|
1
|
""" |
|
2
|
ADR ingestion — structured Architecture Decision Records as Decision nodes. |
|
3
|
|
|
4
|
Parses ADR markdown files in standard MADR format (or a relaxed variant) |
|
5
|
and creates Decision nodes in the navegador knowledge graph. |
|
6
|
|
|
7
|
Usage: |
|
8
|
from navegador.adr import ADRIngester |
|
9
|
|
|
10
|
ingester = ADRIngester(store) |
|
11
|
stats = ingester.ingest("/path/to/docs/decisions") |
|
12
|
""" |
|
13
|
|
|
14
|
from __future__ import annotations |
|
15
|
|
|
16
|
import logging |
|
17
|
import re |
|
18
|
from pathlib import Path |
|
19
|
from typing import Any |
|
20
|
|
|
21
|
from navegador.graph.schema import NodeLabel |
|
22
|
from navegador.graph.store import GraphStore |
|
23
|
|
|
24
|
logger = logging.getLogger(__name__) |
|
25
|
|
|
26
|
# ── Regex helpers ───────────────────────────────────────────────────────────── |
|
27
|
|
|
28
|
_H1 = re.compile(r"^#\s+(.+)$", re.MULTILINE) |
|
29
|
_STATUS = re.compile(r"^#{1,3}\s+Status\s*\n+(.+?)(?=\n#{1,3}\s|\Z)", re.MULTILINE | re.DOTALL) |
|
30
|
_RATIONALE = re.compile( |
|
31
|
r"^#{1,3}\s+Rationale\s*\n+(.+?)(?=\n#{1,3}\s|\Z)", |
|
32
|
re.MULTILINE | re.DOTALL, |
|
33
|
) |
|
34
|
_RATIONALE_FALLBACK = re.compile( |
|
35
|
r"^#{1,3}\s+(?:Decision Outcome|Decision)\s*\n+(.+?)(?=\n#{1,3}\s|\Z)", |
|
36
|
re.MULTILINE | re.DOTALL, |
|
37
|
) |
|
38
|
_DATE = re.compile(r"(?:Date|date)[:\s]+(\d{4}-\d{2}-\d{2})", re.MULTILINE) |
|
39
|
_ADR_NUMBER = re.compile(r"(\d+)", re.ASCII) |
|
40
|
|
|
41
|
|
|
42
|
class ADRIngester: |
|
43
|
""" |
|
44
|
Ingest ADR markdown files from a directory. |
|
45
|
|
|
46
|
Expected file naming convention: NNNN-short-title.md (e.g. 0001-use-falkordb.md) |
|
47
|
Each file is parsed for title, status, rationale, and date, then stored |
|
48
|
as a Decision node. |
|
49
|
""" |
|
50
|
|
|
51
|
def __init__(self, store: GraphStore) -> None: |
|
52
|
self.store = store |
|
53
|
|
|
54
|
# ── Public API ──────────────────────────────────────────────────────────── |
|
55
|
|
|
56
|
def ingest(self, adr_dir: str | Path) -> dict[str, Any]: |
|
57
|
""" |
|
58
|
Parse all ``*.md`` files in *adr_dir* and upsert Decision nodes. |
|
59
|
|
|
60
|
Returns a stats dict: decisions, skipped. |
|
61
|
""" |
|
62
|
adr_dir = Path(adr_dir) |
|
63
|
if not adr_dir.is_dir(): |
|
64
|
logger.warning("ADRIngester: %s is not a directory", adr_dir) |
|
65
|
return {"decisions": 0, "skipped": 0} |
|
66
|
|
|
67
|
md_files = sorted(adr_dir.glob("*.md")) |
|
68
|
decisions = 0 |
|
69
|
skipped = 0 |
|
70
|
|
|
71
|
for md_file in md_files: |
|
72
|
try: |
|
73
|
parsed = self._parse_adr(md_file) |
|
74
|
if parsed is None: |
|
75
|
skipped += 1 |
|
76
|
continue |
|
77
|
self.store.create_node(NodeLabel.Decision, parsed) |
|
78
|
decisions += 1 |
|
79
|
logger.debug("ADR: %s", parsed["name"]) |
|
80
|
except Exception as exc: # noqa: BLE001 |
|
81
|
logger.warning("ADRIngester: failed to parse %s: %s", md_file, exc) |
|
82
|
skipped += 1 |
|
83
|
|
|
84
|
stats = {"decisions": decisions, "skipped": skipped} |
|
85
|
logger.info("ADRIngester: %s", stats) |
|
86
|
return stats |
|
87
|
|
|
88
|
# ── Parsing ─────────────────────────────────────────────────────────────── |
|
89
|
|
|
90
|
def _parse_adr(self, path: Path) -> dict[str, Any] | None: |
|
91
|
""" |
|
92
|
Parse a single ADR markdown file. |
|
93
|
|
|
94
|
Returns a dict suitable for create_node(NodeLabel.Decision, ...) or |
|
95
|
None if the file does not look like an ADR. |
|
96
|
""" |
|
97
|
text = path.read_text(encoding="utf-8", errors="replace") |
|
98
|
|
|
99
|
# Title from first H1 heading |
|
100
|
h1_match = _H1.search(text) |
|
101
|
if h1_match is None: |
|
102
|
return None |
|
103
|
title = h1_match.group(1).strip() |
|
104
|
|
|
105
|
# Strip leading ADR number from title if present (e.g. "1. Use FalkorDB") |
|
106
|
title = re.sub(r"^\d+[.\s]+", "", title).strip() |
|
107
|
|
|
108
|
# Status |
|
109
|
status = "accepted" |
|
110
|
st_match = _STATUS.search(text) |
|
111
|
if st_match: |
|
112
|
raw_status = st_match.group(1).strip().lower().splitlines()[0] |
|
113
|
for known in ("proposed", "accepted", "deprecated", "superseded", "rejected"): |
|
114
|
if known in raw_status: |
|
115
|
status = known |
|
116
|
break |
|
117
|
|
|
118
|
# Rationale — prefer explicit ## Rationale section, fall back to ## Decision |
|
119
|
rationale = "" |
|
120
|
rat_match = _RATIONALE.search(text) or _RATIONALE_FALLBACK.search(text) |
|
121
|
if rat_match: |
|
122
|
rationale = rat_match.group(1).strip() |
|
123
|
|
|
124
|
# Date |
|
125
|
date = "" |
|
126
|
date_match = _DATE.search(text) |
|
127
|
if date_match: |
|
128
|
date = date_match.group(1) |
|
129
|
|
|
130
|
# Use the filename stem as the canonical name (stable, unique) |
|
131
|
name = path.stem |
|
132
|
|
|
133
|
return { |
|
134
|
"name": name, |
|
135
|
"description": title, |
|
136
|
"domain": "", |
|
137
|
"status": status, |
|
138
|
"rationale": rationale, |
|
139
|
"alternatives": "", |
|
140
|
"date": date, |
|
141
|
"file_path": str(path), |
|
142
|
} |
|
143
|
|