File size: 45,251 Bytes
2ecdf7d 0a46902 2ecdf7d 5084a8e 2ecdf7d 0a46902 2ecdf7d 2c1d035 2ecdf7d 0a46902 b8e40db 0a46902 2ecdf7d 5084a8e e16bd16 5084a8e 2ecdf7d 46fa98c 2ecdf7d 46fa98c 43acb00 46fa98c 43acb00 cb22a66 46fa98c cb22a66 1787788 46fa98c 1787788 2ecdf7d 46fa98c cb22a66 46fa98c cb22a66 1787788 46fa98c 1787788 2ecdf7d 46fa98c 2ecdf7d 8e2dddb 2ecdf7d 2c1d035 2ecdf7d 2787e18 70e6f09 939712a 70e6f09 2787e18 5f52d90 2787e18 1787788 46fa98c 1787788 8e2dddb 2ecdf7d 78482d8 46fa98c 2ecdf7d 78482d8 2ecdf7d 5fa3433 46fa98c 2feb0de 46fa98c 2feb0de 43acb00 46fa98c 43acb00 46fa98c 43acb00 46fa98c 43acb00 8e2dddb 2ecdf7d 2feb0de 2ecdf7d 78482d8 2ecdf7d 78482d8 2ecdf7d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 | from __future__ import annotations
import asyncio
import contextlib
import gzip
import io
import json
import random
import re
import struct
import threading
import time
import unittest
import zipfile
from pathlib import Path
from types import SimpleNamespace
from unittest.mock import AsyncMock, patch
from urllib.parse import quote
IMPORT_ERROR = None
try:
from backend import app
import httpx
except ModuleNotFoundError as exc:
app = None
IMPORT_ERROR = exc
from tests.test_async_proxy import FakeResponse, FakeSession
SOURCE = "https://huggingface.co/datasets/VoiceOfML/Test/resolve/main/book29.txt"
OBJECT = "objects/aa/" + "a" * 64
BUCKET_PATH = OBJECT + "/" + "b" * 16 + "/page-manifest.json"
def make_zip(files=None, comment=b"", zip64=False):
output = io.BytesIO()
with zipfile.ZipFile(output, "w", compression=zipfile.ZIP_DEFLATED) as archive:
for name, content in (files if files is not None else {"image.svg": b"<svg/>"}).items():
archive.writestr(name, content)
archive.comment = comment
raw = output.getvalue()
if not zip64:
return raw
offset = len(raw) - 22 - len(comment)
end = list(struct.unpack("<4s4H2LH", raw[offset:offset + 22]))
record64 = struct.pack("<4sQ2H2L4Q", b"PK\x06\x06", 44, 45, 45, 0, 0, end[4], end[4], end[5], end[6])
locator = struct.pack("<4sLQL", b"PK\x06\x07", 0, offset, 1)
end[3:7] = [0xffff, 0xffff, 0xffffffff, 0xffffffff]
return raw[:offset] + record64 + locator + struct.pack("<4s4H2LH", *end) + comment
@unittest.skipIf(app is None, f"backend dependencies unavailable: {IMPORT_ERROR}")
class ReaderArchiveValidationTests(unittest.TestCase):
def test_canonical_paths_preserve_character_encoder_semantics(self):
prefix = "https://huggingface.co/datasets/VoiceOfML/Test/resolve/main/"
def reference(value):
def replace(match):
token = match.group()
if re.fullmatch(r"%[0-9a-fA-F]{2}", token):
char = chr(int(token[1:], 16))
return char if re.fullmatch(r"[A-Za-z0-9._~-]", char) else token.upper()
return quote(token, safe="")
return prefix + re.sub(r"%[0-9a-fA-F]{2}|[^A-Za-z0-9._~/-]", replace, value)
rng = random.Random(20260927)
tokens = [chr(i) for i in range(128) if chr(i) not in "?#"]
tokens += ["\u4e66", "\u00e9", "\U0001f600", "%2f", "%25", "%41", "%7e", "%G0", "%"]
paths = ["%" + first + second for i in range(256)
for first in {f"{i:02x}"[0], f"{i:02X}"[0]}
for second in {f"{i:02x}"[1], f"{i:02X}"[1]}]
paths += ["".join(rng.choices(tokens, k=40)) for _ in range(256)]
for path in paths:
self.assertEqual(app.canonical_reader_source_url(prefix + path), reference(path))
for value in ("https://example.com/%ab", prefix + "a?raw=%ab#fragment"):
self.assertEqual(app.canonical_reader_source_url(value), value)
def test_base36_independent_fixed_vectors(self):
for value, expected in ((0, "0"), (35, "z"), (36, "10"), (72, "20"), (1296, "100"), (2**64 - 1, "3w5e11264sgsf")):
with self.subTest(value=value):
self.assertEqual(app._base36(value), expected)
self.assertEqual(app.reader_short_id(SOURCE), "0ym7e7wbcxau0")
vectors = json.loads(Path(__file__).with_name("reader-contract-vectors.json").read_text())
for value, expected in vectors["ids"]:
self.assertEqual(app.reader_short_id(value), expected)
prefix = "https://huggingface.co/datasets/VoiceOfML/VOMEBOOK/resolve/main/"
for value, expected in vectors["canonicalPaths"]:
self.assertEqual(app.canonical_reader_source_url(prefix + value), prefix + expected)
self.assertEqual(app.canonical_reader_source_url(prefix + expected), prefix + expected)
self.assertEqual(app.reader_short_id(prefix + value), app.reader_short_id(prefix + expected))
self.assertEqual(app.reader_short_id(prefix + vectors["reportedBook"]["path"]), vectors["reportedBook"]["id"])
for record, expected in vectors["ocr"]:
self.assertEqual(app.build_txt_relative_path(record), expected)
def test_valid_standard_zip64_comments_and_prepended_data(self):
for zip64 in (False, True):
for comment in (b"", b"archive comment"):
for prefix in (b"", b"self-extracting prefix"):
with self.subTest(zip64=zip64, comment=comment, prefix=prefix):
raw = prefix + make_zip(comment=comment, zip64=zip64)
app.validate_reader_archive(raw)
self.assertEqual(app.read_reader_archive_member(raw, "image.svg"), b"<svg/>")
app.validate_reader_archive(make_zip(files={}))
def test_gzip_wrapper_has_bounded_expansion(self):
raw = make_zip()
self.assertEqual(app.prepare_reader_archive(gzip.compress(raw)), raw)
with patch.object(app, "READER_RESOURCE_MAX_ARCHIVE_BYTES", len(raw) - 1):
with self.assertRaises(app.ReaderArchiveLimitError):
app.prepare_reader_archive(gzip.compress(raw))
with self.assertRaises(app.ReaderArchiveInvalidError):
app.prepare_reader_archive(b"\x1f\x8btruncated")
def test_directory_limits_checked_before_zipfile_construction(self):
for zip64 in (False, True):
raw = make_zip(zip64=zip64)
for limit in ("READER_ARCHIVE_MAX_CENTRAL_DIRECTORY_BYTES", "READER_ARCHIVE_MAX_ENTRIES"):
with self.subTest(zip64=zip64, limit=limit), patch.object(app, limit, 0), patch.object(app.zipfile, "ZipFile") as parser:
with self.assertRaises(app.ReaderArchiveLimitError):
app.validate_reader_archive(raw)
parser.assert_not_called()
def test_malformed_end_records_and_multidisk_archives(self):
raw = make_zip()
variants = [raw[:-1], b"PKnot-a-zip"]
for offset, fmt, value in ((4, "H", 1), (10, "H", 2), (16, "L", len(raw) + 1), (20, "H", 8)):
bad = bytearray(raw)
struct.pack_into("<" + fmt, bad, len(raw) - 22 + offset, value)
variants.append(bytes(bad))
raw64 = make_zip(zip64=True)
bad64 = bytearray(raw64)
struct.pack_into("<L", bad64, len(raw64) - 22 - 20 + 4, 1)
variants.extend((bytes(bad64), raw64[:-23] + raw64[-22:]))
for index, bad in enumerate(variants):
with self.subTest(index=index), self.assertRaises(app.ReaderArchiveInvalidError):
app.validate_reader_archive(bad)
def test_unsafe_members_and_compression_limits(self):
for name in ("../escape", "/absolute", "a/../escape", "a//escape", "a\\..\\escape"):
with self.subTest(name=name), self.assertRaises(app.ReaderArchiveInvalidError):
app.validate_reader_archive(make_zip({name: b"content"}))
with self.assertRaises(app.ReaderArchiveLimitError):
app.validate_reader_archive(make_zip({"bomb": b"x" * 100000}))
with patch.object(app, "READER_ARCHIVE_MAX_EXPANDED_BYTES", 3), self.assertRaises(app.ReaderArchiveLimitError):
app.validate_reader_archive(make_zip({"small": b"1234"}))
with self.assertRaises(app.ReaderArchiveInvalidError):
app.validate_reader_archive(make_zip().replace(b"image.svg", b"im\x00ge.svg"))
def test_accepted_sidecar_paths_match_the_delivery_validator(self):
files = {}
for mode, filenames in app.READER_ASSET_FILENAMES.items():
for filename in filenames:
for directory in ("", "converter-v1/", "b" * 16 + "/", "b" * 16 + "/converter-v1/"):
path = OBJECT + "/" + directory + filename
# Page manifests have only an optional version directory.
if filename == "page-manifest.json" and "converter" in directory:
continue
if filename in {"document.vcf", "document.ini"}:
continue
files[path] = {"s": 2, "m": mode, "p": path}
for filename in ("document.vcf", "document.ini"):
path = "documents/text/native/" + "d" * 64 + "/" + filename
files[path] = {"s": 2, "m": "t", "p": path, "b": "vomebook/reader-assets-v2"}
files["bucket"] = {"s": 2, "m": "p", "p": BUCKET_PATH, "b": "vomebook/pdf-pages-v2"}
ebook_chapter_path = "chapters/ebook/epub/" + "c" * 64 + "/" + "b" * 16 + "/chapter-manifest.json"
files["ebook-chapters"] = {
"s": 2, "m": "e", "p": ebook_chapter_path, "b": "vomebook/reader-assets-v2",
"c": ebook_chapter_path, "cb": "vomebook/reader-assets-v2",
}
files["spreadsheet-pages"] = {
"s": 2, "m": "p", "p": OBJECT + "/1234567890abcdef/page-manifest.json",
"b": "vomebook/pdf-pages-v2",
}
files["native-audio-bucket"] = {
"s": 2, "m": "a", "p": "media/audio/native/" + "a" * 64 + "/audio.flac",
"b": "vomebook/reader-assets-v2",
}
decoded = app.decode_reader_assets(gzip.compress(json.dumps({"v": 1, "f": files}).encode()))["f"]
self.assertEqual(set(decoded), set(files))
self.assertEqual(decoded["ebook-chapters"]["cb"], "vomebook/reader-assets-v2")
self.assertEqual(
app.reader_asset_url(decoded["spreadsheet-pages"]),
"https://voiceofml-search.hf.space/api/reader-bucket-resource?path="
+ OBJECT + "/1234567890abcdef/page-manifest.json",
)
self.assertEqual(
app.reader_asset_url(decoded["native-audio-bucket"]),
"https://voiceofml-search.hf.space/api/reader-bucket-resource?path="
+ "media/audio/native/" + "a" * 64 + "/audio.flac",
)
self.assertEqual(app.READER_RESOURCE_TYPES["flac"], "audio/flac")
self.assertEqual(app.reader_extension_for_asset(decoded["native-audio-bucket"], "flac"), "audio")
for entry in decoded.values():
if entry.get("b"):
self.assertIsNotNone(app.READER_BUCKET_PATH_RE.fullmatch(entry["p"]))
else:
self.assertIsNotNone(app.READER_ASSET_PATH_RE.fullmatch(entry["p"]))
for bad in (
{"s": 2, "m": "e", "p": OBJECT + "/document.pdf"},
{"s": 2, "m": "p", "p": OBJECT + "/document.pdf", "b": "vomebook/pdf-pages"},
{"s": 2, "m": "p", "p": OBJECT + "/page-manifest.json", "b": "vomebook/pdf-pages"},
):
decoded = app.decode_reader_assets(gzip.compress(json.dumps({"v": 1, "f": {"bad": bad}}).encode()))
self.assertEqual(decoded["f"], {})
def test_pdf_primary_document_survives_decoder_and_metadata_projection(self):
path = "derived/test/" + "c" * 32 + "/document.pdf"
entry = {"s": 2, "m": "p", "p": BUCKET_PATH, "b": "vomebook/pdf-pages-v2",
"pd": path, "pdb": "vomebook/pdf-pages-v2"}
payload = gzip.compress(json.dumps({"v": 1, "f": {"book": entry}}).encode())
self.assertEqual(app.decode_reader_assets(payload)["f"]["book"], entry)
record = {"Repo": "VoiceOfML/Test", "File": "converted", "Extension": "caj", "Folder": []}
resolved = app.reader_resolved_record(record, entry)
self.assertEqual(resolved["extension"], "pdf-pages")
self.assertEqual(resolved["pdf_document"],
"https://voiceofml-search.hf.space/api/reader-bucket-resource?path=" + path)
enriched = app.add_reader_assets({"results": [record]},
{"f": {app.build_reader_asset_key(record): entry}})["results"][0]
self.assertEqual(enriched["ReaderPdfDocument"], resolved["pdf_document"])
for invalid in ({**entry, "pdb": "vomebook/pdf-pages"},
{**entry, "pdb": []},
{**entry, "pd": "../document.pdf"}, {**entry, "pd": BUCKET_PATH}):
raw = gzip.compress(json.dumps({"v": 1, "f": {"book": invalid}}).encode())
self.assertEqual(app.decode_reader_assets(raw)["f"], {})
@unittest.skipIf(app is None, f"backend dependencies unavailable: {IMPORT_ERROR}")
class ReaderBackendTests(unittest.IsolatedAsyncioTestCase):
async def asyncSetUp(self):
self.saved_state = dict(app.app.state._state)
app.initialize_reader_archive_state(app.app)
app.initialize_reader_bucket_state(app.app)
app.app.state.upstream_semaphore = asyncio.Semaphore(1)
self.stack = contextlib.ExitStack()
self.record = {"Repo": "VoiceOfML/Test", "File": "book29", "Extension": "txt", "Folder": []}
self.assets = {"v": 1, "f": {}}
self.stack.enter_context(patch.object(app, "get_reader_assets", AsyncMock(return_value=self.assets)))
self.stack.enter_context(patch.object(app, "records", [self.record]))
self.stack.enter_context(patch.object(app, "reader_record_key_indices", {app.build_reader_asset_key(self.record): 0}))
self.stack.enter_context(patch.object(app, "reader_source_id_indices", {"0ym7e7wbcxau0": [0]}))
self.stack.enter_context(patch("backend.upstream_http.upstream_redirect_cache", {}))
async def asyncTearDown(self):
await app.shutdown_reader_archive_state(app.app)
await app.shutdown_reader_bucket_state(app.app)
self.stack.close()
app.app.state._state.clear()
app.app.state._state.update(self.saved_state)
def session(self, *responses):
session = FakeSession(responses)
app.app.state.http_session = session
return session
async def test_bucket_image_disconnect_stops_stream_and_releases_slot(self):
path = "pages/image/bmp/" + "a" * 64 + "/pages/page-000001.webp"
upstream = FakeResponse(chunks=(b"image",))
self.session(upstream)
with patch.object(app, "get_reader_bucket_resource", AsyncMock()) as buffered:
response = await app.api_reader_bucket_resource(path, SimpleNamespace(method="GET", headers={}))
buffered.assert_not_awaited()
started = asyncio.Event()
async def send(message):
started.set()
await asyncio.Event().wait()
async def receive():
await started.wait()
return {"type": "http.disconnect"}
await asyncio.wait_for(response({"type": "http", "asgi": {"spec_version": "2.0"}}, receive, send), 1)
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_cold_ocr_parse_is_shared_between_concurrent_readers(self):
path = "derived/test/" + "a" * 32 + "/ocr/book-text.json.gz"
pages = [{"page": 1, "text": "ok", "folded": "ok", "spans": [], "layout": {}}]
with patch.object(app, "_load_reader_ocr_book", AsyncMock(return_value=pages)) as loader:
first, second = await asyncio.gather(
app.get_reader_ocr_book(path), app.get_reader_ocr_book(path)
)
self.assertIs(first, second)
loader.assert_awaited_once_with(path)
async def test_uncached_bucket_bodies_stream_without_filesystem_worker(self):
prefix = 'a' * 64
paths = [f'documents/text/txt/{prefix}/document.txt',
f'documents/web/html/{prefix}/document.html',
f'documents/office/docx/{prefix}/document.docx',
f'chapters/ebook/epub/{prefix}/'+'b'*16+'/chapters/chapter-0001.xhtml',
f'media/swf/native/{prefix}/document.swf']
for path in paths:
with self.subTest(path=path):
upstream = FakeResponse(chunks=(b'body',))
self.session(upstream)
with patch.object(app, 'get_reader_bucket_resource', AsyncMock()) as buffered:
response = await app.api_reader_bucket_resource(path, SimpleNamespace(method='GET', headers={}))
buffered.assert_not_awaited()
self.assertEqual(b''.join([chunk async for chunk in response.body_iterator]), b'body')
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_disconnect_while_waiting_for_headers_releases_transport_slot(self):
self.session()
disconnected = asyncio.Event()
started = asyncio.Event()
cancelled = asyncio.Event()
async def open_upstream(*args, **kwargs):
started.set()
try:
await asyncio.Event().wait()
finally:
cancelled.set()
request = SimpleNamespace(method="GET", headers={},
is_disconnected=AsyncMock(side_effect=lambda: disconnected.is_set()))
with patch.object(app, "open_download_response", side_effect=open_upstream):
task = asyncio.create_task(app.stream_reader_document(request, SOURCE))
await asyncio.wait_for(started.wait(), 1)
disconnected.set()
with self.assertRaises(asyncio.CancelledError):
await asyncio.wait_for(task, 1)
self.assertTrue(cancelled.is_set())
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_disconnect_while_queued_does_not_start_upstream(self):
await app.app.state.upstream_semaphore.acquire()
request = SimpleNamespace(method="GET", headers={}, is_disconnected=AsyncMock(return_value=True))
with patch.object(app, "open_download_response", AsyncMock()) as opened:
with self.assertRaises(asyncio.CancelledError):
await asyncio.wait_for(app.stream_reader_document(request, SOURCE), 1)
opened.assert_not_awaited()
self.assertEqual(app.app.state.upstream_semaphore._value, 0)
app.app.state.upstream_semaphore.release()
async def test_cached_range_finishes_when_disconnect_probe_swallows_cancellation(self):
app.app.state.reader_range_cache = app.OrderedDict()
app.app.state.reader_range_cache_bytes = 3
key = (SOURCE, 'bytes=0-2', app.READER_DOCUMENT_MAX_BYTES, None)
app.app.state.reader_range_cache[key] = (time.monotonic(), b'pdf', {
'Content-Type':'application/pdf', 'Content-Length':'3', 'Content-Range':'bytes 0-2/3'})
async def probe():
try:
await asyncio.Event().wait()
except asyncio.CancelledError:
# Starlette's Request probe owns a cancellation scope.
return False
request = SimpleNamespace(method='GET', headers={'Range':'bytes=0-2'}, is_disconnected=probe)
with patch.object(app, 'open_download_response', AsyncMock()) as opened:
response = await asyncio.wait_for(app.stream_reader_document(request, SOURCE), .5)
self.assertEqual(response.status_code, 206)
self.assertEqual(response.body, b'pdf')
self.assertEqual(response.headers['cache-control'], 'no-store')
opened.assert_not_awaited()
async def test_cold_resolve_waits_for_conversion_and_empty_failure_is_not_cacheable(self):
with patch.object(app, "reader_assets_cache", {"v": 1, "f": {}}):
await app.api_reader_resolve("0ym7e7wbcxau0")
app.get_reader_assets.assert_awaited_with(wait=True)
self.assets['f'][app.build_reader_asset_key(self.record)] = {
's':2, 'm':'p', 'p':'documents/pdf/ps/'+'a'*64+'/document.pdf',
'b':'vomebook/reader-assets-v2'}
with patch.object(app, "reader_assets_cache", self.assets):
response = await app.api_reader_resolve("0ym7e7wbcxau0")
app.get_reader_assets.assert_awaited_with(wait=False)
self.assertEqual(json.loads(response.body)['extension'], 'pdf')
self.assets['f'].clear()
response = await app.api_reader_resolve("0ym7e7wbcxau0")
self.assertEqual(response.headers['cache-control'], 'no-store')
async def test_bucket_audio_resource_uses_audio_mime_type(self):
path = "media/audio/native/" + "a" * 64 + "/audio.flac"
with patch.object(app, "get_reader_bucket_resource", AsyncMock(return_value=b"flac")):
response = await app.api_reader_bucket_resource(path)
self.assertEqual(response.media_type, "audio/flac")
self.assertEqual(response.body, b"flac")
async def test_bucket_image_disconnect_stops_stream_and_releases_slot(self):
path = "pages/image/bmp/" + "a" * 64 + "/pages/page-000001.webp"
upstream = FakeResponse(chunks=(b"image",))
self.session(upstream)
with patch.object(app, "get_reader_bucket_resource", AsyncMock()) as buffered:
response = await app.api_reader_bucket_resource(path, SimpleNamespace(method="GET", headers={}))
buffered.assert_not_awaited()
started = asyncio.Event()
async def send(message):
started.set()
await asyncio.Event().wait()
async def receive():
await started.wait()
return {"type": "http.disconnect"}
await asyncio.wait_for(response({"type": "http", "asgi": {"spec_version": "2.0"}}, receive, send), 1)
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_uncached_bucket_bodies_stream_without_filesystem_worker(self):
prefix = 'a' * 64
paths = [f'documents/text/txt/{prefix}/document.txt',
f'documents/web/html/{prefix}/document.html',
f'documents/office/docx/{prefix}/document.docx',
f'chapters/ebook/epub/{prefix}/'+'b'*16+'/chapters/chapter-0001.xhtml',
f'media/swf/native/{prefix}/document.swf']
for path in paths:
with self.subTest(path=path):
upstream = FakeResponse(chunks=(b'body',))
self.session(upstream)
with patch.object(app, 'get_reader_bucket_resource', AsyncMock()) as buffered:
response = await app.api_reader_bucket_resource(path, SimpleNamespace(method='GET', headers={}))
buffered.assert_not_awaited()
self.assertEqual(b''.join([chunk async for chunk in response.body_iterator]), b'body')
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_disconnect_while_waiting_for_headers_releases_transport_slot(self):
self.session()
disconnected = asyncio.Event()
started = asyncio.Event()
cancelled = asyncio.Event()
async def open_upstream(*args, **kwargs):
started.set()
try:
await asyncio.Event().wait()
finally:
cancelled.set()
request = SimpleNamespace(method="GET", headers={},
is_disconnected=AsyncMock(side_effect=lambda: disconnected.is_set()))
with patch.object(app, "open_download_response", side_effect=open_upstream):
task = asyncio.create_task(app.stream_reader_document(request, SOURCE))
await asyncio.wait_for(started.wait(), 1)
disconnected.set()
with self.assertRaises(asyncio.CancelledError):
await asyncio.wait_for(task, 1)
self.assertTrue(cancelled.is_set())
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_disconnect_while_queued_does_not_start_upstream(self):
await app.app.state.upstream_semaphore.acquire()
request = SimpleNamespace(method="GET", headers={}, is_disconnected=AsyncMock(return_value=True))
with patch.object(app, "open_download_response", AsyncMock()) as opened:
with self.assertRaises(asyncio.CancelledError):
await asyncio.wait_for(app.stream_reader_document(request, SOURCE), 1)
opened.assert_not_awaited()
self.assertEqual(app.app.state.upstream_semaphore._value, 0)
app.app.state.upstream_semaphore.release()
async def test_cached_range_finishes_when_disconnect_probe_swallows_cancellation(self):
app.app.state.reader_range_cache = app.OrderedDict()
app.app.state.reader_range_cache_bytes = 3
key = (SOURCE, 'bytes=0-2', app.READER_DOCUMENT_MAX_BYTES, None)
app.app.state.reader_range_cache[key] = (time.monotonic(), b'pdf', {
'Content-Type':'application/pdf', 'Content-Length':'3', 'Content-Range':'bytes 0-2/3'})
async def probe():
try:
await asyncio.Event().wait()
except asyncio.CancelledError:
# Starlette's Request probe owns a cancellation scope.
return False
request = SimpleNamespace(method='GET', headers={'Range':'bytes=0-2'}, is_disconnected=probe)
with patch.object(app, 'open_download_response', AsyncMock()) as opened:
response = await asyncio.wait_for(app.stream_reader_document(request, SOURCE), .5)
self.assertEqual(response.status_code, 206)
self.assertEqual(response.body, b'pdf')
self.assertEqual(response.headers['cache-control'], 'no-store')
opened.assert_not_awaited()
async def content(self, upstream, method="GET"):
self.session(upstream)
return await app.api_reader_content(SimpleNamespace(method=method, headers={}), SOURCE, None)
def assert_isolated(self, response):
policy = response.headers["content-security-policy"]
self.assertIn("sandbox;", policy)
self.assertIn("default-src 'none'", policy)
self.assertNotIn("allow-scripts", policy)
self.assertNotIn("allow-same-origin", policy)
self.assertEqual(response.headers["x-content-type-options"], "nosniff")
async def test_valid_resource_roundtrip_and_literal_http_member_names(self):
files = {"literal%20name.svg": b"percent", "literal name.svg": b"space", "percent%25.svg": b"percent25", "\u4e66.svg": b"unicode"}
upstream = FakeResponse(chunks=(make_zip(files),))
session = self.session(upstream)
async with httpx.AsyncClient(transport=httpx.ASGITransport(app=app.app), base_url="http://test") as client:
for name, body in files.items():
response = await client.get("/api/reader-resource", params={"book": SOURCE, "path": name})
self.assertEqual(response.status_code, 200, response.text)
self.assertEqual(response.content, body)
self.assert_isolated(response)
for unsafe in ("../escape", "/absolute", "a//b", "a/../b", "nul\x00"):
response = await client.get("/api/reader-resource", params={"book": SOURCE, "path": unsafe})
self.assertEqual(response.status_code, 400)
self.assertEqual(len(session.urls), 1)
self.assertTrue(upstream.released)
async def test_archive_failures_are_not_cached_and_member_limits_are_reported(self):
invalid = FakeResponse(chunks=(b"not-a-zip",))
valid = FakeResponse(chunks=(make_zip(),))
self.session(invalid, valid)
self.assertEqual((await app.api_reader_resource(SOURCE, "image.svg")).status_code, 400)
self.assertEqual(app.app.state.reader_archive_cache, {})
self.assertEqual((await app.api_reader_resource(SOURCE, "missing.svg")).status_code, 404)
with patch.object(app, "READER_RESOURCE_MAX_BYTES", 2):
self.assertEqual((await app.api_reader_resource(SOURCE, "image.svg")).status_code, 413)
self.assertTrue(invalid.released and valid.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_archive_cold_requests_share_load_and_cancellation_preserves_waiter(self):
started, release = asyncio.Event(), asyncio.Event()
raw = make_zip()
class SlowResponse(FakeResponse):
def iter_chunked(self, _size):
async def chunks():
started.set()
await release.wait()
yield raw
return chunks()
upstream = SlowResponse()
session = self.session(upstream)
first = asyncio.create_task(app.get_reader_archive(SOURCE))
await asyncio.wait_for(started.wait(), 1)
second = asyncio.create_task(app.get_reader_archive(SOURCE))
await asyncio.sleep(0)
first.cancel()
with self.assertRaises(asyncio.CancelledError):
await first
release.set()
self.assertEqual(await asyncio.wait_for(second, 2), raw)
self.assertEqual(len(session.urls), 1)
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.reader_archive_inflight, {})
async def test_archive_cache_total_bytes_expiry_and_oversize_admission(self):
raw = make_zip()
session = self.session(*(FakeResponse(chunks=(raw,)) for _ in range(6)))
with patch.object(app, "READER_ARCHIVE_CACHE_MAX_BYTES", len(raw) * 2):
for index in range(3):
await app.get_reader_archive(SOURCE + str(index))
cache = app.app.state.reader_archive_cache
self.assertNotIn(SOURCE + "0", cache)
self.assertEqual(sum(len(item[1]) for item in cache.values()), len(raw) * 2)
cache[SOURCE + "2"] = (time.monotonic() - app.READER_ARCHIVE_CACHE_TTL_SECONDS - 1, raw)
await app.get_reader_archive(SOURCE + "2")
self.assertEqual(len(session.urls), 4)
with patch.object(app, "READER_ARCHIVE_CACHE_MAX_BYTES", len(raw) - 1):
await app.get_reader_archive(SOURCE + "big")
await app.get_reader_archive(SOURCE + "big")
self.assertNotIn(SOURCE + "big", cache)
self.assertEqual(len(session.urls), 6)
async def test_archive_shutdown_releases_active_download(self):
started = asyncio.Event()
class BlockedResponse(FakeResponse):
def iter_chunked(self, _size):
async def chunks():
started.set()
await asyncio.Event().wait()
yield b""
return chunks()
upstream = BlockedResponse()
self.session(upstream)
request = asyncio.create_task(app.get_reader_archive(SOURCE))
await asyncio.wait_for(started.wait(), 1)
with patch.object(app, "READER_ARCHIVE_MAX_INFLIGHT", 1):
rejected = await app.api_reader_resource(SOURCE + "other", "image.svg")
self.assertEqual(rejected.status_code, 503)
await app.shutdown_reader_archive_state(app.app)
with self.assertRaises(asyncio.CancelledError):
await request
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
self.assertEqual(app.app.state.reader_archive_cache, {})
async def test_worker_cancellation_retains_slot_until_thread_finishes(self):
app.app.state.reader_archive_worker_semaphore = asyncio.Semaphore(1)
started, release, second_started = threading.Event(), threading.Event(), threading.Event()
def blocking():
started.set()
release.wait(2)
first = asyncio.create_task(app.run_reader_archive_worker(blocking))
self.assertTrue(await asyncio.to_thread(started.wait, 1))
first.cancel()
second = asyncio.create_task(app.run_reader_archive_worker(second_started.set))
try:
with self.assertRaises(asyncio.CancelledError):
await asyncio.wait_for(first, 1)
await asyncio.sleep(0.02)
self.assertEqual(len(app.app.state.reader_archive_workers), 1)
self.assertFalse(second_started.is_set())
finally:
release.set()
results = await asyncio.gather(first, second, return_exceptions=True)
self.assertIsInstance(results[0], asyncio.CancelledError)
self.assertTrue(second_started.is_set())
async def test_archive_shutdown_waits_for_abandoned_worker(self):
started, release = threading.Event(), threading.Event()
def blocking():
started.set()
release.wait(2)
caller = asyncio.create_task(app.run_reader_archive_worker(blocking))
self.assertTrue(await asyncio.to_thread(started.wait, 1))
caller.cancel()
with self.assertRaises(asyncio.CancelledError):
await caller
shutdown = asyncio.create_task(app.shutdown_reader_archive_state(app.app))
try:
await asyncio.sleep(0.02)
self.assertFalse(shutdown.done())
finally:
release.set()
await asyncio.wait_for(shutdown, 1)
self.assertEqual(app.app.state.reader_archive_workers, set())
async def test_content_isolation_for_active_document_mime_types(self):
for mime in ("text/html", "application/xhtml+xml", "image/svg+xml"):
with self.subTest(mime=mime):
response = await self.content(FakeResponse(headers={"Content-Type": mime}, chunks=(b"active content",)))
self.assert_isolated(response)
self.assertEqual(b"".join([part async for part in response.body_iterator]), b"active content")
async def test_compressed_upstream_is_rejected_before_body_and_on_head(self):
for method in ("GET", "HEAD"):
for status in (200, 206):
with self.subTest(method=method, status=status):
upstream = FakeResponse(status, {"Content-Encoding": "gzip", "Content-Length": "35"}, chunks=(gzip.compress(b"reader text\n" * 20),))
with patch.object(upstream, "iter_chunked", side_effect=AssertionError("encoded body must not be read")):
response = await self.content(upstream, method)
self.assertEqual(response.status_code, 502)
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
self.assertIs(app.app.state.http_session.urls[0][1]["auto_decompress"], False)
async def test_stream_cleanup_before_start_and_after_first_body(self):
for fail_at in ("http.response.start", "http.response.body"):
with self.subTest(fail_at=fail_at):
upstream = FakeResponse(chunks=(b"one", b"two"))
with patch.object(upstream, "release", wraps=upstream.release) as released:
response = await self.content(upstream)
async def send(message):
if message["type"] == fail_at:
raise OSError("client disconnected")
async def receive():
return {"type": "http.disconnect"}
with self.assertRaises(Exception):
await response({"type": "http", "asgi": {"spec_version": "2.4"}}, receive, send)
released.assert_called_once()
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_stream_cleanup_for_legacy_asgi_disconnect_before_iteration(self):
upstream = FakeResponse()
response = await self.content(upstream)
started = asyncio.Event()
async def send(message):
started.set()
await asyncio.Event().wait()
async def receive():
await started.wait()
return {"type": "http.disconnect"}
await asyncio.wait_for(response({"type": "http", "asgi": {"spec_version": "2.0"}}, receive, send), 1)
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_stream_cleanup_after_normal_asgi_completion_is_exactly_once(self):
upstream = FakeResponse(chunks=(b"one", b"two"))
messages = []
with patch.object(upstream, "release", wraps=upstream.release) as released:
response = await self.content(upstream)
async def send(message):
messages.append(message)
await response({"type": "http", "asgi": {"spec_version": "2.4"}}, AsyncMock(), send)
released.assert_called_once()
self.assertEqual(b"".join(message.get("body", b"") for message in messages), b"onetwo")
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_malformed_headers_and_not_modified_release_upstream(self):
cases = (
(200, {"Content-Length": "invalid"}, 502),
(206, {"Content-Range": "bytes " + "9" * 5000 + "-1/2"}, 502),
(206, {"Content-Range": "bytes 0-1/2", "Content-Length": "1"}, 502),
(304, {"ETag": "unchanged"}, 304),
)
for status, headers, expected in cases:
with self.subTest(status=status, expected=expected):
upstream = FakeResponse(status, headers)
response = await self.content(upstream)
self.assertEqual(response.status_code, expected)
self.assertTrue(upstream.released)
self.assertEqual(app.app.state.upstream_semaphore._value, 1)
async def test_bucket_get_head_direct_and_by_source_id(self):
self.assets["f"][app.build_reader_asset_key(self.record)] = {"s": 2, "m": "p", "p": BUCKET_PATH, "b": "vomebook/pdf-pages-v2"}
body = b'{"kind":"pdf-pages"}'
with patch.object(app, "read_reader_bucket_resource", return_value=body) as read:
async with httpx.AsyncClient(transport=httpx.ASGITransport(app=app.app), base_url="http://test") as client:
cases = (("/api/reader-bucket-resource", {"path": BUCKET_PATH}), ("/api/reader-content", {"id": "0ym7e7wbcxau0"}))
for route, params in cases:
for method in ("GET", "HEAD"):
with self.subTest(route=route, params=params, method=method):
response = await client.request(method, route, params=params)
self.assertEqual(response.status_code, 200, response.text)
self.assertEqual(response.content, body if method == "GET" else b"")
self.assertEqual(response.headers["content-length"], str(len(body)))
self.assertIn("application/json", response.headers["content-type"])
self.assert_isolated(response)
self.assertEqual(read.call_count, 1)
async def test_cached_manifest_compression_keeps_exact_payload_and_head_length(self):
body = json.dumps({"title": "完整目录", "chapters": ["章节" + str(i) for i in range(300)]},
ensure_ascii=False).encode()
with patch.object(app, "read_reader_bucket_resource", return_value=body) as read:
async with httpx.AsyncClient(transport=httpx.ASGITransport(app=app.app), base_url="http://test") as client:
for encoding in ("gzip", "identity"):
response = await client.get("/api/reader-bucket-resource", params={"path": BUCKET_PATH},
headers={"Accept-Encoding": encoding})
self.assertEqual(response.content, body)
if encoding == "gzip":
self.assertEqual(response.headers["content-encoding"], "gzip")
self.assertLess(int(response.headers["content-length"]), len(body))
else:
self.assertNotIn("content-encoding", response.headers)
response = await client.head("/api/reader-bucket-resource", params={"path": BUCKET_PATH},
headers={"Accept-Encoding": "gzip"})
self.assertEqual(response.content, b"")
self.assertEqual(int(response.headers["content-length"]), len(body))
self.assertNotEqual(response.headers.get("content-encoding"), "gzip")
self.assertEqual(read.call_count, 1)
async def test_missing_v2_bucket_resource_does_not_request_retired_dataset(self):
class Stream:
def __init__(self, payload):
self.payload = payload
def __enter__(self):
return self
def __exit__(self, *_args):
return False
def read(self, _limit):
return self.payload
class FileSystem:
def __init__(self):
self.uris = []
def open(self, uri, _mode):
self.uris.append(uri)
if uri.startswith("hf://buckets/"):
raise FileNotFoundError(uri)
return Stream(b"dataset fallback")
filesystem = FileSystem()
with patch.object(app, "HfFileSystem", return_value=filesystem):
with self.assertRaises(FileNotFoundError):
app.read_reader_bucket_resource("documents/web/native/" + "a" * 64 + "/document.html")
self.assertEqual(filesystem.uris, ["hf://buckets/vomebook/reader-assets-v2/documents/web/native/" + "a" * 64 + "/document.html"])
async def test_bucket_chapter_sidecar_resolves_to_shared_bucket_endpoint(self):
chapter_path = "chapters/ebook/epub/" + "c" * 64 + "/" + "b" * 16 + "/chapter-manifest.json"
self.assets["f"][app.build_reader_asset_key(self.record)] = {
"s": 2, "m": "e", "p": chapter_path, "b": "vomebook/reader-assets-v2",
"c": chapter_path, "cb": "vomebook/reader-assets-v2",
}
resolved = json.loads((await app.api_reader_resolve("0ym7e7wbcxau0")).body)
self.assertEqual(resolved["extension"], "epub-chapters")
self.assertIn("/api/reader-bucket-resource?path=chapters/ebook/", resolved["chapter_manifest"])
self.assertEqual(resolved["chapter_manifest"].count("pdf-pages"), 0)
bucket_body = b'{"kind":"epub-chapters"}'
bucket_url = "/api/reader-bucket-resource?path=" + quote(chapter_path, safe="")
with patch.object(app, "read_reader_bucket_resource", return_value=bucket_body) as read:
response = await app.api_reader_content(
SimpleNamespace(method="GET", headers={}), bucket_url, None)
self.assertEqual(response.status_code, 200)
self.assertEqual(response.body, bucket_body)
self.assertEqual(read.call_args.args[0], chapter_path)
async def test_cold_resolve_waits_for_conversion_and_empty_failure_is_not_cacheable(self):
with patch.object(app, "reader_assets_cache", {"v": 1, "f": {}}):
await app.api_reader_resolve("0ym7e7wbcxau0")
app.get_reader_assets.assert_awaited_with(wait=True)
self.assets['f'][app.build_reader_asset_key(self.record)] = {
's':2, 'm':'p', 'p':'documents/pdf/ps/'+'a'*64+'/document.pdf',
'b':'vomebook/reader-assets-v2'}
with patch.object(app, "reader_assets_cache", self.assets):
response = await app.api_reader_resolve("0ym7e7wbcxau0")
app.get_reader_assets.assert_awaited_with(wait=False)
self.assertEqual(json.loads(response.body)['extension'], 'pdf')
self.assets['f'].clear()
response = await app.api_reader_resolve("0ym7e7wbcxau0")
self.assertEqual(response.headers['cache-control'], 'no-store')
async def test_bucket_content_rejects_arbitrary_relative_and_ambiguous_paths(self):
for url in (
"https://[broken", "/api/other?path=x", "//evil.test/api/reader-bucket-resource?path=x",
"/api/reader-bucket-resource?path=x&path=y",
"/api/reader-bucket-resource?bucket=vomebook%2Fpdf-optimized&path=x",
"/api/reader-bucket-resource?path=../escape",
"https://huggingface.co/buckets/vomebook/pdf-optimized/resolve/objects/aa/document.pdf",
):
response = await app.api_reader_content(SimpleNamespace(method="GET", headers={}), url, None)
self.assertIn(response.status_code, (400, 403))
async def test_source_id_does_not_hash_originals_and_conversion_precedence_agrees(self):
asset = {"s": 2, "m": "p", "p": OBJECT + "/page-manifest.json", "c": OBJECT + "/chapters-v1/chapter-manifest.json", "f": OBJECT + "/pdf-v1/document.pdf"}
self.assets["f"][app.build_reader_asset_key(self.record)] = asset
resolved = json.loads((await app.api_reader_resolve("0ym7e7wbcxau0")).body)
enriched = app.add_reader_assets({"results": [self.record]}, self.assets)["results"][0]
self.assertEqual(resolved["extension"], "pdf-pages")
self.assertEqual(resolved["extension"], enriched["ReaderExtension"])
self.assertEqual(resolved["chapter_manifest"], enriched["ReaderChapterManifest"])
self.assertEqual(resolved["fallback"], enriched["ReaderFallback"])
if __name__ == "__main__":
unittest.main()
|