[composite_id] = CompositeUnit(
+                sbml_id=composite_id,
+                units=local_units,
+            )
+
+    def parse_compartments(self, sbml_model: libsbml.Model) -> None:
+        for compartment in sbml_model.getListOfCompartments():
+            sbml_id = _name_to_py(compartment.getId())
+            size = compartment.getSize()
+            if str(size) == "nan":
+                size = 0
+            self.compartments[sbml_id] = Compartment(
+                name=compartment.getName(),
+                dimensions=compartment.getSpatialDimensions(),
+                size=size,
+                units=compartment.getUnits(),
+                is_constant=compartment.getConstant(),
+            )
+
+    def parse_parameters(self, sbml_model: libsbml.Model) -> None:
+        for parameter in sbml_model.getListOfParameters():
+            self.parameters[_name_to_py(parameter.getId())] = Parameter(
+                value=_nan_to_zero(parameter.getValue()),
+                is_constant=parameter.getConstant(),
+            )
+
+    def parse_variables(self, sbml_model: libsbml.Model) -> None:
+        for compound in sbml_model.getListOfSpecies():
+            compound_id = _name_to_py(compound.getId())
+            if bool(compound.getConversionFactor()):
+                msg = "Conversion factors are not implemented. "
+                raise NotImplementedError(msg)
+
+            # NOTE: What the shit is this?
+            initial_amount = compound.getInitialAmount()
+            if str(initial_amount) == "nan":
+                initial_amount = compound.getInitialConcentration()
+                is_concentration = str(initial_amount) != "nan"
+            else:
+                is_concentration = False
+
+            has_boundary_condition = compound.getBoundaryCondition()
+            if has_boundary_condition:
+                self.boundary_species.add(compound_id)
+
+            self.variables[compound_id] = Compound(
+                compartment=compound.getCompartment(),
+                initial_amount=_nan_to_zero(initial_amount),
+                substance_units=compound.getSubstanceUnits(),
+                has_only_substance_units=compound.getHasOnlySubstanceUnits(),
+                has_boundary_condition=has_boundary_condition,
+                is_constant=compound.getConstant(),
+                is_concentration=is_concentration,
+            )
+
+    def parse_functions(self, sbml_model: libsbml.Model) -> None:
+        for func in sbml_model.getListOfFunctionDefinitions():
+            func_name = func.getName()
+            sbml_id = func.getId()
+            if sbml_id is None or sbml_id == "":
+                sbml_id = func_name
+            elif func_name is None or func_name == "":
+                func_name = sbml_id
+            func_name = _name_to_py(func_name)
+
+            if (node := func.getMath()) is None:
+                continue
+            body, args = parse_sbml_math(node=node)
+
+            self.functions[func_name] = Function(
+                body=body,
+                args=args,
+            )
+
+    ###############################################################################
+    # Different kinds of derived values
+    ###############################################################################
+
+    def parse_initial_assignments(self, sbml_model: libsbml.Model) -> None:
+        for assignment in sbml_model.getListOfInitialAssignments():
+            name = _name_to_py(assignment.getSymbol())
+
+            node = assignment.getMath()
+            if node is None:
+                warnings.warn(
+                    f"Unusable math for {name}",
+                    stacklevel=1,
+                )
+                continue
+
+            body, args = parse_sbml_math(node)
+            self.initial_assignment[name] = Derived(
+                body=body,
+                args=args,
+            )
+
+    def _parse_algebraic_rule(self, rule: libsbml.AlgebraicRule) -> None:
+        msg = f"Algebraic rules are not implemented for {rule.getId()}"
+        raise NotImplementedError(msg)
+
+    def _parse_assignment_rule(self, rule: libsbml.AssignmentRule) -> None:
+        if (node := rule.getMath()) is None:
+            return
+
+        name: str = _name_to_py(rule.getId())
+        body, args = parse_sbml_math(node=node)
+
+        self.derived[name] = Derived(
+            body=body,
+            args=args,
+        )
+
+    def _parse_rate_rule(self, rule: libsbml.RateRule) -> None:
+        msg = f"Skipping rate rule {rule.getId()}"
+        raise NotImplementedError(msg)
+
+    def parse_rules(self, sbml_model: libsbml.Model) -> None:
+        """Parse rules and separate them by type."""
+        for rule in sbml_model.getListOfRules():
+            if rule.element_name == "algebraicRule":
+                self._parse_algebraic_rule(rule=rule)
+            elif rule.element_name == "assignmentRule":
+                self._parse_assignment_rule(rule=rule)
+            elif rule.element_name == "rateRule":
+                self._parse_rate_rule(rule=rule)
+            else:
+                msg = "Unknown rate type"
+                raise ValueError(msg)
+
+    def _parse_local_parameters(
+        self, reaction_id: str, kinetic_law: libsbml.KineticLaw
+    ) -> dict[str, str]:
+        """Parse local parameters."""
+        parameters_to_update = {}
+        for parameter in kinetic_law.getListOfLocalParameters():
+            old_id = _name_to_py(parameter.getId())
+            if old_id in self.parameters:
+                new_id = f"{reaction_id}__{old_id}"
+                parameters_to_update[old_id] = new_id
+            else:
+                new_id = old_id
+            self.parameters[new_id] = Parameter(
+                value=_nan_to_zero(parameter.getValue()),
+                is_constant=parameter.getConstant(),
+            )
+        # Some models apparently also write local parameters in this
+        for parameter in kinetic_law.getListOfParameters():
+            old_id = _name_to_py(parameter.getId())
+            if old_id in self.parameters:
+                new_id = f"{reaction_id}__{old_id}"
+                parameters_to_update[old_id] = new_id
+            else:
+                new_id = old_id
+            self.parameters[new_id] = Parameter(
+                value=_nan_to_zero(parameter.getValue()),
+                is_constant=parameter.getConstant(),
+            )
+        return parameters_to_update
+
+    def parse_reactions(self, sbml_model: libsbml.Model) -> None:
+        for reaction in sbml_model.getListOfReactions():
+            sbml_id = _name_to_py(reaction.getId())
+            kinetic_law = reaction.getKineticLaw()
+            if kinetic_law is None:
+                continue
+            parameters_to_update = self._parse_local_parameters(
+                reaction_id=sbml_id,
+                kinetic_law=kinetic_law,
+            )
+
+            node = reaction.getKineticLaw().getMath()
+            # FIXME: convert substance amount to concentration here
+            body, args = parse_sbml_math(node=node)
+
+            # Update parameter references
+            for old, new in parameters_to_update.items():
+                pat = re.compile(f"({old})" + r"\b")
+                body = pat.sub(new, body)
+
+            for i, arg in enumerate(args):
+                args[i] = parameters_to_update.get(arg, arg)
+
+            dynamic_stoichiometry: dict[str, str] = {}
+            parsed_reactants: defaultdict[str, int] = defaultdict(int)
+            for substrate in reaction.getListOfReactants():
+                species = _name_to_py(substrate.getSpecies())
+                if (ref := substrate.getId()) != "":
+                    dynamic_stoichiometry[species] = ref
+                    self.parameters[ref] = Parameter(0.0, is_constant=False)
+
+                elif species not in self.boundary_species:
+                    factor = substrate.getStoichiometry()
+                    if str(factor) == "nan":
+                        msg = f"Cannot parse stoichiometry: {factor}"
+                        raise ValueError(msg)
+                    parsed_reactants[species] -= factor
+
+            parsed_products: defaultdict[str, int] = defaultdict(int)
+            for product in reaction.getListOfProducts():
+                species = _name_to_py(product.getSpecies())
+                if (ref := product.getId()) != "":
+                    dynamic_stoichiometry[species] = ref
+                    self.parameters[ref] = Parameter(0.0, is_constant=False)
+                elif species not in self.boundary_species:
+                    factor = product.getStoichiometry()
+                    if str(factor) == "nan":
+                        msg = f"Cannot parse stoichiometry: {factor}"
+                        raise ValueError(msg)
+                    parsed_products[species] += factor
+
+            # Combine stoichiometries
+            # Hint: you can't just combine the dictionaries, as you have cases like
+            # S1 + S2 -> 2S2, which have to be combined to S1 -> S2
+            stoichiometries = dict(parsed_reactants)
+            for species, value in parsed_products.items():
+                if species in stoichiometries:
+                    stoichiometries[species] = stoichiometries[species] + value
+                else:
+                    stoichiometries[species] = value
+
+            self.reactions[sbml_id] = Reaction(
+                body=body,
+                stoichiometry=stoichiometries | dynamic_stoichiometry,
+                args=args,
+            )
+
+    def _convert_substance_amount_to_concentration(
+        self,
+    ) -> None:
+        """Convert substance amount to concentration if has_only_substance_units is false.
+
+        The compounds in the test are supplied in mole if has_only_substance_units is false.
+        In that case, the reaction equation has to be reformed like this:
+        k1 * S1 * compartment -> k1 * S1
+        or in other words the species have to be divided by the compartment to get
+        concentration units.
+        """
+        for reaction in self.reactions.values():
+            function_body = reaction.body
+            removed_compartments = set()
+            for arg in reaction.args:
+                # the parsed species part is important to not
+                # introduce conversion on things that aren't species
+                if (species := self.variables.get(arg, None)) is None:
+                    continue
+
+                if not species.has_only_substance_units:
+                    compartment = species.compartment
+                    if compartment is not None:
+                        if self.compartments[compartment].dimensions == 0:
+                            continue
+                        if species.is_concentration:
+                            if compartment not in removed_compartments:
+                                pattern = f"({compartment})" + r"\b"
+                                repl = f"({compartment} / {compartment})"
+                                function_body = re.sub(pattern, repl, function_body)
+                                removed_compartments.add(compartment)
+                        else:
+                            # \b is word boundary
+                            pattern = f"({arg})" + r"\b"
+                            repl = f"({arg} / {compartment})"
+                            function_body = re.sub(pattern, repl, function_body)
+                        if compartment not in reaction.args:
+                            reaction.args.append(compartment)
+
+            # Simplify the function
+            try:
+                reaction.body = str(sympy.parse_expr(function_body))
+            except AttributeError:
+                # E.g. when math.factorial is called
+                # FIXME: do the sympy conversion before?
+                reaction.body = function_body
+
+
+def _handle_fn(name: str, body: str, args: list[str]) -> Callable[..., float]:
+    func_args = ", ".join(args)
+    func_str = "\n".join(
+        [
+            f"def {name}({func_args}):",
+            f"{INDENT}return {body}",
+            "",
+        ]
+    )
+    try:
+        exec(func_str, globals(), None)  # noqa: S102
+    except SyntaxError as e:
+        msg = f"Invalid function definition: {func_str}"
+        raise SyntaxError(msg) from e
+    python_func = globals()[name]
+    python_func.__source__ = func_str
+    return python_func  # type: ignore
+
+
+def _codegen_fn(name: str, body: str, args: list[str]) -> str:
+    func_args = ", ".join(args)
+    return "\n".join(
+        [
+            f"def {name}({func_args}):",
+            f"{INDENT}return {body}",
+            "",
+        ]
+    )
+
+
+def _codegen_initial_assignment(
+    k: str,
+    sbml: Parser,
+    parameters: dict[str, float],
+) -> str:
+    if k in parameters:
+        return f"m.update_parameter('{k}', _init_{k}(*(args[i] for i in {sbml.initial_assignment[k].args})) )"
+
+    ass = sbml.initial_assignment[k]
+    fn = f"_init_{k}(*(args[i] for i in {ass.args}))"
+
+    if (species := sbml.variables.get(k)) is not None:
+        compartment = species.compartment
+        if compartment is not None and (
+            not species.is_concentration
+            or (species.has_only_substance_units and species.is_concentration)
+        ):
+            size = 1 if (c := sbml.compartments.get(compartment)) is None else c.size
+            if size != 0:
+                fn = f"{fn} * {size}"
+    elif k in sbml.compartments:
+        pass
+    else:
+        msg = "Initial assignment targeting unknown type"
+        raise NotImplementedError(msg)
+
+    return f"m.update_variable('{k}', {fn})"
+
+
+def _codgen(name: str, sbml: Parser) -> Path:
+    import itertools as it
+
+    functions = {
+        k: _codegen_fn(k, body=v.body, args=v.args)
+        for k, v in it.chain(
+            sbml.functions.items(),
+            sbml.derived.items(),
+            sbml.reactions.items(),
+            {f"_init_{k}": v for k, v in sbml.initial_assignment.items()}.items(),
+        )
+    }
+
+    parameters = {
+        k: v.value for k, v in sbml.parameters.items() if k not in sbml.derived
+    }
+    variables = {
+        k: v.initial_amount for k, v in sbml.variables.items() if k not in sbml.derived
+    }
+    for k, v in sbml.compartments.items():
+        if k in sbml.derived:
+            continue
+        if v.is_constant:
+            parameters[k] = v.size
+        else:
+            variables[k] = v.size
+
+    derived_str = "\n    ".join(
+        f"m.add_derived('{k}', fn={k}, args={v.args})" for k, v in sbml.derived.items()
+    )
+    rxn_str = "\n    ".join(
+        f"m.add_reaction('{k}', fn={k}, args={rxn.args}, stoichiometry={rxn.stoichiometry})"
+        for k, rxn in sbml.reactions.items()
+    )
+
+    functions_str = "\n\n".join(functions.values())
+
+    parameters_str = f"m.add_parameters({parameters})" if len(parameters) > 0 else ""
+    variables_str = f"m.add_variables({variables})" if len(variables) > 0 else ""
+
+    # Initial assignments
+    initial_assignment_order = _sort_dependencies(
+        available=set(sbml.initial_assignment) ^ set(parameters) ^ set(variables),
+        elements=[(k, set(v.args)) for k, v in sbml.initial_assignment.items()],
+    )
+
+    if len(initial_assignment_order) > 0:
+        initial_assignment_source = "\n    ".join(
+            _codegen_initial_assignment(k, sbml, parameters=parameters)
+            for k in initial_assignment_order
+        )
+    else:
+        initial_assignment_source = ""
+
+    file = f"""
+import math
+
+import numpy as np
+import scipy
+
+from modelbase2 import Model
+
+{functions_str}
+
+def get_model() -> Model:
+    m = Model()
+    {parameters_str}
+    {variables_str}
+    {derived_str}
+    {rxn_str}
+    args = m.get_args()
+    {initial_assignment_source}
+    return m
+"""
+    path = default_tmp_dir(None, remove_old_cache=False) / f"{name}.py"
+    with path.open("w+") as f:
+        f.write(file)
+    return path
+
+
+def import_from_path(module_name: str, file_path: Path) -> Callable[[], Model]:
+    import sys
+    from importlib import util
+
+    spec = unwrap(util.spec_from_file_location(module_name, file_path))
+    module = util.module_from_spec(spec)
+    sys.modules[module_name] = module
+    unwrap(spec.loader).exec_module(module)
+    return module.get_model
+
+
+def valid_filename(value: str) -> str:
+    import re
+    import unicodedata
+
+    value = (
+        unicodedata.normalize("NFKD", value).encode("ascii", "ignore").decode("ascii")
+    )
+    value = re.sub(r"[^\w\s-]", "", value.lower())
+    value = re.sub(r"[-\s]+", "_", value).strip("-_")
+    return f"mb_{value}"
+
+
+def read(file: Path) -> Model:
+    """Import a metabolic model from an SBML file.
+
+    Args:
+        file: Path to the SBML file to import.
+
+    Returns:
+        Model: Imported model instance.
+
+    """
+    name = valid_filename(file.stem)
+    sbml = Parser().parse(file=file)
+
+    model_fn = import_from_path(name, _codgen(name, sbml))
+    return model_fn()