Add a structural check over HQ's own records

Nothing in this repository was verified by anything but reading, which is how
a superseded decision stayed live in the constitution and in the to-be README
at the same time. Both were found by a person looking, and nothing stopped a
third.

Five checks: links resolve; `decisions:`/`extends:` name records that exist and
are accepted; a governing document citing a superseded record must name its
replacement in the same paragraph; supersession is symmetric; filename number
matches heading number.

Each was made to fail before it was made to pass. The live-citation check was
verified against a reconstruction of the actual incident -- the to-be README
citing ADR 0017 as live guidance -- and reports it with file and line.

It found one thing nobody had noticed: ADR 0018 never declared that it
superseded 0011, though 0011 has named 0018 as its superseder since August.
Fixed.

Deliberately not checked, and said so in the README: 02-DECISIONS and
01-RESEARCH may cite superseded records freely, because a decision record
discusses history and research records what was observed. 00-as-is may rest on
one, per 0056. Flagging those would put noise on correct documents, and a check
that cries wolf gets suppressed -- which costs more than not having it.

Two bugs found by running it: the frontmatter reader iterated an inline list as
characters, and the as-is exemption was missing entirely.
This commit is contained in:
2026-08-27 20:18:35 +02:00
parent e1f4c7d9e0
commit 03874f3fe2
3 changed files with 321 additions and 0 deletions
+45
View File
@@ -0,0 +1,45 @@
# Checks
```
python3 00-META/checks/records.py
```
Non-zero exit on any problem, so it can be a gate rather than a report.
**Why this exists.** Until now nothing in this repository was verified by anything but reading,
which is how a superseded decision stayed live in the constitution for days and in
`01-to-be/README.md` alongside it. Both were found by a person looking. `how-we-build` §5 says
*an unenforced rule is indistinguishable from a wrong one, and costs more, because people
believe it* — this repository was carrying several.
**Every check here failed on something real before it passed.** A check that has never failed is
indistinguishable from one that cannot.
| Check | Asserts | Found |
|---|---|---|
| `links` | every relative link resolves | — (run ad hoc during authoring; now permanent) |
| `rests-on` | `decisions:` and `extends:` name records that exist and are **accepted** | the class behind both incidents |
| `live-citation` | a governing document citing a **superseded** record names its replacement in the same paragraph | `01-to-be/README.md` citing ADR 0017 as live guidance |
| `supersession` | if A says it was superseded by B, B says it supersedes A | ADR 0018 never declared that it superseded 0011 |
| `numbering` | the number in the filename is the number in the heading | — |
## What is deliberately not checked
- **`02-DECISIONS/` and `01-RESEARCH/` may cite superseded records freely.** A decision record
discusses history; research records what was observed. Flagging those would produce noise on
correct documents, and a check that cries wolf gets suppressed — which costs more than not
having it.
- **`03-DESIGN/00-as-is/` may rest on a superseded record.** It describes what runs, and what
runs was built under whatever was decided at the time
([ADR 0056](../../02-DECISIONS/0056-the-authority-is-the-control-plane-not-a-database.md):
*as-is describing a superseded decision is exactly what as-is is for*).
- **Whether a citation's prose is still true.** Only whether the record it points at is live.
A document can cite an accepted record and describe it wrongly, and nothing here notices.
So "governing" means `00-META/` and `03-DESIGN/01-to-be/` — the documents that tell somebody
what to do.
## Adding a check
State what incident it would have caught, and make it fail before you make it pass. A check
whose failure has never been observed is a guess about its own correctness.
+275
View File
@@ -0,0 +1,275 @@
#!/usr/bin/env python3
"""Structural checks over HQ's own records.
Every check here exists because the thing it checks for actually happened. See README.md
for which incident is behind which check. Run from the repository root:
python3 00-META/checks/records.py
Exits non-zero if anything fails, so it can be a gate rather than a report.
"""
import os
import re
import sys
ROOT = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
# Where a citation is *guidance* rather than history. A document here tells somebody what to
# do, so a link to a superseded record is an instruction to follow a withdrawn decision.
# 02-DECISIONS and 01-RESEARCH are deliberately absent: they record what was decided and what
# was observed, and both legitimately discuss superseded records at length.
GOVERNING = ("00-META/", "03-DESIGN/01-to-be/")
LINK = re.compile(r"\[[^\]]*\]\((?!https?:|mailto:)([^)]+)\)")
ADR_FILE = re.compile(r"^(\d{4})-")
def markdown_files():
for base, dirs, files in os.walk(ROOT):
dirs[:] = [d for d in dirs if d not in (".git", ".claude")]
for name in sorted(files):
if name.endswith(".md"):
yield os.path.join(base, name)
def rel(path):
return os.path.relpath(path, ROOT)
def read(path):
with open(path, encoding="utf-8") as handle:
return handle.read()
def frontmatter(text):
"""Minimal frontmatter reader — enough for the fields these checks use.
Not a YAML parser on purpose: a dependency in a repository that has none, to read four
scalar fields and one list, would cost more than it returns.
"""
if not text.startswith("---\n"):
return {}
end = text.find("\n---", 4)
if end == -1:
return {}
fields, key = {}, None
for line in text[4:end].split("\n"):
item = re.match(r"^\s+-\s+(.*)$", line)
if item and key:
fields.setdefault(key, []).append(item.group(1).strip())
continue
pair = re.match(r"^([A-Za-z_-]+):\s*(.*)$", line)
if pair:
key = pair.group(1)
value = pair.group(2).strip()
if value.startswith("[") and value.endswith("]"):
# inline list: `decisions: [a, b]`, `code: []`
inner = value[1:-1].strip()
fields[key] = [v.strip() for v in inner.split(",") if v.strip()]
elif value:
fields[key] = value
else:
fields[key] = []
return fields
def load_records():
"""Every decision record, by its four-digit number."""
records = {}
folder = os.path.join(ROOT, "02-DECISIONS")
for name in sorted(os.listdir(folder)):
match = ADR_FILE.match(name)
if not name.endswith(".md") or not match:
continue
path = os.path.join(folder, name)
text = read(path)
records[match.group(1)] = {
"number": match.group(1),
"name": name,
"path": path,
"text": text,
"front": frontmatter(text),
}
return records
class Failures:
def __init__(self):
self.items = []
def add(self, check, location, message):
self.items.append((check, location, message))
def report(self):
if not self.items:
print("records: all checks passed")
return 0
by_check = {}
for check, location, message in self.items:
by_check.setdefault(check, []).append((location, message))
for check in sorted(by_check):
print(f"\n{check} — {len(by_check[check])} problem(s)")
for location, message in by_check[check]:
print(f" {location}\n {message}")
print(f"\nrecords: {len(self.items)} problem(s)")
return 1
def check_links(failures):
"""Every relative link resolves to something that exists."""
for path in markdown_files():
folder = os.path.dirname(path)
for number, line in enumerate(read(path).split("\n"), 1):
for target in LINK.findall(line):
target = target.split("#")[0].strip()
if not target:
continue
if not os.path.exists(os.path.normpath(os.path.join(folder, target))):
failures.add("links", f"{rel(path)}:{number}", f"link does not resolve: {target}")
def check_rests_on(failures, records):
"""A document's `decisions:` and a record's `extends:` must name a live record.
These are the load-bearing citations: the document declares that it rests on that
decision. Resting on a withdrawn one is the defect this whole check set exists for.
"""
for path in markdown_files():
front = frontmatter(read(path))
cited = list(front.get("decisions", []) or [])
extends = front.get("extends")
if isinstance(extends, str) and extends:
cited.append(extends)
for entry in cited:
match = ADR_FILE.match(os.path.basename(entry))
if not match:
failures.add("rests-on", rel(path), f"not a decision record: {entry}")
continue
number = match.group(1)
if number not in records:
failures.add("rests-on", rel(path), f"no such record: {entry}")
continue
status = records[number]["front"].get("status")
if status != "accepted":
# An as-is document describes what runs, and what runs was built under
# whatever was decided at the time. ADR 0056: "as-is describing a superseded
# decision is exactly what as-is is for."
if rel(path).startswith("03-DESIGN/00-as-is/"):
continue
# An extension that supersedes legitimately names what it replaced.
this = ADR_FILE.match(os.path.basename(path))
supersedes = records[number]["front"].get("superseded-by", "")
if this and supersedes and os.path.basename(path) in str(supersedes):
continue
failures.add(
"rests-on",
rel(path),
f"rests on ADR {number}, which is '{status}' — a document may not rest on a "
f"record that is not accepted",
)
def check_live_citations(failures, records):
"""In a governing document, a link to a superseded record must name its replacement.
The reader of a rule needs to know the rule was withdrawn, and needs somewhere to go.
Naming the superseder in the same paragraph is both, and it is what a person would
write anyway.
"""
superseded = {
number: record["front"].get("superseded-by", "")
for number, record in records.items()
if record["front"].get("status") == "superseded"
}
for path in markdown_files():
if not any(rel(path).startswith(prefix) for prefix in GOVERNING):
continue
text = read(path)
offset = 0
for paragraph in text.split("\n\n"):
line_no = text[:offset].count("\n") + 1
offset += len(paragraph) + 2
targets = LINK.findall(paragraph)
for target in targets:
match = ADR_FILE.match(os.path.basename(target.split("#")[0]))
if not match or match.group(1) not in superseded:
continue
number = match.group(1)
replacement = os.path.basename(str(superseded[number]))
if not replacement:
failures.add(
"live-citation",
f"{rel(path)}:{line_no}",
f"cites superseded ADR {number}, which names no superseder",
)
continue
if not any(replacement in t for t in targets):
failures.add(
"live-citation",
f"{rel(path)}:{line_no}",
f"cites superseded ADR {number} without naming its replacement "
f"({replacement}) in the same paragraph",
)
def check_supersession_symmetry(failures, records):
"""If A says it was superseded by B, B must say it supersedes A."""
for number, record in records.items():
front = record["front"]
status = front.get("status")
by = os.path.basename(str(front.get("superseded-by", "")))
if status == "superseded" and not by:
failures.add("supersession", rel(record["path"]), "marked superseded but names no superseder")
continue
if by and status != "superseded":
failures.add("supersession", rel(record["path"]), f"names a superseder but status is '{status}'")
if not by:
continue
match = ADR_FILE.match(by)
if not match or match.group(1) not in records:
failures.add("supersession", rel(record["path"]), f"superseder does not exist: {by}")
continue
other = records[match.group(1)]
claims = os.path.basename(str(other["front"].get("supersedes", "")))
if claims != record["name"]:
failures.add(
"supersession",
rel(other["path"]),
f"ADR {number} says this supersedes it; this record does not say so "
f"(supersedes: {claims or 'absent'})",
)
def check_numbering(failures, records):
"""The number in the filename is the number in the heading."""
for number, record in records.items():
heading = re.search(r"^# (\d+)\.", record["text"], re.M)
if not heading:
failures.add("numbering", rel(record["path"]), "no '# N. Title' heading")
elif heading.group(1) != number.lstrip("0"):
failures.add(
"numbering",
rel(record["path"]),
f"filename says {number}, heading says {heading.group(1)}",
)
def main():
failures = Failures()
records = load_records()
check_links(failures)
check_rests_on(failures, records)
check_live_citations(failures, records)
check_supersession_symmetry(failures, records)
check_numbering(failures, records)
print(f"records: {len(records)} decision records checked")
return failures.report()
if __name__ == "__main__":
sys.exit(main())
@@ -3,6 +3,7 @@ status: accepted
date: 2026-08-23
deciders: jochen
reconstructed: false
supersedes: 0011-the-installer-owns-linking.md
extends: 0011-the-installer-owns-linking.md
---