crawlee 1.0.4b3__py3-none-any.whl → 1.0.4b4__py3-none-any.whl

crawlee/_utils/urls.py

@@ -7,6 +7,7 @@ from yarl import URL
 
 if TYPE_CHECKING:
     from collections.abc import Iterator
+    from logging import Logger
 
 
 def is_url_absolute(url: str) -> bool:
@@ -22,13 +23,19 @@ def convert_to_absolute_url(base_url: str, relative_url: str) -> str:
     return str(URL(base_url).join(URL(relative_url)))
 
 
-def to_absolute_url_iterator(base_url: str, urls: Iterator[str]) -> Iterator[str]:
+def to_absolute_url_iterator(base_url: str, urls: Iterator[str], logger: Logger | None = None) -> Iterator[str]:
     """Convert an iterator of relative URLs to absolute URLs using a base URL."""
     for url in urls:
         if is_url_absolute(url):
             yield url
         else:
-            yield convert_to_absolute_url(base_url, url)
+            converted_url = convert_to_absolute_url(base_url, url)
+            # Skip the URL if conversion fails, probably due to an incorrect format, such as 'mailto:'.
+            if not is_url_absolute(converted_url):
+                if logger:
+                    logger.debug(f'Could not convert URL "{url}" to absolute using base URL "{base_url}". Skipping it.')
+                continue
+            yield converted_url
 
 
 _http_url_adapter = TypeAdapter(AnyHttpUrl)

crawlee/crawlers/_abstract_http/_abstract_http_crawler.py

@@ -167,7 +167,9 @@ class AbstractHttpCrawler(
             kwargs.setdefault('strategy', 'same-hostname')
 
             links_iterator: Iterator[str] = iter(self._parser.find_links(parsed_content, selector=selector))
-            links_iterator = to_absolute_url_iterator(context.request.loaded_url or context.request.url, links_iterator)
+            links_iterator = to_absolute_url_iterator(
+                context.request.loaded_url or context.request.url, links_iterator, logger=context.log
+            )
 
             if robots_txt_file:
                 skipped, links_iterator = partition(lambda url: robots_txt_file.is_allowed(url), links_iterator)

crawlee/crawlers/_playwright/_playwright_crawler.py

@@ -366,7 +366,9 @@ class PlaywrightCrawler(BasicCrawler[PlaywrightCrawlingContext, StatisticsState]
             links_iterator: Iterator[str] = iter(
                 [url for element in elements if (url := await element.get_attribute('href')) is not None]
             )
-            links_iterator = to_absolute_url_iterator(context.request.loaded_url or context.request.url, links_iterator)
+            links_iterator = to_absolute_url_iterator(
+                context.request.loaded_url or context.request.url, links_iterator, logger=context.log
+            )
 
             if robots_txt_file:
                 skipped, links_iterator = partition(lambda url: robots_txt_file.is_allowed(url), links_iterator)

{crawlee-1.0.4b3.dist-info → crawlee-1.0.4b4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: crawlee
-Version: 1.0.4b3
+Version: 1.0.4b4
 Summary: Crawlee for Python
 Project-URL: Apify Homepage, https://apify.com
 Project-URL: Changelog, https://crawlee.dev/python/docs/changelog

{crawlee-1.0.4b3.dist-info → crawlee-1.0.4b4.dist-info}/RECORD

@@ -37,7 +37,7 @@ crawlee/_utils/sitemap.py,sha256=UI9EJiFiyFvV5_flVUtdsEVz8ZsJeRERPtcx8ZsqjTU,166
 crawlee/_utils/system.py,sha256=tA8AP__9vsJ9OTLTnAYAKkxc8U5-IEna0N_hqYBybUo,4294
 crawlee/_utils/time.py,sha256=WK17P939r65dLz2rWvL59OEJoxgzdinw-ND9WuG4DuU,2353
 crawlee/_utils/try_import.py,sha256=QI_58ifc2l0Rxehzu6xcofQrRAVeLzZuBTTTHttLl8s,1310
-crawlee/_utils/urls.py,sha256=NN27TA6KMU5V_j5TCZ4o33UIXw4pB9a-wGlmDQtYT8E,1294
+crawlee/_utils/urls.py,sha256=fEYXJxBT02f-DIYKF_h7PdaKAShfXBs99-dHDjDX03A,1725
 crawlee/_utils/wait.py,sha256=RfiXhp5VUBxOEtEMtru7_jNfKDr2BJCcFge5qGg2gxk,2848
 crawlee/_utils/web.py,sha256=nnKhg8pUSWz0RY64Qd-_GPNBX1fWI2hXS-gzcfQ-rig,364
 crawlee/browsers/__init__.py,sha256=TghkrNSbI_k87UgVBlgNNcEm8Ot05pSLEAPRSv6YsUs,1064
@@ -53,7 +53,7 @@ crawlee/crawlers/__init__.py,sha256=9VmFahav3rjE-2Bxa5PAhBgkYXP0k5SSAEpdG2xMZ7c,
 crawlee/crawlers/_types.py,sha256=xbGTJQirgz5wUbfr12afMR4q-_5AWP7ngF2e8K5P8l0,355
 crawlee/crawlers/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 crawlee/crawlers/_abstract_http/__init__.py,sha256=QCjn8x7jpo8FwEeSRw10TVj_0La2v9mLEiQWdk2RoTw,273
-crawlee/crawlers/_abstract_http/_abstract_http_crawler.py,sha256=ZG6A4DNGZYbOBXi0Th6K6CHDi2SqWO5VpxcnjypDO-A,11503
+crawlee/crawlers/_abstract_http/_abstract_http_crawler.py,sha256=DEiErZi7j2FHMgyVELPy09GyHo5Gx4UDpuiN6D3sGNk,11553
 crawlee/crawlers/_abstract_http/_abstract_http_parser.py,sha256=Y5o_hiW_0mQAte5GFqkUxscwKEFpWrBYRsLKP1cfBwE,3521
 crawlee/crawlers/_abstract_http/_http_crawling_context.py,sha256=Rno_uJ8ivmyRxFQv2MyY_z9B5WPHSEd5MAPz31_1ZIo,2179
 crawlee/crawlers/_abstract_http/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -85,7 +85,7 @@ crawlee/crawlers/_parsel/_parsel_crawling_context.py,sha256=sZB26RcRLjSoD15myEOM
 crawlee/crawlers/_parsel/_parsel_parser.py,sha256=yWBfuXUHMriK4DRnyrXTQoGeqX5WV9bOEkBp_g0YCvQ,1540
 crawlee/crawlers/_parsel/_utils.py,sha256=MbRwx-cdjlq1zLzFYf64M3spOGQ6yxum4FvP0sdqA_Q,2693
 crawlee/crawlers/_playwright/__init__.py,sha256=6Cahe6VEF82o8CYiP8Cmp58Cmb6Rb8uMeyy7wnwe5ms,837
-crawlee/crawlers/_playwright/_playwright_crawler.py,sha256=YI_EvJApfabuBY5TZq7OdBI-45ASiDE2GfsIC4qpd8A,23756
+crawlee/crawlers/_playwright/_playwright_crawler.py,sha256=QfZVWj6A0H1idC0yQT-WAxlWTk7janB4TtKDtf8htt8,23806
 crawlee/crawlers/_playwright/_playwright_crawling_context.py,sha256=Oi0tMBXHaEDlFjqG01DzgB7Ck52bjVjz-X__eMioxas,1249
 crawlee/crawlers/_playwright/_playwright_http_client.py,sha256=Nfm69dqX85k68jN1p3ljZWbn8egqDWPIPRykXyXsoQs,3977
 crawlee/crawlers/_playwright/_playwright_pre_nav_crawling_context.py,sha256=fEI2laWhmJdWiGoMF5JBLBsim9NtENfagZt6FFd2Rgo,1387
@@ -187,8 +187,8 @@ crawlee/storages/_request_queue.py,sha256=bjBOGbpMaGUsqJPVB-JD2VShziPAYMI-GvWKKp
 crawlee/storages/_storage_instance_manager.py,sha256=72n0YlPwNpSQDJSPf4TxnI2GvIK6L-ZiTmHRbFcoVU0,8164
 crawlee/storages/_utils.py,sha256=Yz-5tEBYKYCFJemYT29--uGJqoJLApLDLgPcsnbifRw,439
 crawlee/storages/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-crawlee-1.0.4b3.dist-info/METADATA,sha256=nMdKXjkypc9_hf4w200FOndiYblo5_ZvwSuHRmYKqpk,29314
-crawlee-1.0.4b3.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-crawlee-1.0.4b3.dist-info/entry_points.txt,sha256=1p65X3dA-cYvzjtlxLL6Kn1wpY-3uEDVqJLp53uNPeo,45
-crawlee-1.0.4b3.dist-info/licenses/LICENSE,sha256=AsFjHssKjj4LGd2ZCqXn6FBzMqcWdjQre1byPPSypVw,11355
-crawlee-1.0.4b3.dist-info/RECORD,,
+crawlee-1.0.4b4.dist-info/METADATA,sha256=G2GqZKPuhTSf7bDXOKQ1b8WdzsMgIUCatpH1dTAu2nM,29314
+crawlee-1.0.4b4.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+crawlee-1.0.4b4.dist-info/entry_points.txt,sha256=1p65X3dA-cYvzjtlxLL6Kn1wpY-3uEDVqJLp53uNPeo,45
+crawlee-1.0.4b4.dist-info/licenses/LICENSE,sha256=AsFjHssKjj4LGd2ZCqXn6FBzMqcWdjQre1byPPSypVw,11355
+crawlee-1.0.4b4.dist-info/RECORD,,