feat(detection-core): build exact family catalog resolution

This commit is contained in:
Marius Ciepluch
2026-10-31 14:32:37 +01:00
parent 93404c5e1e
commit 741338f207
3 changed files with 438 additions and 1 deletions
+1 -1
View File
@@ -80,7 +80,7 @@
- Refuse undeclared fields, roles, decisions, reasons, or determinate outcomes without their minimum evidence. - Refuse undeclared fields, roles, decisions, reasons, or determinate outcomes without their minimum evidence.
- Done when the family registration matrix accepts only declared combinations and reports unsupported scope without cross-family fallback. - Done when the family registration matrix accepts only declared combinations and reports unsupported scope without cross-family fallback.
- _Requirements: 1.3, 2.1, 2.4, 3.1, 3.2, 3.3, 3.4, 4.4, 9.1_ - _Requirements: 1.3, 2.1, 2.4, 3.1, 3.2, 3.3, 3.4, 4.4, 9.1_
- [ ] 3.2 Build exact family catalog resolution and capability declarations - [x] 3.2 Build exact family catalog resolution and capability declarations
- Re-derive every contract and catalog digest and require canonical ordering and unique family/version/schema identities. - Re-derive every contract and catalog digest and require canonical ordering and unique family/version/schema identities.
- Resolve a contract only when family, digest, and ecosystem identify exactly one entry. - Resolve a contract only when family, digest, and ecosystem identify exactly one entry.
- Validate that each family capability names its contract, readiness fixtures, development population, and admission boundary. - Validate that each family capability names its contract, readiness fixtures, development population, and admission boundary.
+180
View File
@@ -0,0 +1,180 @@
"""Exact immutable family catalogs and capability declarations."""
from __future__ import annotations
import re
from dataclasses import dataclass
from openultrasast.core import canonical
from openultrasast.core.family import FamilyContract, family_contract_digest, validate_contract
FAMILY_CATALOG_SCHEMA = "openultrasast.detection-core.family-catalog.v2"
_DIGEST = re.compile(r"[0-9a-f]{64}")
class FamilyCatalogError(ValueError):
"""A family catalog violates its canonical closed-set contract."""
class FamilyCatalogResolutionError(FamilyCatalogError, LookupError):
"""No exact family, digest, and ecosystem catalog entry exists."""
class FamilyCapabilityError(FamilyCatalogError):
"""A family capability declaration is incomplete or substituted."""
def _text(value: object, field: str, error: type[FamilyCatalogError] = FamilyCatalogError) -> None:
if not isinstance(value, str) or not value:
raise error(f"{field} must be nonempty text")
def _digest(value: object, field: str, error: type[FamilyCatalogError] = FamilyCatalogError) -> None:
if not isinstance(value, str) or _DIGEST.fullmatch(value) is None:
raise error(f"{field} must be a full lowercase SHA-256 digest")
@dataclass(frozen=True)
class FamilyCatalogEntry:
contract: FamilyContract
contract_digest: str
def __post_init__(self) -> None:
if not isinstance(self.contract, FamilyContract):
raise FamilyCatalogError("entry contract must be a FamilyContract")
_digest(self.contract_digest, "contract digest")
validate_contract(self.contract)
if self.contract_digest != family_contract_digest(self.contract):
raise FamilyCatalogError("contract digest does not match contract content")
def catalog_entry_key(entry: FamilyCatalogEntry) -> tuple[str, str, str, str]:
if not isinstance(entry, FamilyCatalogEntry):
raise FamilyCatalogError("catalog entry must be a FamilyCatalogEntry")
contract = entry.contract
return (contract.family_id, contract.family_version, contract.schema_version, entry.contract_digest)
def _entry_value(entry: FamilyCatalogEntry) -> canonical.JsonValue:
contract = entry.contract
return {
"family_id": contract.family_id,
"family_version": contract.family_version,
"schema_version": contract.schema_version,
"ecosystems": contract.ecosystems,
"contract_digest": entry.contract_digest,
}
def derive_catalog_digest(entries: tuple[FamilyCatalogEntry, ...]) -> str:
if not isinstance(entries, tuple) or not entries:
raise FamilyCatalogError("entries must be a nonempty tuple")
if not all(isinstance(entry, FamilyCatalogEntry) for entry in entries):
raise FamilyCatalogError("entries must contain FamilyCatalogEntry values")
return canonical.content_digest({"schema_version": FAMILY_CATALOG_SCHEMA, "entries": tuple(_entry_value(entry) for entry in entries)})
@dataclass(frozen=True)
class FamilyCatalog:
entries: tuple[FamilyCatalogEntry, ...]
catalog_digest: str
def __post_init__(self) -> None:
validate_catalog(self)
def _validate_unique_entries(entries: tuple[FamilyCatalogEntry, ...]) -> None:
identities = tuple((item.contract.family_id, item.contract.family_version, item.contract.schema_version) for item in entries)
if len(set(identities)) != len(identities):
raise FamilyCatalogError("duplicate family contract identity")
digests = tuple(item.contract_digest for item in entries)
if len(set(digests)) != len(digests):
raise FamilyCatalogError("duplicate contract digest")
def validate_catalog(value: FamilyCatalog) -> None:
if not isinstance(value, FamilyCatalog):
raise FamilyCatalogError("catalog must be a FamilyCatalog")
if not isinstance(value.entries, tuple) or not value.entries:
raise FamilyCatalogError("entries must be a nonempty tuple")
if not all(isinstance(entry, FamilyCatalogEntry) for entry in value.entries):
raise FamilyCatalogError("entries must contain FamilyCatalogEntry values")
_digest(value.catalog_digest, "catalog digest")
if value.entries != tuple(sorted(value.entries, key=catalog_entry_key)):
raise FamilyCatalogError("catalog entries must be in canonical order")
_validate_unique_entries(value.entries)
for entry in value.entries:
validate_contract(entry.contract)
if entry.contract_digest != family_contract_digest(entry.contract):
raise FamilyCatalogError("contract digest does not match contract content")
if value.catalog_digest != derive_catalog_digest(value.entries):
raise FamilyCatalogError("catalog digest does not match canonical entries")
def resolve_contract(
value: FamilyCatalog,
family_id: str,
contract_digest: str,
ecosystem: str,
) -> FamilyContract:
validate_catalog(value)
_text(family_id, "family_id", FamilyCatalogResolutionError)
_digest(contract_digest, "contract digest", FamilyCatalogResolutionError)
_text(ecosystem, "ecosystem", FamilyCatalogResolutionError)
matches = tuple(
entry.contract
for entry in value.entries
if entry.contract.family_id == family_id and entry.contract_digest == contract_digest and ecosystem in entry.contract.ecosystems
)
if len(matches) != 1:
raise FamilyCatalogResolutionError("family, digest, and ecosystem must resolve one exact contract")
return matches[0]
@dataclass(frozen=True)
class FamilyCapabilityDeclaration:
arm_id: str
family_contract_digest: str
readiness_fixture_digest: str
development_population_digest: str
admission_boundary_digest: str
def __post_init__(self) -> None:
_validate_capability_fields(self)
def _validate_capability_fields(declaration: FamilyCapabilityDeclaration) -> None:
_text(declaration.arm_id, "arm_id", FamilyCapabilityError)
for value, field in (
(declaration.family_contract_digest, "family contract digest"),
(declaration.readiness_fixture_digest, "readiness fixture digest"),
(declaration.development_population_digest, "development population digest"),
(declaration.admission_boundary_digest, "admission boundary digest"),
):
_digest(value, field, FamilyCapabilityError)
def validate_capability(declaration: FamilyCapabilityDeclaration, value: FamilyCatalog) -> None:
if not isinstance(declaration, FamilyCapabilityDeclaration):
raise FamilyCapabilityError("declaration must be a FamilyCapabilityDeclaration")
_validate_capability_fields(declaration)
validate_catalog(value)
matches = tuple(entry for entry in value.entries if entry.contract_digest == declaration.family_contract_digest)
if len(matches) != 1:
raise FamilyCapabilityError("capability contract digest must identify one exact catalog entry")
__all__ = [
"FAMILY_CATALOG_SCHEMA",
"FamilyCapabilityDeclaration",
"FamilyCapabilityError",
"FamilyCatalog",
"FamilyCatalogEntry",
"FamilyCatalogError",
"FamilyCatalogResolutionError",
"catalog_entry_key",
"derive_catalog_digest",
"resolve_contract",
"validate_capability",
"validate_catalog",
]
+257
View File
@@ -0,0 +1,257 @@
"""Exact family-catalog resolution and capability declaration contracts."""
from dataclasses import replace
from typing import cast
import pytest
import openultrasast.core.canonical as canonical
import openultrasast.core.evidence as evidence
import openultrasast.core.family as family
import openultrasast.core.family_catalog as catalog
import openultrasast.core.semantics as semantics
def _digest(value: str) -> str:
return canonical.content_digest(value)
def _contract(
family_id: str = "obligation",
family_version: str = "1.0.0",
schema_version: str = evidence.EVIDENCE_PACKET_SCHEMA,
ecosystems: tuple[str, ...] = ("python",),
) -> family.FamilyContract:
role = family.RoleSpec(
"operation",
(family.FieldSpec("operation", semantics.FieldKind.TEXT, True),),
(semantics.RelationKind.OPERATION_EFFECT,),
family.RelationOriginPolicy.PRODUCT_OR_REVIEWED_FACT,
)
return family.FamilyContract(
family_id,
family_version,
schema_version,
ecosystems,
family.ObjectSpec(
"operation",
(family.FieldSpec("operation", semantics.FieldKind.TEXT, True),),
("operation",),
),
(role,),
(
family.DecisionRule(
"unsafe",
evidence.NeutralOutcome.VIOLATION,
("operation",),
(semantics.RelationStatus.COMPLETE,),
),
family.DecisionRule(
"unknown",
evidence.NeutralOutcome.INDETERMINATE,
("operation",),
(
semantics.RelationStatus.COMPLETE,
semantics.RelationStatus.INCOMPLETE,
semantics.RelationStatus.UNRESOLVED,
),
),
),
(family.ResidualQuestion("operation-known", ("operation",), ("unsafe", "unknown")),),
("missing_operation", "unsupported_scope"),
)
def _entry(contract: family.FamilyContract) -> catalog.FamilyCatalogEntry:
return catalog.FamilyCatalogEntry(contract, family.family_contract_digest(contract))
def _catalog(*contracts: family.FamilyContract) -> catalog.FamilyCatalog:
entries = tuple(sorted((_entry(contract) for contract in contracts), key=catalog.catalog_entry_key))
return catalog.FamilyCatalog(entries, catalog.derive_catalog_digest(entries))
def _capability(contract: family.FamilyContract) -> catalog.FamilyCapabilityDeclaration:
return catalog.FamilyCapabilityDeclaration(
"obligation-arm",
family.family_contract_digest(contract),
_digest("readiness fixtures"),
_digest("development population"),
_digest("admission boundary"),
)
def test_catalog_rederives_contract_and_catalog_digests() -> None:
contract = _contract()
value = _catalog(contract)
catalog.validate_catalog(value)
assert value.entries[0].contract_digest == family.family_contract_digest(contract)
assert value.catalog_digest == catalog.derive_catalog_digest(value.entries)
def test_catalog_requires_canonical_entry_order() -> None:
first = _contract("alpha")
second = _contract("zeta")
entries = (_entry(second), _entry(first))
with pytest.raises(catalog.FamilyCatalogError, match="canonical order"):
catalog.FamilyCatalog(entries, catalog.derive_catalog_digest(tuple(reversed(entries))))
def test_catalog_refuses_duplicate_contract_identity_and_digest() -> None:
contract = _contract()
entry = _entry(contract)
with pytest.raises(catalog.FamilyCatalogError, match="duplicate family contract identity"):
catalog.FamilyCatalog((entry, entry), _digest("catalog"))
duplicate_digest = _entry(_contract("other"))
object.__setattr__(duplicate_digest, "contract_digest", entry.contract_digest)
entries = tuple(sorted((entry, duplicate_digest), key=catalog.catalog_entry_key))
with pytest.raises(catalog.FamilyCatalogError, match="duplicate contract digest"):
catalog.FamilyCatalog(entries, _digest("catalog"))
def test_catalog_refuses_substituted_contract_and_catalog_digest() -> None:
contract = _contract()
with pytest.raises(catalog.FamilyCatalogError, match="contract digest"):
catalog.FamilyCatalogEntry(contract, _digest("substituted"))
value = _catalog(contract)
with pytest.raises(catalog.FamilyCatalogError, match="catalog digest"):
catalog.FamilyCatalog(value.entries, _digest("substituted catalog"))
def test_catalog_identity_includes_family_version_and_schema() -> None:
contracts = (
_contract(family_version="1.0.0"),
_contract(family_version="2.0.0"),
_contract(family_version="3.0.0", schema_version="evidence-v3"),
)
value = _catalog(*contracts)
assert len(value.entries) == 3
catalog.validate_catalog(value)
def test_catalog_collections_are_immutable_tuples() -> None:
value = _catalog(_contract())
with pytest.raises(catalog.FamilyCatalogError, match="entries"):
replace(value, entries=cast(tuple[catalog.FamilyCatalogEntry, ...], list(value.entries)))
def test_resolve_contract_requires_exact_family_digest_and_ecosystem() -> None:
python = _contract("obligation", ecosystems=("python",))
java = _contract("taint", ecosystems=("java",))
value = _catalog(python, java)
assert catalog.resolve_contract(value, "obligation", family.family_contract_digest(python), "python") is python
for family_id, digest, ecosystem in (
("missing", family.family_contract_digest(python), "python"),
("obligation", family.family_contract_digest(java), "python"),
("obligation", family.family_contract_digest(python), "java"),
):
with pytest.raises(catalog.FamilyCatalogResolutionError, match="exact"):
catalog.resolve_contract(value, family_id, digest, ecosystem)
def test_resolution_never_falls_back_to_another_version() -> None:
first = _contract(family_version="1.0.0")
second = _contract(family_version="2.0.0")
value = _catalog(first, second)
assert (
catalog.resolve_contract(
value,
first.family_id,
family.family_contract_digest(second),
"python",
)
is second
)
with pytest.raises(catalog.FamilyCatalogResolutionError, match="exact"):
catalog.resolve_contract(value, first.family_id, _digest("unknown version"), "python")
def test_capability_names_exact_catalog_contract_and_boundaries() -> None:
contract = _contract()
value = _catalog(contract)
capability = _capability(contract)
catalog.validate_capability(capability, value)
assert capability.family_contract_digest == value.entries[0].contract_digest
@pytest.mark.parametrize(
"field",
(
"family_contract_digest",
"readiness_fixture_digest",
"development_population_digest",
"admission_boundary_digest",
),
)
@pytest.mark.parametrize("bad", ("", "not-a-digest", "A" * 64, None))
def test_capability_refuses_every_missing_or_corrupt_digest(field: str, bad: object) -> None:
capability = _capability(_contract())
with pytest.raises((catalog.FamilyCapabilityError, canonical.DigestFormatError), match="digest"):
_replace_capability_digest(capability, field, cast(str, bad))
def _replace_capability_digest(
capability: catalog.FamilyCapabilityDeclaration,
field: str,
value: str,
) -> catalog.FamilyCapabilityDeclaration:
if field == "family_contract_digest":
return replace(capability, family_contract_digest=value)
if field == "readiness_fixture_digest":
return replace(capability, readiness_fixture_digest=value)
if field == "development_population_digest":
return replace(capability, development_population_digest=value)
return replace(capability, admission_boundary_digest=value)
@pytest.mark.parametrize("bad", ("", 7, None))
def test_capability_requires_nonempty_text_arm_id(bad: object) -> None:
capability = _capability(_contract())
with pytest.raises(catalog.FamilyCapabilityError, match="arm_id"):
replace(capability, arm_id=cast(str, bad))
def test_capability_refuses_contract_substitution() -> None:
contract = _contract()
other = _contract("other")
value = _catalog(contract, other)
capability = _capability(contract)
catalog.validate_capability(capability, value)
substituted = replace(capability, family_contract_digest=_digest("not in catalog"))
with pytest.raises(catalog.FamilyCapabilityError, match="catalog"):
catalog.validate_capability(substituted, value)
@pytest.mark.parametrize("bad", ("", 7, None))
def test_capability_validator_rechecks_tampered_arm_id(bad: object) -> None:
contract = _contract()
value = _catalog(contract)
capability = _capability(contract)
object.__setattr__(capability, "arm_id", bad)
with pytest.raises(catalog.FamilyCapabilityError, match="arm_id"):
catalog.validate_capability(capability, value)
@pytest.mark.parametrize(
"field",
(
"family_contract_digest",
"readiness_fixture_digest",
"development_population_digest",
"admission_boundary_digest",
),
)
@pytest.mark.parametrize("bad", ("", 7, None, "not-a-digest", "A" * 64))
def test_capability_validator_rechecks_every_tampered_digest(field: str, bad: object) -> None:
contract = _contract()
value = _catalog(contract)
capability = _capability(contract)
object.__setattr__(capability, field, bad)
with pytest.raises(catalog.FamilyCapabilityError, match="digest"):
catalog.validate_capability(capability, value)
def test_catalog_digest_covers_contract_identity_and_content() -> None:
contract = _contract()
changed = replace(contract, indeterminate_reasons=contract.indeterminate_reasons + ("new_reason",))
assert _catalog(contract).catalog_digest != _catalog(changed).catalog_digest