| 1 |
"""Versioned frontmatter validation — run migrations before Pydantic. |
| 2 |
|
| 3 |
The plain `DlmFrontmatter.model_validate(raw)` path refuses unknown |
| 4 |
keys (`extra="forbid"`) and would therefore reject *any* v2+ document |
| 5 |
with a field added after v1. That turns "forgot to run `dlm migrate`" |
| 6 |
into a confusing `SchemaValidationError` instead of an actionable |
| 7 |
version-drift message. |
| 8 |
|
| 9 |
This module threads the raw dict through the migration registry first |
| 10 |
(bringing a v1 dict up to `CURRENT_SCHEMA_VERSION`) and only then hands |
| 11 |
it to Pydantic. Older-but-runnable documents parse cleanly; newer |
| 12 |
documents raise a typed `DlmVersionError` pointing at the dlm upgrade. |
| 13 |
|
| 14 |
One entry point: |
| 15 |
|
| 16 |
validate_versioned(raw: dict, *, path: Path | None) -> DlmFrontmatter |
| 17 |
|
| 18 |
Callers: `dlm.doc.parser._validate_frontmatter` (post-YAML, pre-Pydantic). |
| 19 |
""" |
| 20 |
|
| 21 |
from __future__ import annotations |
| 22 |
|
| 23 |
from pathlib import Path |
| 24 |
|
| 25 |
from pydantic import ValidationError |
| 26 |
|
| 27 |
from dlm.doc.errors import DlmVersionError, SchemaValidationError, UnsupportedMigrationError |
| 28 |
from dlm.doc.migrations.dispatch import apply_pending |
| 29 |
from dlm.doc.schema import CURRENT_SCHEMA_VERSION, DlmFrontmatter |
| 30 |
|
| 31 |
|
| 32 |
def validate_versioned(raw: dict[str, object], *, path: Path | None = None) -> DlmFrontmatter: |
| 33 |
"""Dispatch: migrate (if needed) then Pydantic-validate. |
| 34 |
|
| 35 |
Raises: |
| 36 |
DlmVersionError: `raw["dlm_version"]` is newer than this parser |
| 37 |
supports, or a required intermediate migrator is missing. |
| 38 |
SchemaValidationError: Pydantic validation failed after any |
| 39 |
applicable migrations. |
| 40 |
""" |
| 41 |
version = raw.get("dlm_version", 1) |
| 42 |
# `isinstance(v, int)` is True for `bool`, so exclude it explicitly — |
| 43 |
# `dlm_version: true` would otherwise coerce to version 1 silently. |
| 44 |
if not isinstance(version, int) or isinstance(version, bool): |
| 45 |
raise SchemaValidationError( |
| 46 |
f"dlm_version must be an integer, got {type(version).__name__}", |
| 47 |
path=path, |
| 48 |
line=2, |
| 49 |
) |
| 50 |
try: |
| 51 |
migrated, applied = apply_pending(raw, target_version=CURRENT_SCHEMA_VERSION) |
| 52 |
except UnsupportedMigrationError: |
| 53 |
raise |
| 54 |
except DlmVersionError as exc: |
| 55 |
raise DlmVersionError(exc.message, path=path, line=2) from exc |
| 56 |
if applied: |
| 57 |
# The parser's own error path logs migrations as an info line; the |
| 58 |
# migration framework's data-rewrite happens here without |
| 59 |
# side-effects on the source file. `dlm migrate <path>` is the |
| 60 |
# write-path — parsing never mutates the on-disk document. |
| 61 |
pass |
| 62 |
|
| 63 |
try: |
| 64 |
return DlmFrontmatter.model_validate(migrated) |
| 65 |
except ValidationError as exc: |
| 66 |
raise SchemaValidationError( |
| 67 |
_format_pydantic_error(exc), |
| 68 |
path=path, |
| 69 |
line=2, |
| 70 |
) from exc |
| 71 |
|
| 72 |
|
| 73 |
def _format_pydantic_error(exc: ValidationError) -> str: |
| 74 |
"""Collapse Pydantic's error-tree into a single-line message.""" |
| 75 |
parts = [] |
| 76 |
for err in exc.errors(): |
| 77 |
loc = ".".join(str(p) for p in err.get("loc", ())) or "<root>" |
| 78 |
msg = err.get("msg", "invalid value") |
| 79 |
parts.append(f"{loc}: {msg}") |
| 80 |
return "; ".join(parts) or "validation failed" |