feat: complete evidence restructuring worktree
This commit is contained in:
Codex
@@ -226,7 +226,7 @@ _EVIDENCE_ID = re.compile(r"^evidence:[a-z0-9]+(?:-[a-z0-9]+)*$")
|
||||
|
||||
|
||||
class CuratedEvidence(StrictModel):
|
||||
schema_version: Literal[1]
|
||||
schema_version: Literal[1, 2]
|
||||
id: str
|
||||
title: str
|
||||
kind: EvidenceKind
|
||||
@@ -247,6 +247,439 @@ class CuratedEvidence(StrictModel):
|
||||
return self
|
||||
|
||||
|
||||
_V2_LABELS = {
|
||||
"en": {
|
||||
"column": "Column",
|
||||
"columns": "Columns",
|
||||
"concept": "Concept",
|
||||
"definition": "Definition",
|
||||
"description": "Description",
|
||||
"empty": "No items",
|
||||
"input": "Input",
|
||||
"interpretation": "Interpretation",
|
||||
"label": "Label",
|
||||
"output": "Output",
|
||||
"question": "Question",
|
||||
"review_items": "Review items",
|
||||
"field": "Field",
|
||||
"rule": "Rule",
|
||||
"sql": "SQL",
|
||||
"supporting_excerpts": "Supporting excerpts",
|
||||
"synonyms": "Synonyms",
|
||||
"tables": "Tables",
|
||||
"url": "URL",
|
||||
"value": "Value",
|
||||
"values": "Values",
|
||||
"meaning": "Meaning",
|
||||
"variants": "Variants",
|
||||
},
|
||||
"it": {
|
||||
"column": "Colonna",
|
||||
"columns": "Colonne",
|
||||
"concept": "Concetto",
|
||||
"definition": "Definizione",
|
||||
"description": "Descrizione",
|
||||
"empty": "Nessun elemento",
|
||||
"input": "Input",
|
||||
"interpretation": "Interpretazione",
|
||||
"label": "Etichetta",
|
||||
"output": "Output",
|
||||
"question": "Domanda",
|
||||
"review_items": "Elementi da rivedere",
|
||||
"field": "Campo",
|
||||
"rule": "Regola",
|
||||
"sql": "SQL",
|
||||
"supporting_excerpts": "Estratti di supporto",
|
||||
"synonyms": "Sinonimi",
|
||||
"tables": "Tabelle",
|
||||
"url": "URL",
|
||||
"value": "Valore",
|
||||
"values": "Valori",
|
||||
"meaning": "Significato",
|
||||
"variants": "Varianti",
|
||||
},
|
||||
}
|
||||
_V2_FIELD = re.compile(
|
||||
r"<!-- tht:field:([a-z_]+) -->\n(.*?)\n<!-- /tht:field:\1 -->",
|
||||
re.DOTALL,
|
||||
)
|
||||
_V2_EXCERPT_SEPARATOR = "<!-- tht:excerpt-separator -->"
|
||||
_V2_EMPTY_LIST = "<!-- tht:empty-list -->"
|
||||
_V2_REVIEW_SEPARATOR = "<!-- tht:review-separator -->"
|
||||
_V2_REVIEW_FIELD = "<!-- tht:review-field -->"
|
||||
|
||||
|
||||
def _v2_labels(language: str) -> dict[str, str]:
|
||||
return _V2_LABELS["it" if language.lower().startswith("it") else "en"]
|
||||
|
||||
|
||||
def _render_v2_field(name: str, label: str, content: str) -> str:
|
||||
closing_marker = f"<!-- /tht:field:{name} -->"
|
||||
if "<!-- tht:field:" in content or "<!-- /tht:field:" in content:
|
||||
raise ValueError(f"curated evidence {name} contains a reserved marker")
|
||||
return (
|
||||
f"<!-- tht:field:{name} -->\n"
|
||||
f"## {label}\n\n"
|
||||
f"{content}\n"
|
||||
f"{closing_marker}"
|
||||
)
|
||||
|
||||
|
||||
def _render_v2_excerpt(value: str) -> str:
|
||||
if _V2_EXCERPT_SEPARATOR in value:
|
||||
raise ValueError("supporting excerpt contains a reserved marker")
|
||||
quoted = "\n".join(">" if not line else f"> {line}" for line in value.split("\n"))
|
||||
return quoted
|
||||
|
||||
|
||||
def _render_v2_list(
|
||||
values: tuple[str, ...], *, code: bool = False, empty_label: str = "No items",
|
||||
) -> str:
|
||||
if not values:
|
||||
return f"{_V2_EMPTY_LIST}\n_{empty_label}._"
|
||||
if any("\n" in value for value in values):
|
||||
raise ValueError("curated evidence list values must be single-line")
|
||||
if code and any("`" in value for value in values):
|
||||
raise ValueError("curated evidence code values must not contain backticks")
|
||||
return "\n".join(f"- `{value}`" if code else f"- {value}" for value in values)
|
||||
|
||||
|
||||
def _parse_v2_list(
|
||||
value: str, *, code: bool = False, empty_label: str = "No items",
|
||||
) -> tuple[str, ...]:
|
||||
if value == f"{_V2_EMPTY_LIST}\n_{empty_label}._":
|
||||
return ()
|
||||
parsed: list[str] = []
|
||||
for line in value.split("\n"):
|
||||
if not line.startswith("- "):
|
||||
raise ValueError("curated evidence list is malformed")
|
||||
item = line[2:]
|
||||
if code:
|
||||
if len(item) < 2 or not item.startswith("`") or not item.endswith("`"):
|
||||
raise ValueError("curated evidence code list is malformed")
|
||||
item = item[1:-1]
|
||||
parsed.append(item)
|
||||
return tuple(parsed)
|
||||
|
||||
|
||||
def _escape_v2_table_value(value: str) -> str:
|
||||
return value.replace("\\", "\\\\").replace("|", "\\|").replace("\n", "\\n")
|
||||
|
||||
|
||||
def _unescape_v2_table_value(value: str) -> str:
|
||||
output: list[str] = []
|
||||
index = 0
|
||||
while index < len(value):
|
||||
if value[index] != "\\":
|
||||
output.append(value[index])
|
||||
index += 1
|
||||
continue
|
||||
if index + 1 >= len(value):
|
||||
raise ValueError("curated evidence table escape is malformed")
|
||||
escaped = value[index + 1]
|
||||
if escaped not in {"\\", "|", "n"}:
|
||||
raise ValueError("curated evidence table escape is malformed")
|
||||
output.append("\n" if escaped == "n" else escaped)
|
||||
index += 2
|
||||
return "".join(output)
|
||||
|
||||
|
||||
def _render_v2_values(values: dict[str, str], labels: dict[str, str]) -> str:
|
||||
if any("`" in value for value in values):
|
||||
raise ValueError("curated evidence enum values must not contain backticks")
|
||||
rows = [
|
||||
f"| {labels['value']} | {labels['meaning']} |",
|
||||
"| --- | --- |",
|
||||
]
|
||||
if not values:
|
||||
rows.append(f"| _{labels['empty']}._ | |")
|
||||
return "\n".join(rows)
|
||||
rows.extend(
|
||||
f"| `{_escape_v2_table_value(value)}` | {_escape_v2_table_value(meaning)} |"
|
||||
for value, meaning in sorted(values.items())
|
||||
)
|
||||
return "\n".join(rows)
|
||||
|
||||
|
||||
def _parse_v2_values(value: str, labels: dict[str, str]) -> dict[str, str]:
|
||||
lines = value.split("\n")
|
||||
if (
|
||||
len(lines) < 3
|
||||
or lines[0] != f"| {labels['value']} | {labels['meaning']} |"
|
||||
or lines[1] != "| --- | --- |"
|
||||
):
|
||||
raise ValueError("curated evidence values table is malformed")
|
||||
if lines[2:] == [f"| _{labels['empty']}._ | |"]:
|
||||
return {}
|
||||
parsed: dict[str, str] = {}
|
||||
for line in lines[2:]:
|
||||
if not line.startswith("| ") or not line.endswith(" |"):
|
||||
raise ValueError("curated evidence values table is malformed")
|
||||
cells = re.split(r"(?<!\\)\s\|\s", line[2:-2], maxsplit=1)
|
||||
if len(cells) != 2 or not cells[0].startswith("`") or not cells[0].endswith("`"):
|
||||
raise ValueError("curated evidence values table is malformed")
|
||||
key = _unescape_v2_table_value(cells[0][1:-1])
|
||||
if key in parsed:
|
||||
raise ValueError("curated evidence enum value appears more than once")
|
||||
parsed[key] = _unescape_v2_table_value(cells[1])
|
||||
return parsed
|
||||
|
||||
|
||||
def _render_v2_payload(value: CuratedEvidence, labels: dict[str, str]) -> list[str]:
|
||||
payload = value.payload
|
||||
if value.kind == "glossary":
|
||||
fields = [_render_v2_field("definition", labels["definition"], payload.definition)]
|
||||
if payload.synonyms:
|
||||
fields.append(_render_v2_field(
|
||||
"synonyms", labels["synonyms"], _render_v2_list(payload.synonyms),
|
||||
))
|
||||
if payload.variants:
|
||||
fields.append(_render_v2_field(
|
||||
"variants", labels["variants"], _render_v2_list(payload.variants),
|
||||
))
|
||||
return fields
|
||||
if value.kind == "domain":
|
||||
return [_render_v2_field("rule", labels["rule"], payload.rule)]
|
||||
if value.kind == "enum":
|
||||
return [
|
||||
_render_v2_field("column", labels["column"], f"`{payload.column}`"),
|
||||
_render_v2_field("values", labels["values"], _render_v2_values(payload.values, labels)),
|
||||
]
|
||||
if value.kind == "example":
|
||||
return [
|
||||
_render_v2_field("question", labels["question"], payload.question),
|
||||
_render_v2_field("interpretation", labels["interpretation"], payload.interpretation),
|
||||
]
|
||||
if value.kind == "mapping":
|
||||
return [
|
||||
_render_v2_field("concept", labels["concept"], payload.concept),
|
||||
_render_v2_field("tables", labels["tables"], _render_v2_list(
|
||||
payload.tables, code=True, empty_label=labels["empty"],
|
||||
)),
|
||||
_render_v2_field("columns", labels["columns"], _render_v2_list(
|
||||
payload.columns, code=True, empty_label=labels["empty"],
|
||||
)),
|
||||
]
|
||||
if value.kind == "normalization":
|
||||
return [
|
||||
_render_v2_field("input", labels["input"], payload.input),
|
||||
_render_v2_field("output", labels["output"], payload.output),
|
||||
_render_v2_field("rule", labels["rule"], payload.rule),
|
||||
]
|
||||
if value.kind == "formula":
|
||||
if "```" in payload.sql:
|
||||
raise ValueError("curated evidence SQL contains a reserved Markdown fence")
|
||||
return [
|
||||
_render_v2_field("concept", labels["concept"], payload.concept),
|
||||
_render_v2_field("columns", labels["columns"], _render_v2_list(
|
||||
payload.columns, code=True, empty_label=labels["empty"],
|
||||
)),
|
||||
_render_v2_field("sql", labels["sql"], f"```sql\n{payload.sql}\n```"),
|
||||
]
|
||||
if value.kind == "reference":
|
||||
return [
|
||||
_render_v2_field("label", labels["label"], payload.label),
|
||||
_render_v2_field("url", labels["url"], f"<{payload.url}>"),
|
||||
_render_v2_field("description", labels["description"], payload.description),
|
||||
]
|
||||
raise ValueError(f"unsupported curated evidence kind {value.kind}")
|
||||
|
||||
|
||||
def _render_v2_review_items(value: CuratedEvidence, labels: dict[str, str]) -> str:
|
||||
rendered: list[str] = []
|
||||
for item in value.review_items:
|
||||
if "`" in item.code or (item.field is not None and "`" in item.field):
|
||||
raise ValueError("curated evidence review identifiers must not contain backticks")
|
||||
if _V2_REVIEW_SEPARATOR in item.message or _V2_REVIEW_FIELD in item.message:
|
||||
raise ValueError("curated evidence review message contains a reserved marker")
|
||||
block = f"### `{item.code}`\n\n{item.message}"
|
||||
if item.field is not None:
|
||||
block += f"\n\n{_V2_REVIEW_FIELD}\n**{labels['field']}:** `{item.field}`"
|
||||
rendered.append(block)
|
||||
return f"\n{_V2_REVIEW_SEPARATOR}\n".join(rendered)
|
||||
|
||||
|
||||
def _render_v2_body(value: CuratedEvidence) -> str:
|
||||
if "\n" in value.title:
|
||||
raise ValueError("curated evidence title must be single-line in v2")
|
||||
labels = _v2_labels(value.language)
|
||||
fields = [
|
||||
*_render_v2_payload(value, labels),
|
||||
_render_v2_field(
|
||||
"supporting_excerpts",
|
||||
labels["supporting_excerpts"],
|
||||
f"\n{_V2_EXCERPT_SEPARATOR}\n".join(
|
||||
_render_v2_excerpt(excerpt)
|
||||
for excerpt in value.provenance.supporting_excerpts
|
||||
),
|
||||
),
|
||||
]
|
||||
if value.review_items:
|
||||
fields.append(_render_v2_field(
|
||||
"review_items",
|
||||
labels["review_items"],
|
||||
_render_v2_review_items(value, labels),
|
||||
))
|
||||
return f"# {value.title}\n\n" + "\n\n".join(fields) + "\n"
|
||||
|
||||
|
||||
def _parse_v2_field_content(name: str, block: str) -> str:
|
||||
try:
|
||||
heading, content = block.split("\n\n", 1)
|
||||
except ValueError as error:
|
||||
raise ValueError(f"curated evidence field {name} is malformed") from error
|
||||
if not heading.startswith("## ") or not content:
|
||||
raise ValueError(f"curated evidence field {name} is malformed")
|
||||
return content
|
||||
|
||||
|
||||
def _parse_v2_excerpts(block: str) -> tuple[str, ...]:
|
||||
excerpts: list[str] = []
|
||||
for raw_excerpt in block.split(f"\n{_V2_EXCERPT_SEPARATOR}\n"):
|
||||
lines = raw_excerpt.split("\n")
|
||||
if any(line != ">" and not line.startswith("> ") for line in lines):
|
||||
raise ValueError("curated evidence supporting excerpt is malformed")
|
||||
excerpts.append("\n".join(line[2:] if line.startswith("> ") else "" for line in lines))
|
||||
if not excerpts:
|
||||
raise ValueError("curated evidence supporting excerpts are malformed")
|
||||
return tuple(excerpts)
|
||||
|
||||
|
||||
def _parse_inline_code(value: str, name: str) -> str:
|
||||
if len(value) < 2 or not value.startswith("`") or not value.endswith("`"):
|
||||
raise ValueError(f"curated evidence field {name} must be inline code")
|
||||
return value[1:-1]
|
||||
|
||||
|
||||
def _parse_v2_payload(
|
||||
kind: str, fields: dict[str, str], labels: dict[str, str],
|
||||
) -> tuple[dict, set[str]]:
|
||||
if kind == "glossary":
|
||||
expected = {"definition"}
|
||||
payload: dict = {"definition": fields.get("definition")}
|
||||
for name in ("synonyms", "variants"):
|
||||
if name in fields:
|
||||
expected.add(name)
|
||||
payload[name] = _parse_v2_list(fields[name], empty_label=labels["empty"])
|
||||
else:
|
||||
payload[name] = ()
|
||||
return payload, expected
|
||||
if kind == "domain":
|
||||
return {"rule": fields.get("rule")}, {"rule"}
|
||||
if kind == "enum":
|
||||
return {
|
||||
"column": _parse_inline_code(fields.get("column", ""), "column"),
|
||||
"values": _parse_v2_values(fields.get("values", ""), labels),
|
||||
}, {"column", "values"}
|
||||
if kind == "example":
|
||||
return {
|
||||
"question": fields.get("question"),
|
||||
"interpretation": fields.get("interpretation"),
|
||||
}, {"question", "interpretation"}
|
||||
if kind == "mapping":
|
||||
return {
|
||||
"concept": fields.get("concept"),
|
||||
"tables": _parse_v2_list(
|
||||
fields.get("tables", ""), code=True, empty_label=labels["empty"],
|
||||
),
|
||||
"columns": _parse_v2_list(
|
||||
fields.get("columns", ""), code=True, empty_label=labels["empty"],
|
||||
),
|
||||
}, {"concept", "tables", "columns"}
|
||||
if kind == "normalization":
|
||||
return {
|
||||
"input": fields.get("input"),
|
||||
"output": fields.get("output"),
|
||||
"rule": fields.get("rule"),
|
||||
}, {"input", "output", "rule"}
|
||||
if kind == "formula":
|
||||
sql = fields.get("sql", "")
|
||||
if not sql.startswith("```sql\n") or not sql.endswith("\n```"):
|
||||
raise ValueError("curated evidence SQL block is malformed")
|
||||
return {
|
||||
"concept": fields.get("concept"),
|
||||
"columns": _parse_v2_list(
|
||||
fields.get("columns", ""), code=True, empty_label=labels["empty"],
|
||||
),
|
||||
"sql": sql.removeprefix("```sql\n").removesuffix("\n```"),
|
||||
}, {"concept", "columns", "sql"}
|
||||
if kind == "reference":
|
||||
url = fields.get("url", "")
|
||||
if not url.startswith("<") or not url.endswith(">"):
|
||||
raise ValueError("curated evidence reference URL is malformed")
|
||||
return {
|
||||
"label": fields.get("label"),
|
||||
"url": url[1:-1],
|
||||
"description": fields.get("description"),
|
||||
}, {"label", "url", "description"}
|
||||
raise ValueError("curated evidence body kind is unsupported")
|
||||
|
||||
|
||||
def _parse_v2_review_items(value: str) -> tuple[ReviewItem, ...]:
|
||||
items: list[ReviewItem] = []
|
||||
for raw_item in value.split(f"\n{_V2_REVIEW_SEPARATOR}\n"):
|
||||
try:
|
||||
heading, detail = raw_item.split("\n\n", 1)
|
||||
except ValueError as error:
|
||||
raise ValueError("curated evidence review item is malformed") from error
|
||||
if not heading.startswith("### `") or not heading.endswith("`"):
|
||||
raise ValueError("curated evidence review item code is malformed")
|
||||
code = heading.removeprefix("### `").removesuffix("`")
|
||||
field = None
|
||||
marker = f"\n\n{_V2_REVIEW_FIELD}\n"
|
||||
if marker in detail:
|
||||
message, rendered_field = detail.split(marker, 1)
|
||||
match = re.fullmatch(r"\*\*[^*]+:\*\* `([^`]+)`", rendered_field)
|
||||
if match is None:
|
||||
raise ValueError("curated evidence review item field is malformed")
|
||||
field = match.group(1)
|
||||
else:
|
||||
message = detail
|
||||
if not code or not message:
|
||||
raise ValueError("curated evidence review item is malformed")
|
||||
items.append(ReviewItem(code=code, message=message, field=field))
|
||||
return tuple(items)
|
||||
|
||||
|
||||
def _parse_v2_body(data: dict, body: str) -> dict:
|
||||
kind = data.get("kind")
|
||||
body_owned = {"payload", "review_items"}
|
||||
if isinstance(kind, str):
|
||||
body_owned.add(kind)
|
||||
if body_owned.intersection(data):
|
||||
raise ValueError("curated evidence v2 frontmatter contains body-owned fields")
|
||||
title = data.get("title")
|
||||
if not isinstance(title, str) or not body.startswith(f"# {title}\n"):
|
||||
raise ValueError("curated evidence body title must match its metadata")
|
||||
fields: dict[str, str] = {}
|
||||
for match in _V2_FIELD.finditer(body):
|
||||
name = match.group(1)
|
||||
if name in fields:
|
||||
raise ValueError(f"curated evidence field {name} appears more than once")
|
||||
fields[name] = _parse_v2_field_content(name, match.group(2))
|
||||
skeleton = _V2_FIELD.sub("", body).strip()
|
||||
if skeleton != f"# {title}":
|
||||
raise ValueError("curated evidence body contains unstructured content")
|
||||
labels = _v2_labels(str(data.get("language", "")))
|
||||
payload, payload_fields = _parse_v2_payload(kind, fields, labels)
|
||||
common_fields = {"supporting_excerpts"}
|
||||
if "review_items" in fields:
|
||||
common_fields.add("review_items")
|
||||
if set(fields) != payload_fields | common_fields:
|
||||
raise ValueError("curated evidence body fields do not match its kind")
|
||||
provenance = data.get("provenance")
|
||||
if not isinstance(provenance, dict) or "supporting_excerpts" in provenance:
|
||||
raise ValueError("curated evidence v2 provenance is malformed")
|
||||
provenance["supporting_excerpts"] = _parse_v2_excerpts(fields["supporting_excerpts"])
|
||||
data["review_items"] = (
|
||||
_parse_v2_review_items(fields["review_items"])
|
||||
if "review_items" in fields
|
||||
else []
|
||||
)
|
||||
data["payload"] = payload
|
||||
return data
|
||||
|
||||
|
||||
def parse_curated_markdown(text: str, *, path: Path | None = None) -> CuratedEvidence:
|
||||
"""Parse the canonical frontmatter representation of one Curated Evidence unit."""
|
||||
if not text.startswith("---\n"):
|
||||
@@ -260,11 +693,14 @@ def parse_curated_markdown(text: str, *, path: Path | None = None) -> CuratedEvi
|
||||
data = dict(raw)
|
||||
except (TypeError, ValueError) as error:
|
||||
raise ValueError("curated evidence frontmatter must be a mapping") from error
|
||||
if body.strip():
|
||||
raise ValueError("curated evidence must not contain an ignored body")
|
||||
kind = data.get("kind")
|
||||
if "payload" not in data and kind in _PAYLOAD_TYPE_BY_KIND:
|
||||
data["payload"] = data.pop(kind, None)
|
||||
if data.get("schema_version") == 2:
|
||||
data = _parse_v2_body(data, body)
|
||||
else:
|
||||
if body.strip():
|
||||
raise ValueError("curated evidence must not contain an ignored body")
|
||||
kind = data.get("kind")
|
||||
if "payload" not in data and kind in _PAYLOAD_TYPE_BY_KIND:
|
||||
data["payload"] = data.pop(kind, None)
|
||||
evidence = CuratedEvidence.model_validate(data)
|
||||
if path is not None:
|
||||
_validate_kind_directory(path, evidence.kind)
|
||||
@@ -273,6 +709,11 @@ def parse_curated_markdown(text: str, *, path: Path | None = None) -> CuratedEvi
|
||||
|
||||
def dump_curated_markdown(value: CuratedEvidence) -> str:
|
||||
"""Render canonical frontmatter with a human-readable kind-specific payload key."""
|
||||
if value.schema_version == 2:
|
||||
data = value.model_dump(mode="json", exclude={"payload", "review_items"})
|
||||
data["provenance"].pop("supporting_excerpts")
|
||||
frontmatter = yaml.safe_dump(data, allow_unicode=True, sort_keys=False)
|
||||
return f"---\n{frontmatter}---\n{_render_v2_body(value)}"
|
||||
data = value.model_dump(mode="json", exclude={"payload"})
|
||||
data[value.kind] = value.payload.model_dump(mode="json")
|
||||
frontmatter = yaml.safe_dump(data, allow_unicode=True, sort_keys=False)
|
||||
|
||||
Reference in New Issue
Block a user