power-openapi-models 0.1.0__tar.gz → 0.3.0__tar.gz

{power_openapi_models-0.1.0/src/power_openapi_models.egg-info → power_openapi_models-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: power-openapi-models
-Version: 0.1.0
+Version: 0.3.0
 Summary: Typed Python models for the Sienna power system data format
 Author: Sienna Platform
 License-Expression: BSD-3-Clause
@@ -22,7 +22,10 @@ Requires-Python: >=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: pydantic>=2.0
+Provides-Extra: source-version
+Requires-Dist: jsonschema>=4; extra == "source-version"
 Provides-Extra: dev
+Requires-Dist: jsonschema>=4; extra == "dev"
 Requires-Dist: ruff>=0.6; extra == "dev"
 Requires-Dist: pytest>=8; extra == "dev"
 Requires-Dist: pyright>=1.1.380; extra == "dev"
@@ -111,6 +114,20 @@ print(UnitSystem.NATURAL_UNITS.value)
 `base_power`. Getting this wrong is the most common source of
 wrong-by-a-factor-of-100 bugs across the ecosystem.
 
+### Timestamps may omit the offset
+
+The schema says RFC 3339, which requires an offset, but producers write
+`2024-01-01T00:00:00`, so all three packages accept it and read it as UTC. An offset is
+kept; a bare date or text that is not a timestamp is rejected. Fields are typed
+`power_openapi_models.timestamps.UtcDatetime`, so they always hold an aware `datetime`.
+
+```python
+from power_openapi_models.infrastructure_core.models import DataSource
+
+source = DataSource(id=1, fields=[], retrieved_at="2024-01-01T00:00:00")
+assert source.retrieved_at.utcoffset().total_seconds() == 0
+```
+
 ### Discriminated unions carry a `.root`
 
 Several types are one of several shapes, chosen by a discriminator field.
@@ -167,9 +184,34 @@ consumer's job.
 ## Round-tripping
 
 `read_document` followed by `write_document` reproduces the input byte for
-byte. Fields the input never carried stay omitted rather than being written
-back as explicit nulls, top-level keys are sorted, and the file ends with a
-newline.
+byte, for a document in canonical form. Optional properties that are absent,
+null or equal to their schema default (an empty `trading_hub_associations`,
+say) are omitted rather than written back, top-level keys are sorted after the
+`schema_version` stamp, and the file ends with a newline.
+
+## Schema versions
+
+Both documents carry a required `schema_version`: the schema release that wrote
+them. `read_document` and `read_portfolio_document` check it on the raw JSON,
+before decoding, and raise `SchemaVersionError` (with `outcome`, `reader` and
+`document` attributes) unless the document is from this package's schema version
+or an older one on the same compatibility line. `check_schema_version(raw)` runs
+the same rule on a parsed dict and returns the outcome without raising (a
+non-object root raises pydantic's `ValidationError`).
+
+```python
+from power_openapi_models.document import get_source_schema_version, read_document
+
+doc = read_document("../fixtures/case14_operations.NATURAL_UNITS.json")
+print(get_source_schema_version(doc))
+```
+
+`write_document(doc, path)` stamps this package's version, so reading then writing
+upgrades a document; `upgrade_document(src, dst)` does exactly that. To keep the
+version a document was read at, pass `schema_version="source"`. When that version
+is older, the encoded document is validated against that version's strict schema
+first, and the write fails listing every path that version does not know; nothing
+is dropped. That check needs `pip install "power-openapi-models[source-version]"`.
 
 ## Worked example
 

{power_openapi_models-0.1.0 → power_openapi_models-0.3.0}/README.md

@@ -78,6 +78,20 @@ print(UnitSystem.NATURAL_UNITS.value)
 `base_power`. Getting this wrong is the most common source of
 wrong-by-a-factor-of-100 bugs across the ecosystem.
 
+### Timestamps may omit the offset
+
+The schema says RFC 3339, which requires an offset, but producers write
+`2024-01-01T00:00:00`, so all three packages accept it and read it as UTC. An offset is
+kept; a bare date or text that is not a timestamp is rejected. Fields are typed
+`power_openapi_models.timestamps.UtcDatetime`, so they always hold an aware `datetime`.
+
+```python
+from power_openapi_models.infrastructure_core.models import DataSource
+
+source = DataSource(id=1, fields=[], retrieved_at="2024-01-01T00:00:00")
+assert source.retrieved_at.utcoffset().total_seconds() == 0
+```
+
 ### Discriminated unions carry a `.root`
 
 Several types are one of several shapes, chosen by a discriminator field.
@@ -134,9 +148,34 @@ consumer's job.
 ## Round-tripping
 
 `read_document` followed by `write_document` reproduces the input byte for
-byte. Fields the input never carried stay omitted rather than being written
-back as explicit nulls, top-level keys are sorted, and the file ends with a
-newline.
+byte, for a document in canonical form. Optional properties that are absent,
+null or equal to their schema default (an empty `trading_hub_associations`,
+say) are omitted rather than written back, top-level keys are sorted after the
+`schema_version` stamp, and the file ends with a newline.
+
+## Schema versions
+
+Both documents carry a required `schema_version`: the schema release that wrote
+them. `read_document` and `read_portfolio_document` check it on the raw JSON,
+before decoding, and raise `SchemaVersionError` (with `outcome`, `reader` and
+`document` attributes) unless the document is from this package's schema version
+or an older one on the same compatibility line. `check_schema_version(raw)` runs
+the same rule on a parsed dict and returns the outcome without raising (a
+non-object root raises pydantic's `ValidationError`).
+
+```python
+from power_openapi_models.document import get_source_schema_version, read_document
+
+doc = read_document("../fixtures/case14_operations.NATURAL_UNITS.json")
+print(get_source_schema_version(doc))
+```
+
+`write_document(doc, path)` stamps this package's version, so reading then writing
+upgrades a document; `upgrade_document(src, dst)` does exactly that. To keep the
+version a document was read at, pass `schema_version="source"`. When that version
+is older, the encoded document is validated against that version's strict schema
+first, and the write fails listing every path that version does not know; nothing
+is dropped. That check needs `pip install "power-openapi-models[source-version]"`.
 
 ## Worked example
 

{power_openapi_models-0.1.0 → power_openapi_models-0.3.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "power-openapi-models"
-version = "0.1.0"
+version = "0.3.0"
 description = "Typed Python models for the Sienna power system data format"
 readme = "README.md"
 license = "BSD-3-Clause"
@@ -41,7 +41,9 @@ Issues = "https://github.com/Sienna-Platform/power-openapi-models/issues"
 Changelog = "https://github.com/Sienna-Platform/power-openapi-models/blob/main/CHANGELOG.md"
 
 [project.optional-dependencies]
+source-version = ["jsonschema>=4"]
 dev = [
+    "jsonschema>=4",
     "ruff>=0.6",
     "pytest>=8",
     "pyright>=1.1.380",
@@ -54,7 +56,7 @@ dev = [
 where = ["src"]
 
 [tool.setuptools.package-data]
-power_openapi_models = ["py.typed", "_schema_version.txt"]
+power_openapi_models = ["py.typed", "_schema_version.txt", "bundles/*/*.json"]
 
 [tool.ruff]
 line-length = 100

{power_openapi_models-0.1.0 → power_openapi_models-0.3.0}/src/power_openapi_models/__init__.py

@@ -17,6 +17,7 @@ from power_openapi_models import (
     operations,
     timeseries,
 )
+from power_openapi_models._versioning import SchemaVersionError, check_schema_version
 
 try:
     __version__ = _version("power-openapi-models")
@@ -47,6 +48,8 @@ __all__ = [
     "investments",
     "operations",
     "timeseries",
+    "SchemaVersionError",
+    "check_schema_version",
     "__version__",
     "__schema_version__",
 ]

power_openapi_models-0.3.0/src/power_openapi_models/_schema_version.txt

@@ -0,0 +1 @@
+v0.3.0

power_openapi_models-0.3.0/src/power_openapi_models/_versioning.py

@@ -0,0 +1,195 @@
+"""Hand-written (NOT generated): the schema-version reader rule and the strict-bundle
+check behind `write_document(..., schema_version="source")`.
+
+The rule, its outcomes and the canonical message texts are defined once, in
+SiennaSchemas' `docs/VERSIONING.md`; `scripts/schema_version.py` there is the reference
+implementation. `check_schema_version` must agree with `tests/fixtures/versioning/cases.json`.
+"""
+
+from __future__ import annotations
+
+import json
+import re
+from pathlib import Path
+from typing import Literal
+
+from pydantic import ValidationError
+
+Outcome = Literal["missing", "malformed", "incompatible", "newer", "upgradable", "current"]
+
+READER_VERSION = (
+    (Path(__file__).parent / "_schema_version.txt").read_text().strip().removeprefix("v")
+)
+BUNDLE_DIR = Path(__file__).parent / "bundles"
+
+_PATTERN = re.compile(r"(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)(-[0-9A-Za-z.-]+)?")
+
+
+class SchemaVersionError(ValueError):
+    """A document whose `schema_version` this reader cannot read."""
+
+    def __init__(self, outcome: Outcome, reader: str, document: object, message: str):
+        super().__init__(message)
+        self.outcome = outcome
+        self.reader = reader
+        self.document = document
+
+
+def _parse(version: object) -> tuple[tuple[int, str], ...] | None:
+    """Components as (digit count, digits): the pattern forbids leading zeros, so this
+    orders like the integers without a size limit."""
+    if not isinstance(version, str):
+        return None
+    m = _PATTERN.fullmatch(version)
+    if m is None:
+        return None
+    pre = (0, m[4][1:]) if m[4] else (0, "")
+    return (*((len(m[i]), m[i]) for i in (1, 2, 3)), pre)
+
+
+def _prerelease(version) -> bool:
+    return bool(version[3][1])
+
+
+def _line(version) -> tuple[tuple[int, str], ...]:
+    major, minor = version[:2]
+    return (major, minor) if major[1] == "0" else (major,)
+
+
+def _line_name(version) -> str:
+    return ".".join(digits for _, digits in _line(version))
+
+
+def _classify(raw: object, reader: str) -> Outcome:
+    if not isinstance(raw, dict):
+        raise ValidationError.from_exception_data(
+            "document", [{"type": "dict_type", "loc": (), "input": raw}]
+        )
+    if "schema_version" not in raw:
+        return "missing"
+    d = _parse(raw["schema_version"])
+    if d is None:
+        return "malformed"
+    r = _parse(reader)
+    if r is None:
+        raise ValueError(f"reader schema version {reader!r} is not a valid version")
+    if (_prerelease(d) or _prerelease(r)) and d != r:
+        return "incompatible"
+    if _line(d) != _line(r):
+        return "incompatible"
+    if d[:3] > r[:3]:
+        return "newer"
+    if d[:3] < r[:3]:
+        return "upgradable"
+    return "current"
+
+
+def _message(outcome: Outcome, reader: str, document: object) -> str:
+    if outcome == "missing":
+        return (
+            "document has no schema_version: it predates versioning; re-export it with a "
+            "current producer (psy5 bundles: PowerSystemsUpdater)"
+        )
+    if outcome == "malformed":
+        encoded = json.dumps(document, separators=(",", ":"), ensure_ascii=False)
+        return f"document schema_version {encoded} is not a valid version"
+    d = _parse(document)
+    r = _parse(reader)
+    assert d is not None and r is not None
+    if outcome == "incompatible":
+        if _prerelease(d) or _prerelease(r):
+            return (
+                f"document written by schema {document} cannot be read by schema {reader}: "
+                "dev builds read only their own output"
+            )
+        return (
+            f"document written by schema {document} (line {_line_name(d)}) cannot be read by "
+            f"schema {reader} (line {_line_name(r)}): documents do not cross compatibility "
+            "lines; migrating between lines is a separate upgrade tool's job "
+            "(psy5 bundles: PowerSystemsUpdater)"
+        )
+    return (
+        f"document written by schema {document}; this reader understands up to {reader}; "
+        f"update the model package to one built from >= {document}"
+    )
+
+
+def _check(raw: object, reader: str) -> Outcome:
+    outcome = _classify(raw, reader)
+    if outcome in ("upgradable", "current"):
+        return outcome
+    document = raw.get("schema_version")
+    raise SchemaVersionError(outcome, reader, document, _message(outcome, reader, document))
+
+
+def check_schema_version(raw: object) -> Outcome:
+    """Classify the raw parsed JSON of a document against this package's schema version.
+
+    Pure: returns the outcome and never raises `SchemaVersionError` (the read path does).
+    A non-object root is a format error and raises pydantic's `ValidationError`, as
+    reading it does. Run before decoding, so a newer document reports its version
+    rather than an unknown key.
+    """
+    return _classify(raw, READER_VERSION)
+
+
+def _covered(error, branch: int, root: dict) -> int:
+    """How many instance keys the `branch` of an anyOf/oneOf `error` declares."""
+    sub = error.schema[error.validator][branch]
+    while "$ref" in sub:
+        node = root
+        for part in sub["$ref"].removeprefix("#/").split("/"):
+            node = node[part]
+        sub = node
+    if not isinstance(error.instance, dict):
+        return 0
+    return sum(1 for key in error.instance if key in sub.get("properties", {}))
+
+
+def _leaves(error, root: dict):
+    """The offending (path, message) pairs under one validation error."""
+    if error.validator in ("anyOf", "oneOf") and error.context:
+        branches: dict[int, list] = {}
+        for e in error.context:
+            branches.setdefault(e.schema_path[0], []).extend(_leaves(e, root))
+        pick = min(
+            branches,
+            key=lambda i: (len(branches[i]), -_covered(error, i, root)),
+        )
+        yield from branches[pick]
+        return
+    path = "/" + "/".join(str(p) for p in error.absolute_path)
+    if error.validator == "additionalProperties" and isinstance(error.instance, dict):
+        known = error.schema.get("properties", {})
+        for key in error.instance:
+            if key not in known:
+                yield f"{path.rstrip('/')}/{key}", "is not a property of this schema version"
+        return
+    yield path, error.message
+
+
+def validate_source(tree: dict, document: str, bundle_name: str, bundle_dir: Path) -> None:
+    """Validate the encoded `tree` against the strict bundle of schema `document`.
+
+    Raises one `ValueError` listing every offending path.
+    """
+    try:
+        from jsonschema import Draft7Validator
+    except ImportError as e:
+        raise ImportError(
+            f"writing at the source schema version {document} needs a validator; "
+            "install power-openapi-models[source-version]"
+        ) from e
+    bundle = Path(bundle_dir) / document / f"{bundle_name}.json"
+    if not bundle.is_file():
+        raise FileNotFoundError(f"strict bundle for schema {document} not found at {bundle}")
+    root = json.loads(bundle.read_text())
+    found = sorted(
+        {leaf for e in Draft7Validator(root).iter_errors(tree) for leaf in _leaves(e, root)}
+    )
+    if found:
+        lines = "\n".join(f"  {path}: {message}" for path, message in found)
+        raise ValueError(
+            f'document does not fit schema {document}; save with schema_version="current" '
+            f"to write the current version:\n{lines}"
+        )