tests/test_reader_pdf_loading.py

"""Actual module imports and PDF.js with cold cache and transport faults."""
import contextlib
import json
import re
import threading
import time
import unittest
import urllib.parse
from unittest.mock import patch

from tests.test_reader_performance import (
    PlaywrightError, StaticHandler, minimal_pdf, static_server, sync_playwright,
)

MODULE = '**/static/vendor/pdf.min.*.mjs*'
SOURCE = 'https://huggingface.co/datasets/VoiceOfML/Test/resolve/main/cold-retry.pdf'


@unittest.skipIf(sync_playwright is None, 'Playwright is unavailable')
class PdfLoadingTests(unittest.TestCase):
    @classmethod
    def setUpClass(cls):
        resources = contextlib.ExitStack()
        cls.addClassCleanup(resources.close)
        cls.origin = resources.enter_context(static_server())
        playwright = resources.enter_context(sync_playwright())
        try:
            cls.browser = playwright.chromium.launch(headless=True, args=['--no-sandbox'])
        except PlaywrightError as error:
            raise unittest.SkipTest(str(error))
        cls.addClassCleanup(cls.browser.close)

    def setUp(self):
        self.context = self.browser.new_context(service_workers='block')
        self.addCleanup(self.context.close)
        self.page = self.context.new_page()
        self.page.set_default_timeout(10000)
        self.errors, self.modules, self.documents = [], [], []
        self.page.on('pageerror', lambda error: self.errors.append(str(error)))
        self.page.on('request', lambda request: self.modules.append(request.url)
                     if '/vendor/pdf.min.' in request.url else None)
        self.page.route('**/api/reader-content**', self.document)
        cdp = self.context.new_cdp_session(self.page)
        cdp.send('Network.enable')
        cdp.send('Network.setCacheDisabled', {'cacheDisabled': True})

    def document(self, route):
        self.documents.append(route.request.url)
        route.fulfill(content_type='application/pdf', body=minimal_pdf())

    def open(self):
        self.page.goto(self.origin + '/static/reader.html?' + urllib.parse.urlencode(
            dict(url=SOURCE, ext='pdf', title='Cold retry')), wait_until='domcontentloaded')

    def ready(self):
        self.page.locator('.reader-page canvas.ready').wait_for(state='attached')
        self.assertEqual(self.page.locator('#status').text_content(), '1 \u9875')
        self.assertEqual(self.errors, [])

    def test_cold_success_uses_one_module_request(self):
        self.open()
        self.ready()
        self.assertEqual(len(self.modules), 1)
        self.assertNotIn('reader-module-retry', self.modules[0])

    def test_pdf_engine_preloads_before_id_resolution(self):
        pending = []
        self.page.route('**/api/reader-resolve?**', lambda route: pending.append(route))
        with self.page.expect_worker() as opened:
            self.page.goto(self.origin + '/static/reader.html?id=398vk0yyy8m29&ext=pdf', wait_until='domcontentloaded')
        worker = opened.value
        worker.evaluate('''() => new Promise((resolve, reject) => {
          const deadline = setTimeout(() => reject(Error('Worker did not initialize')), 5000);
          const check = () => globalThis.pdfjsWorker ? (clearTimeout(deadline), resolve(true)) : setTimeout(check, 10);
          check();
        })''')
        self.assertEqual(len(pending), 1)
        self.assertEqual(self.documents, [])
        pending[0].fulfill(content_type='application/json', body=json.dumps({
            'url': SOURCE, 'download': SOURCE, 'extension': 'pdf',
        }))
        self.ready()
        self.assertEqual(len(self.page.workers), 1)
        self.assertEqual(len(pending), 1)

    def test_early_resolve_runs_before_main_module_and_retries_transient_http_once(self):
        held, lookups = [], []
        self.page.route('**/static/reader.js*', lambda route: held.append(route))
        def lookup(route):
            lookups.append(route)
            if len(lookups) == 1:
                route.fulfill(status=503, body='try again')
            else:
                route.fulfill(content_type='application/json', body=json.dumps({
                    'url': SOURCE, 'download': SOURCE, 'extension': 'pdf',
                }))
        self.page.route('**/api/reader-resolve?**', lookup)
        self.page.goto(self.origin + '/static/reader.html?id=398vk0yyy8m29&ext=pdf', wait_until='commit')
        self.page.wait_for_function('() => !!window.__VOICE_READER_RESOLVE__')
        self.assertEqual(self.page.evaluate('() => window.__VOICE_READER_RESOLVE__.promise.then(x => x.status)'), 503)
        self.assertEqual(self.documents, [])
        self.page.unroute('**/static/reader.js*')
        for route in held: route.continue_()
        self.ready()
        self.assertEqual(len(lookups), 2)

    def test_disposal_terminates_prepared_worker_and_pending_resolve(self):
        self.page.add_init_script('''window.__lookupAborted = false;
          const nativeFetch = window.fetch;
          window.fetch = (url, options) => String(url).includes('/api/reader-resolve?')
            ? new Promise((resolve, reject) => options.signal.addEventListener('abort', () => {
                window.__lookupAborted = true; reject(new DOMException('closed', 'AbortError'));
              }, {once:true})) : nativeFetch(url, options);''')
        with self.page.expect_worker() as opened:
            self.page.goto(self.origin + '/static/reader.html?id=398vk0yyy8m29&ext=pdf', wait_until='domcontentloaded')
        with opened.value.expect_event('close'):
            self.page.evaluate("() => window.dispatchEvent(new Event('pagehide'))")
        self.assertTrue(self.page.evaluate('() => window.__lookupAborted'))
        self.assertEqual(self.documents, [])
        self.assertEqual(self.errors, [])

    def test_cached_resolution_skips_early_network_lookup(self):
        self.page.add_init_script('sessionStorage.setItem("reader-resolve:398vk0yyy8m29", ' +
            json.dumps(json.dumps({'url': SOURCE, 'download': SOURCE, 'extension': 'pdf'})) + ');')
        lookups = []
        self.page.route('**/api/reader-resolve?**', lambda route: (lookups.append(route.request.url), route.abort()))
        self.page.goto(self.origin + '/static/reader.html?id=398vk0yyy8m29&ext=pdf', wait_until='domcontentloaded')
        self.ready()
        self.assertEqual(lookups, [])

    def test_worker_timeout_terminates_once_and_exposes_retry(self):
        self.page.add_init_script('''const native = window.setTimeout.bind(window);
          window.setTimeout = (fn, ms, ...args) => native(fn, ms === 20000 ? 250 : ms, ...args);''')
        self.page.route(MODULE, lambda route: route.fulfill(content_type='text/javascript', body='''
          export const GlobalWorkerOptions = {};
          export class PDFWorker {
            promise = new Promise(() => {});
            destroy() { window.__workerDestroys = (window.__workerDestroys || 0) + 1; }
          }
          export function getDocument() { throw Error('must not fetch before Worker readiness'); }
        '''))
        self.open()
        self.page.locator('#reader-engine-retry').wait_for()
        self.assertEqual(self.page.evaluate('() => window.__workerDestroys'), 1)
        self.assertEqual(self.documents, [])
        self.assertEqual(self.errors, [])

    def test_proxy_failure_reuses_prepared_worker_for_original_source(self):
        self.page.unroute('**/api/reader-content**')
        self.page.route('**/api/reader-content**', lambda route: route.fulfill(status=403, body='forbidden'))
        self.page.route(SOURCE, self.document)
        self.open()
        self.ready()
        self.assertEqual(self.documents, [SOURCE])
        self.assertEqual(len(self.page.workers), 1)

    def test_resolved_non_pdf_releases_speculative_worker(self):
        pending = []
        self.page.route('**/api/reader-resolve?**', lambda route: pending.append(route))
        self.page.unroute('**/api/reader-content**')
        self.page.route('**/api/reader-content**', lambda route: route.fulfill(content_type='text/plain', body='Resolved text'))
        with self.page.expect_worker() as opened:
            self.page.goto(self.origin + '/static/reader.html?id=398vk0yyy8m29&ext=pdf', wait_until='domcontentloaded')
        with opened.value.expect_event('close'):
            pending[0].fulfill(content_type='application/json', body=json.dumps({
                'url': SOURCE.replace('.pdf', '.txt'), 'extension': 'txt',
            }))
        self.page.locator('#content').get_by_text('Resolved text').wait_for()
        self.assertEqual(self.errors, [])

    def test_large_original_pdf_renders_using_bounded_ranges_and_stops_transfer(self):
        # A valid large stream object before the xref, not padding after EOF.
        small = minimal_pdf()
        xref = small.index(b'xref\n')
        attachment = b'x' * (4 * 1024 * 1024)
        extra = b'6 0 obj\n<< /Length ' + str(len(attachment)).encode() + b' >>\nstream\n' + attachment + b'\nendstream\nendobj\n'
        suffix = small[xref:].replace(b'0 6\n', b'0 7\n').replace(b'trailer\n', f'{xref:010d} 00000 n \ntrailer\n'.encode()).replace(b'/Size 6', b'/Size 7')
        suffix = re.sub(rb'startxref\n\d+', b'startxref\n' + str(xref + len(extra)).encode(), suffix)
        payload = small[:xref] + extra + suffix
        requests, stopped = [], threading.Event()
        original = StaticHandler.do_GET
        def serve(handler):
            if not handler.path.startswith('/range-fixture.pdf'):
                return original(handler)
            match = re.fullmatch(r'bytes=(\d+)-(\d*)', handler.headers.get('Range', ''))
            start, end = (int(match[1]), min(int(match[2] or len(payload)-1), len(payload)-1)) if match else (0, len(payload)-1)
            request = {'range': (start, end) if match else None, 'sent': 0}
            requests.append(request)
            handler.send_response(206 if match else 200)
            handler.send_header('Content-Type', 'application/pdf')
            handler.send_header('Content-Length', str(end-start+1))
            handler.send_header('Accept-Ranges', 'bytes')
            if match:
                handler.send_header('Content-Range', f'bytes {start}-{end}/{len(payload)}')
            handler.end_headers()
            try:
                for offset in range(start, end+1, 16384):
                    chunk = payload[offset:min(offset+16384, end+1)]
                    handler.wfile.write(chunk)
                    handler.wfile.flush()
                    request['sent'] += len(chunk)
                    time.sleep(.005)
            except (BrokenPipeError, ConnectionResetError):
                pass
            finally:
                if not match: stopped.set()
        self.page.unroute('**/api/reader-content**')
        self.page.add_init_script('''const originalFetch = window.fetch;
          window.fetch = (url, init) => originalFetch(String(url).includes('/api/reader-content?') ? '/range-fixture.pdf' : url, init);''')
        with patch.object(StaticHandler, 'do_GET', serve):
            self.open()
            self.ready()
            self.assertTrue(stopped.wait(2), 'full-file transfer must be cancelled')
            ranges = [item['range'] for item in requests if item['range']]
            self.assertIn((0, 262143), ranges)
            self.assertTrue(any(end == len(payload)-1 for _, end in ranges))
            self.assertLess(sum(item['sent'] for item in requests), 1024 * 1024)
            self.assertEqual(len(self.modules), 1)

    def test_published_range_artifact_selects_small_chunk(self):
        source = 'https://huggingface.co/datasets/vomebook/Reader-Assets/resolve/main/' \
            'objects/aa/' + 'a' * 64 + '/pdf-range-v1-0123456789abcdef-objects/document.pdf'
        self.page.route(MODULE, lambda route: route.fulfill(content_type='text/javascript', body='''
          export const GlobalWorkerOptions = {};
          export class PDFWorker { promise = Promise.resolve(); destroy() {} }
          export function getDocument(options) {
            window.__pdfRangeChunk = options.rangeChunkSize;
            const page = {
              getViewport: ({scale}) => ({width: 600 * scale, height: 800 * scale,
                convertToViewportPoint: (x, y) => [x, y]}),
              getTextContent: async () => ({items: []}),
              render: () => ({promise: Promise.resolve()}),
            };
            return {promise: Promise.resolve({numPages: 1,
              getPage: async () => page, getOutline: async () => null}),
              destroy: () => Promise.resolve()};
          }
        '''))
        self.page.goto(self.origin + '/static/reader.html?' + urllib.parse.urlencode(
            dict(url=source, ext='pdf', title='Optimized PDF')), wait_until='domcontentloaded')
        self.ready()
        self.assertEqual(self.page.evaluate('() => window.__pdfRangeChunk'), 262144)

    def test_first_connection_failure_recovers_with_real_pdf_engine(self):
        def serve(route):
            if 'reader-module-retry' not in route.request.url:
                route.abort('connectionreset')
            else:
                route.continue_()
        self.page.route(MODULE, serve)
        self.open()
        self.ready()
        self.assertEqual(len(self.modules), 2)
        self.assertIn('reader-module-retry=1', self.modules[1])
        self.assertEqual(len(self.documents), 1)

    def test_http_failure_recovers_on_last_attempt(self):
        self.page.route(MODULE, lambda route: route.continue_()
                        if 'reader-module-retry=2' in route.request.url
                        else route.fulfill(status=503, body='temporarily unavailable'))
        self.open()
        self.ready()
        self.assertEqual(len(self.modules), 3)

    def test_exhaustion_has_bounded_requests_and_manual_reload_recovers(self):
        self.page.route(MODULE, lambda route: route.abort('connectionreset'))
        self.open()
        self.page.locator('#reader-engine-retry').wait_for()
        self.assertEqual(len(self.modules), 3)
        self.assertEqual(self.documents, [])
        self.assertEqual(self.page.locator('#content').get_attribute('data-error-code'), 'READER_ENGINE_NETWORK')
        self.assertNotIn('Failed to fetch', self.page.locator('.reader-error').inner_text())
        self.page.unroute(MODULE)
        self.page.locator('#reader-engine-retry').click()
        self.ready()
        self.assertEqual(len(self.documents), 1)

    def test_disposal_cancels_retry_backoff(self):
        self.page.route(MODULE, lambda route: route.abort('connectionreset'))
        self.open()
        self.page.wait_for_timeout(100)
        self.page.evaluate("() => window.dispatchEvent(new Event('pagehide'))")
        self.page.wait_for_timeout(1800)
        self.assertEqual(len(self.modules), 1)
        self.assertEqual(self.documents, [])
        self.assertEqual(self.errors, [])

    def test_timed_out_import_cannot_start_document_after_disposal(self):
        held = []
        self.page.add_init_script('''const native = window.setTimeout.bind(window);
          window.setTimeout = (fn, ms, ...args) => native(fn, ms === 20000 ? 50 : ms, ...args);''')
        self.page.route(MODULE, lambda route: held.append(route))
        self.open()
        self.page.locator('#reader-engine-retry').wait_for()
        self.assertEqual(len(self.modules), 3)
        for route in held:
            route.fulfill(content_type='text/javascript', body='''
              export const GlobalWorkerOptions = {};
              export function getDocument() { window.__latePdfStarted=true; throw Error('late'); }
            ''')
        self.page.wait_for_timeout(200)
        self.assertFalse(self.page.evaluate('() => !!window.__latePdfStarted'))
        self.assertEqual(self.documents, [])
        self.assertEqual(self.errors, [])

    def test_syntax_error_is_not_retried_as_network_failure(self):
        self.page.route(MODULE, lambda route: route.fulfill(content_type='text/javascript', body='export const = ;'))
        self.open()
        self.page.locator('.reader-error').wait_for()
        self.assertEqual(len(self.modules), 1)
        self.assertEqual(self.page.locator('#reader-engine-retry').count(), 0)

    def test_first_http_failure_recovers_under_real_service_worker(self):
        original = StaticHandler.do_GET
        requests = []
        def serve(handler):
            if '/vendor/pdf.min.' in handler.path:
                requests.append(handler.path)
                if len(requests) == 1:
                    handler.send_error(503)
                    return
            if handler.path == '/manifest.json' or handler.path.startswith('/icons/'):
                handler.path = '/static' + handler.path
            original(handler)
        with patch.object(StaticHandler, 'do_GET', serve), \
                self.browser.new_context(service_workers='allow') as context:
            page = context.new_page()
            context.route('**/api/reader-content**', lambda route: route.fulfill(
                content_type='application/pdf', body=minimal_pdf()))
            page.goto(self.origin + '/static/manifest.json')
            page.evaluate('''async () => {
              await navigator.serviceWorker.register('/static/sw.js',{scope:'/static/'});
              await Promise.race([navigator.serviceWorker.ready,
                new Promise((_, reject) => setTimeout(() => reject(Error('SW activation timed out')), 10000))]);
            }''')
            page.goto(self.origin + '/static/reader.html?' + urllib.parse.urlencode(
                dict(url=SOURCE, ext='pdf', title='Cold SW retry')))
            page.locator('.reader-page canvas.ready').wait_for(state='attached', timeout=15000)
            self.assertTrue(page.evaluate('() => !!navigator.serviceWorker.controller'))
            self.assertEqual(len(requests), 2)
            self.assertIn('reader-module-retry=1', requests[1])