open-standards/jouleclaw/jouleclaw-rs/crates/jouleclaw-hydra-client/clients/python/jclaw_hydra/shacl.py

"""Minimal SHACL Core validator — Python mirror of jouleclaw-shacl."""

from __future__ import annotations

import re
from dataclasses import dataclass, field
from enum import Enum
from typing import Any, Dict, List, Optional


JC_NS = "https://jouleclaw.transaction.science/vocab/v1#"


class NodeKind(str, Enum):
    IRI = "IRI"
    LITERAL = "Literal"
    BLANK_NODE_OR_IRI = "BlankNodeOrIRI"


class Datatype(str, Enum):
    STRING = "string"
    INTEGER = "integer"
    NON_NEGATIVE_INTEGER = "non_negative_integer"
    DOUBLE = "double"
    BOOLEAN = "boolean"
    DATETIME = "datetime"


@dataclass
class Violation:
    shape: str
    path: str
    message: str


@dataclass
class ValidationReport:
    conforms: bool
    violations: List[Violation] = field(default_factory=list)

    def to_dict(self) -> Dict[str, Any]:
        return {
            "conforms": self.conforms,
            "violations": [
                {"shape": v.shape, "path": v.path, "message": v.message}
                for v in self.violations
            ],
        }


@dataclass
class PropertyShape:
    path: str
    min_count: Optional[int] = None
    max_count: Optional[int] = None
    datatype: Optional[Datatype] = None
    node_kind: Optional[NodeKind] = None
    in_values: Optional[List[Any]] = None
    pattern: Optional[str] = None
    min_inclusive: Optional[float] = None
    max_inclusive: Optional[float] = None
    node: Optional[str] = None


@dataclass
class NodeShape:
    target_class: str
    properties: List[PropertyShape] = field(default_factory=list)


class ShapeGraph:
    def __init__(self) -> None:
        self._shapes: Dict[str, NodeShape] = {}

    def add(self, shape: NodeShape) -> "ShapeGraph":
        self._shapes[shape.target_class] = shape
        return self

    def get(self, target_class: str) -> Optional[NodeShape]:
        return self._shapes.get(target_class)

    def __len__(self) -> int:
        return len(self._shapes)


def validate(graph: ShapeGraph, target_class: str, value: Any) -> ValidationReport:
    shape = graph.get(target_class)
    if shape is None:
        raise KeyError(f"no shape registered for {target_class}")
    violations: List[Violation] = []
    _validate_against(shape, value, graph, violations)
    return ValidationReport(conforms=not violations, violations=violations)


def _validate_against(
    shape: NodeShape,
    value: Any,
    graph: ShapeGraph,
    out: List[Violation],
) -> None:
    if not isinstance(value, dict):
        out.append(Violation(shape.target_class, "", f"expected object, got {type(value).__name__}"))
        return
    for prop in shape.properties:
        present = prop.path in value and value[prop.path] is not None
        if not present:
            if (prop.min_count or 0) > 0:
                out.append(
                    Violation(
                        shape.target_class,
                        prop.path,
                        f"missing required property `{prop.path}`",
                    )
                )
            continue
        _check_property(prop, value[prop.path], shape, graph, out)


def _check_property(
    prop: PropertyShape,
    value: Any,
    parent: NodeShape,
    graph: ShapeGraph,
    out: List[Violation],
) -> None:
    if isinstance(value, list):
        n = len(value)
        if prop.min_count is not None and n < prop.min_count:
            out.append(_v(parent, prop, f"min_count {prop.min_count} not met ({n} present)"))
        if prop.max_count is not None and n > prop.max_count:
            out.append(_v(parent, prop, f"max_count {prop.max_count} exceeded ({n} present)"))
        for item in value:
            _check_value(prop, item, parent, graph, out)
        return
    _check_value(prop, value, parent, graph, out)


def _check_value(
    prop: PropertyShape,
    value: Any,
    parent: NodeShape,
    graph: ShapeGraph,
    out: List[Violation],
) -> None:
    # node_kind
    if prop.node_kind is not None:
        if prop.node_kind == NodeKind.IRI and not isinstance(value, str):
            out.append(_v(parent, prop, f"node_kind IRI not satisfied"))
        elif prop.node_kind == NodeKind.LITERAL and not isinstance(value, (str, int, float, bool)):
            out.append(_v(parent, prop, f"node_kind Literal not satisfied"))
        elif prop.node_kind == NodeKind.BLANK_NODE_OR_IRI and not isinstance(value, (dict, str)):
            out.append(_v(parent, prop, f"node_kind BlankNodeOrIRI not satisfied"))
    # datatype
    if prop.datatype is not None and not _matches_datatype(prop.datatype, value):
        out.append(_v(parent, prop, f"datatype {prop.datatype.value} not satisfied"))
    # in
    if prop.in_values is not None and value not in prop.in_values:
        out.append(_v(parent, prop, f"value not in allowed set ({len(prop.in_values)} options)"))
    # pattern
    if prop.pattern is not None and isinstance(value, str):
        if not re.search(prop.pattern, value):
            out.append(_v(parent, prop, f"pattern `{prop.pattern}` not matched"))
    # range
    if isinstance(value, (int, float)) and not isinstance(value, bool):
        f = float(value)
        if prop.min_inclusive is not None and f < prop.min_inclusive:
            out.append(_v(parent, prop, f"min_inclusive {prop.min_inclusive} not met ({f})"))
        if prop.max_inclusive is not None and f > prop.max_inclusive:
            out.append(_v(parent, prop, f"max_inclusive {prop.max_inclusive} exceeded ({f})"))
    # node delegation
    if prop.node is not None:
        nested = graph.get(prop.node)
        if nested is None:
            out.append(_v(parent, prop, f"nested shape `{prop.node}` not in graph"))
        else:
            _validate_against(nested, value, graph, out)


def _matches_datatype(dt: Datatype, value: Any) -> bool:
    if dt == Datatype.STRING:
        return isinstance(value, str)
    if dt == Datatype.INTEGER:
        return isinstance(value, int) and not isinstance(value, bool)
    if dt == Datatype.NON_NEGATIVE_INTEGER:
        return isinstance(value, int) and not isinstance(value, bool) and value >= 0
    if dt == Datatype.DOUBLE:
        return isinstance(value, (int, float)) and not isinstance(value, bool)
    if dt == Datatype.BOOLEAN:
        return isinstance(value, bool)
    if dt == Datatype.DATETIME:
        if not isinstance(value, str) or len(value) < 19:
            return False
        return _looks_like_rfc3339(value)
    return False


def _looks_like_rfc3339(s: str) -> bool:
    # YYYY-MM-DDThh:mm:ss…  the same shape jouleclaw-shacl accepts
    return bool(
        re.match(
            r"^\d{4}-\d{2}-\d{2}[T ]\d{2}:\d{2}:\d{2}",
            s,
        )
    )


def _v(parent: NodeShape, prop: PropertyShape, message: str) -> Violation:
    return Violation(shape=parent.target_class, path=prop.path, message=message)