|
1
|
""" |
|
2
|
RepoIngester — walks a repository, parses source files with tree-sitter, |
|
3
|
and writes nodes + edges into the GraphStore. |
|
4
|
|
|
5
|
Supported languages (all via tree-sitter): |
|
6
|
Python .py |
|
7
|
TypeScript .ts .tsx |
|
8
|
JavaScript .js .jsx |
|
9
|
Go .go |
|
10
|
Rust .rs |
|
11
|
Java .java |
|
12
|
Kotlin .kt .kts |
|
13
|
C# .cs |
|
14
|
PHP .php |
|
15
|
Ruby .rb |
|
16
|
Swift .swift |
|
17
|
C .c .h |
|
18
|
C++ .cpp .hpp .cc .cxx |
|
19
|
|
|
20
|
Infrastructure-as-Code: |
|
21
|
HCL .tf .hcl (Terraform / OpenTofu) |
|
22
|
Puppet .pp |
|
23
|
Bash .sh .bash .zsh |
|
24
|
Ansible .yml .yaml (detected heuristically, not via extension) |
|
25
|
""" |
|
26
|
|
|
27
|
import hashlib |
|
28
|
import logging |
|
29
|
import time |
|
30
|
from pathlib import Path |
|
31
|
|
|
32
|
from navegador.graph import queries |
|
33
|
from navegador.graph.schema import NodeLabel |
|
34
|
from navegador.graph.store import GraphStore |
|
35
|
|
|
36
|
logger = logging.getLogger(__name__) |
|
37
|
|
|
38
|
# File extensions → language key |
|
39
|
LANGUAGE_MAP: dict[str, str] = { |
|
40
|
".py": "python", |
|
41
|
".ts": "typescript", |
|
42
|
".tsx": "typescript", |
|
43
|
".js": "javascript", |
|
44
|
".jsx": "javascript", |
|
45
|
".go": "go", |
|
46
|
".rs": "rust", |
|
47
|
".java": "java", |
|
48
|
".kt": "kotlin", |
|
49
|
".kts": "kotlin", |
|
50
|
".cs": "csharp", |
|
51
|
".php": "php", |
|
52
|
".rb": "ruby", |
|
53
|
".swift": "swift", |
|
54
|
".c": "c", |
|
55
|
".h": "c", |
|
56
|
".cpp": "cpp", |
|
57
|
".hpp": "cpp", |
|
58
|
".cc": "cpp", |
|
59
|
".cxx": "cpp", |
|
60
|
".tf": "hcl", |
|
61
|
".hcl": "hcl", |
|
62
|
".pp": "puppet", |
|
63
|
".sh": "bash", |
|
64
|
".bash": "bash", |
|
65
|
".zsh": "bash", |
|
66
|
} |
|
67
|
|
|
68
|
|
|
69
|
class RepoIngester: |
|
70
|
""" |
|
71
|
Parses a local code repository and populates a GraphStore. |
|
72
|
|
|
73
|
Usage: |
|
74
|
store = GraphStore.sqlite(".navegador/graph.db") |
|
75
|
ingester = RepoIngester(store) |
|
76
|
stats = ingester.ingest("/path/to/repo") |
|
77
|
|
|
78
|
Args: |
|
79
|
store: The graph store to write nodes and edges into. |
|
80
|
redact: When True, file contents are scanned for sensitive patterns |
|
81
|
(API keys, passwords, tokens, …) and any matches are replaced |
|
82
|
with ``[REDACTED]`` before the content is stored in graph nodes. |
|
83
|
""" |
|
84
|
|
|
85
|
def __init__(self, store: GraphStore, redact: bool = False) -> None: |
|
86
|
self.store = store |
|
87
|
self.redact = redact |
|
88
|
self._parsers: dict[str, "LanguageParser"] = {} |
|
89
|
if redact: |
|
90
|
from navegador.security import SensitiveContentDetector |
|
91
|
|
|
92
|
self._detector = SensitiveContentDetector() |
|
93
|
else: |
|
94
|
self._detector = None # type: ignore[assignment] |
|
95
|
|
|
96
|
def ingest( |
|
97
|
self, |
|
98
|
repo_path: str | Path, |
|
99
|
clear: bool = False, |
|
100
|
incremental: bool = False, |
|
101
|
) -> dict[str, int]: |
|
102
|
""" |
|
103
|
Ingest a repository into the graph. |
|
104
|
|
|
105
|
Args: |
|
106
|
repo_path: Path to the repository root. |
|
107
|
clear: If True, wipe the graph before ingesting. |
|
108
|
incremental: If True, skip files whose content hash hasn't changed. |
|
109
|
|
|
110
|
Returns: |
|
111
|
Dict with counts: files, functions, classes, edges, skipped. |
|
112
|
""" |
|
113
|
repo_path = Path(repo_path).resolve() |
|
114
|
if not repo_path.exists(): |
|
115
|
raise FileNotFoundError(f"Repository not found: {repo_path}") |
|
116
|
|
|
117
|
if clear: |
|
118
|
self.store.clear() |
|
119
|
|
|
120
|
# Create repository node |
|
121
|
self.store.create_node( |
|
122
|
NodeLabel.Repository, |
|
123
|
{ |
|
124
|
"name": repo_path.name, |
|
125
|
"path": str(repo_path), |
|
126
|
"file_path": "", |
|
127
|
}, |
|
128
|
) |
|
129
|
|
|
130
|
stats: dict[str, int] = { |
|
131
|
"files": 0, |
|
132
|
"functions": 0, |
|
133
|
"classes": 0, |
|
134
|
"edges": 0, |
|
135
|
"skipped": 0, |
|
136
|
} |
|
137
|
|
|
138
|
for source_file in self._iter_source_files(repo_path): |
|
139
|
language = LANGUAGE_MAP.get(source_file.suffix) |
|
140
|
if not language: |
|
141
|
continue |
|
142
|
|
|
143
|
rel_path = str(source_file.relative_to(repo_path)) |
|
144
|
content_hash = _file_hash(source_file) |
|
145
|
|
|
146
|
if incremental and self._file_unchanged(rel_path, content_hash): |
|
147
|
stats["skipped"] += 1 |
|
148
|
continue |
|
149
|
|
|
150
|
if incremental: |
|
151
|
self._clear_file_subgraph(rel_path) |
|
152
|
|
|
153
|
parse_path, effective_root = self._maybe_redact_to_tmp(source_file, repo_path) |
|
154
|
try: |
|
155
|
parser = self._get_parser(language) |
|
156
|
file_stats = parser.parse_file(parse_path, effective_root, self.store) |
|
157
|
stats["files"] += 1 |
|
158
|
stats["functions"] += file_stats.get("functions", 0) |
|
159
|
stats["classes"] += file_stats.get("classes", 0) |
|
160
|
stats["edges"] += file_stats.get("edges", 0) |
|
161
|
|
|
162
|
self._store_file_hash(rel_path, content_hash) |
|
163
|
except Exception: |
|
164
|
logger.exception("Failed to parse %s", source_file) |
|
165
|
finally: |
|
166
|
# Remove the temporary redacted directory if one was created |
|
167
|
if effective_root is not repo_path: |
|
168
|
import shutil |
|
169
|
|
|
170
|
shutil.rmtree(effective_root, ignore_errors=True) |
|
171
|
|
|
172
|
# Ansible pass — heuristically detect and parse Ansible YAML files |
|
173
|
self._ingest_ansible(repo_path, stats, incremental) |
|
174
|
|
|
175
|
logger.info( |
|
176
|
"Ingested %s: %d files, %d functions, %d classes, %d skipped", |
|
177
|
repo_path.name, |
|
178
|
stats["files"], |
|
179
|
stats["functions"], |
|
180
|
stats["classes"], |
|
181
|
stats["skipped"], |
|
182
|
) |
|
183
|
return stats |
|
184
|
|
|
185
|
def watch( |
|
186
|
self, |
|
187
|
repo_path: str | Path, |
|
188
|
interval: float = 2.0, |
|
189
|
callback=None, |
|
190
|
) -> None: |
|
191
|
""" |
|
192
|
Watch a repo for changes and re-ingest incrementally. |
|
193
|
|
|
194
|
Args: |
|
195
|
repo_path: Path to the repository root. |
|
196
|
interval: Seconds between polls. |
|
197
|
callback: Optional callable receiving stats dict after each cycle. |
|
198
|
If callback returns False, the watch loop stops. |
|
199
|
""" |
|
200
|
repo_path = Path(repo_path).resolve() |
|
201
|
if not repo_path.exists(): |
|
202
|
raise FileNotFoundError(f"Repository not found: {repo_path}") |
|
203
|
|
|
204
|
# Initial full ingest |
|
205
|
stats = self.ingest(repo_path, incremental=True) |
|
206
|
if callback and callback(stats) is False: |
|
207
|
return |
|
208
|
|
|
209
|
while True: |
|
210
|
time.sleep(interval) |
|
211
|
stats = self.ingest(repo_path, incremental=True) |
|
212
|
if callback and callback(stats) is False: |
|
213
|
return |
|
214
|
|
|
215
|
def _file_unchanged(self, rel_path: str, content_hash: str) -> bool: |
|
216
|
result = self.store.query(queries.FILE_HASH, {"path": rel_path}) |
|
217
|
rows = result.result_set or [] |
|
218
|
if not rows or rows[0][0] is None: |
|
219
|
return False |
|
220
|
return rows[0][0] == content_hash |
|
221
|
|
|
222
|
def _clear_file_subgraph(self, rel_path: str) -> None: |
|
223
|
self.store.query(queries.DELETE_FILE_SUBGRAPH, {"path": rel_path}) |
|
224
|
|
|
225
|
def _store_file_hash(self, rel_path: str, content_hash: str) -> None: |
|
226
|
self.store.query( |
|
227
|
"MATCH (f:File {path: $path}) SET f.content_hash = $hash", |
|
228
|
{"path": rel_path, "hash": content_hash}, |
|
229
|
) |
|
230
|
|
|
231
|
def _maybe_redact_to_tmp(self, source_file: Path, repo_root: Path) -> tuple[Path, Path]: |
|
232
|
""" |
|
233
|
If redaction is enabled, return a *(parse_path, effective_repo_root)* |
|
234
|
tuple where *parse_path* can be passed to ``parser.parse_file`` and |
|
235
|
``parse_path.relative_to(effective_repo_root)`` still yields the |
|
236
|
correct relative path for graph node naming. |
|
237
|
|
|
238
|
When redaction is disabled or the file has no sensitive content, both |
|
239
|
returned values are the originals unchanged. |
|
240
|
|
|
241
|
The caller is responsible for deleting the temp directory when it is |
|
242
|
no longer needed. |
|
243
|
""" |
|
244
|
if not self.redact or self._detector is None: |
|
245
|
return source_file, repo_root |
|
246
|
|
|
247
|
try: |
|
248
|
original = source_file.read_text(encoding="utf-8", errors="replace") |
|
249
|
except OSError: |
|
250
|
return source_file, repo_root |
|
251
|
|
|
252
|
redacted = self._detector.redact(original) |
|
253
|
if redacted == original: |
|
254
|
return source_file, repo_root |
|
255
|
|
|
256
|
# Mirror the file at the same relative path inside a temp directory so |
|
257
|
# that parse_path.relative_to(tmp_root) == source_file.relative_to(repo_root). |
|
258
|
import tempfile |
|
259
|
|
|
260
|
rel = source_file.relative_to(repo_root) |
|
261
|
tmp_root = Path(tempfile.mkdtemp()) |
|
262
|
tmp_file = tmp_root / rel |
|
263
|
tmp_file.parent.mkdir(parents=True, exist_ok=True) |
|
264
|
tmp_file.write_text(redacted, encoding="utf-8") |
|
265
|
return tmp_file, tmp_root |
|
266
|
|
|
267
|
def _iter_source_files(self, repo_path: Path): |
|
268
|
skip_dirs = { |
|
269
|
".git", |
|
270
|
".venv", |
|
271
|
"venv", |
|
272
|
"node_modules", |
|
273
|
"__pycache__", |
|
274
|
"dist", |
|
275
|
"build", |
|
276
|
".next", |
|
277
|
"target", # Rust / Java (Maven/Gradle) |
|
278
|
"vendor", # Go modules cache |
|
279
|
".gradle", # Gradle cache |
|
280
|
} |
|
281
|
for path in repo_path.rglob("*"): |
|
282
|
if path.is_file() and path.suffix in LANGUAGE_MAP: |
|
283
|
if not any(part in skip_dirs for part in path.parts): |
|
284
|
yield path |
|
285
|
|
|
286
|
def _ingest_ansible(self, repo_path: Path, stats: dict[str, int], incremental: bool) -> None: |
|
287
|
"""Detect and parse Ansible YAML files (playbooks, roles, tasks).""" |
|
288
|
from navegador.ingestion.ansible import AnsibleParser |
|
289
|
|
|
290
|
is_ansible_file = AnsibleParser.is_ansible_file |
|
291
|
|
|
292
|
ansible_parser: AnsibleParser | None = None |
|
293
|
|
|
294
|
for path in repo_path.rglob("*.yml"): |
|
295
|
if not path.is_file(): |
|
296
|
continue |
|
297
|
if any(part in (".git", ".venv", "venv", "node_modules") for part in path.parts): |
|
298
|
continue |
|
299
|
if not is_ansible_file(path, repo_path): |
|
300
|
continue |
|
301
|
|
|
302
|
rel_path = str(path.relative_to(repo_path)) |
|
303
|
content_hash = _file_hash(path) |
|
304
|
|
|
305
|
if incremental and self._file_unchanged(rel_path, content_hash): |
|
306
|
stats["skipped"] += 1 |
|
307
|
continue |
|
308
|
|
|
309
|
if incremental: |
|
310
|
self._clear_file_subgraph(rel_path) |
|
311
|
|
|
312
|
if ansible_parser is None: |
|
313
|
ansible_parser = AnsibleParser() |
|
314
|
try: |
|
315
|
file_stats = ansible_parser.parse_file(path, repo_path, self.store) |
|
316
|
stats["files"] += 1 |
|
317
|
stats["functions"] += file_stats.get("functions", 0) |
|
318
|
stats["classes"] += file_stats.get("classes", 0) |
|
319
|
stats["edges"] += file_stats.get("edges", 0) |
|
320
|
self._store_file_hash(rel_path, content_hash) |
|
321
|
except Exception: |
|
322
|
logger.exception("Failed to parse Ansible file %s", path) |
|
323
|
|
|
324
|
# Also check .yaml extension |
|
325
|
for path in repo_path.rglob("*.yaml"): |
|
326
|
if not path.is_file(): |
|
327
|
continue |
|
328
|
if any(part in (".git", ".venv", "venv", "node_modules") for part in path.parts): |
|
329
|
continue |
|
330
|
if not is_ansible_file(path, repo_path): |
|
331
|
continue |
|
332
|
|
|
333
|
rel_path = str(path.relative_to(repo_path)) |
|
334
|
content_hash = _file_hash(path) |
|
335
|
|
|
336
|
if incremental and self._file_unchanged(rel_path, content_hash): |
|
337
|
stats["skipped"] += 1 |
|
338
|
continue |
|
339
|
|
|
340
|
if incremental: |
|
341
|
self._clear_file_subgraph(rel_path) |
|
342
|
|
|
343
|
if ansible_parser is None: |
|
344
|
ansible_parser = AnsibleParser() |
|
345
|
try: |
|
346
|
file_stats = ansible_parser.parse_file(path, repo_path, self.store) |
|
347
|
stats["files"] += 1 |
|
348
|
stats["functions"] += file_stats.get("functions", 0) |
|
349
|
stats["classes"] += file_stats.get("classes", 0) |
|
350
|
stats["edges"] += file_stats.get("edges", 0) |
|
351
|
self._store_file_hash(rel_path, content_hash) |
|
352
|
except Exception: |
|
353
|
logger.exception("Failed to parse Ansible file %s", path) |
|
354
|
|
|
355
|
def _get_parser(self, language: str) -> "LanguageParser": |
|
356
|
if language not in self._parsers: |
|
357
|
if language == "python": |
|
358
|
from navegador.ingestion.python import PythonParser |
|
359
|
|
|
360
|
self._parsers[language] = PythonParser() |
|
361
|
elif language in ("typescript", "javascript"): |
|
362
|
from navegador.ingestion.typescript import TypeScriptParser |
|
363
|
|
|
364
|
self._parsers[language] = TypeScriptParser(language) |
|
365
|
elif language == "go": |
|
366
|
from navegador.ingestion.go import GoParser |
|
367
|
|
|
368
|
self._parsers[language] = GoParser() |
|
369
|
elif language == "rust": |
|
370
|
from navegador.ingestion.rust import RustParser |
|
371
|
|
|
372
|
self._parsers[language] = RustParser() |
|
373
|
elif language == "java": |
|
374
|
from navegador.ingestion.java import JavaParser |
|
375
|
|
|
376
|
self._parsers[language] = JavaParser() |
|
377
|
elif language == "kotlin": |
|
378
|
from navegador.ingestion.kotlin import KotlinParser |
|
379
|
|
|
380
|
self._parsers[language] = KotlinParser() |
|
381
|
elif language == "csharp": |
|
382
|
from navegador.ingestion.csharp import CSharpParser |
|
383
|
|
|
384
|
self._parsers[language] = CSharpParser() |
|
385
|
elif language == "php": |
|
386
|
from navegador.ingestion.php import PHPParser |
|
387
|
|
|
388
|
self._parsers[language] = PHPParser() |
|
389
|
elif language == "ruby": |
|
390
|
from navegador.ingestion.ruby import RubyParser |
|
391
|
|
|
392
|
self._parsers[language] = RubyParser() |
|
393
|
elif language == "swift": |
|
394
|
from navegador.ingestion.swift import SwiftParser |
|
395
|
|
|
396
|
self._parsers[language] = SwiftParser() |
|
397
|
elif language == "c": |
|
398
|
from navegador.ingestion.c import CParser |
|
399
|
|
|
400
|
self._parsers[language] = CParser() |
|
401
|
elif language == "cpp": |
|
402
|
from navegador.ingestion.cpp import CppParser |
|
403
|
|
|
404
|
self._parsers[language] = CppParser() |
|
405
|
elif language == "hcl": |
|
406
|
from navegador.ingestion.hcl import HCLParser |
|
407
|
|
|
408
|
self._parsers[language] = HCLParser() |
|
409
|
elif language == "puppet": |
|
410
|
from navegador.ingestion.puppet import PuppetParser |
|
411
|
|
|
412
|
self._parsers[language] = PuppetParser() |
|
413
|
elif language == "bash": |
|
414
|
from navegador.ingestion.bash import BashParser |
|
415
|
|
|
416
|
self._parsers[language] = BashParser() |
|
417
|
else: |
|
418
|
raise ValueError(f"Unsupported language: {language}") |
|
419
|
return self._parsers[language] |
|
420
|
|
|
421
|
|
|
422
|
def _file_hash(path: Path) -> str: |
|
423
|
"""SHA-256 content hash for a file.""" |
|
424
|
h = hashlib.sha256() |
|
425
|
h.update(path.read_bytes()) |
|
426
|
return h.hexdigest() |
|
427
|
|
|
428
|
|
|
429
|
class LanguageParser: |
|
430
|
"""Base class for language-specific AST parsers.""" |
|
431
|
|
|
432
|
def parse_file(self, path: Path, repo_root: Path, store: GraphStore) -> dict[str, int]: |
|
433
|
raise NotImplementedError |
|
434
|
|