diff --git a/swh/lister/cgit/lister.py b/swh/lister/cgit/lister.py index c0d9113..5ca9445 100644 --- a/swh/lister/cgit/lister.py +++ b/swh/lister/cgit/lister.py @@ -1,217 +1,234 @@ # Copyright (C) 2019-2021 The Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from datetime import datetime, timezone import logging import re from typing import Any, Dict, Iterator, List, Optional from urllib.parse import urljoin, urlparse from bs4 import BeautifulSoup import requests from requests.exceptions import HTTPError from tenacity.before_sleep import before_sleep_log from swh.lister import USER_AGENT from swh.lister.pattern import CredentialsType, StatelessLister from swh.lister.utils import throttling_retry from swh.scheduler.interface import SchedulerInterface from swh.scheduler.model import ListedOrigin logger = logging.getLogger(__name__) Repositories = List[Dict[str, Any]] class CGitLister(StatelessLister[Repositories]): """Lister class for CGit repositories. This lister will retrieve the list of published git repositories by parsing the HTML page(s) of the index retrieved at `url`. The lister currently defines 2 listing behaviors: - If the `base_git_url` is provided, the listed origin urls are computed out of the base git url link and the one listed in the main listed page (resulting in less HTTP queries than the 2nd behavior below). This is expected to be the main deployed behavior. - Otherwise (with no `base_git_url`), for each found git repository listed, one extra HTTP query is made at the given url found in the main listing page to gather published "Clone" URLs to be used as origin URL for that git repo. If several "Clone" urls are provided, prefer the http/https one, if any, otherwise fallback to the first one. """ LISTER_NAME = "cgit" def __init__( self, scheduler: SchedulerInterface, url: str, instance: Optional[str] = None, credentials: Optional[CredentialsType] = None, base_git_url: Optional[str] = None, ): """Lister class for CGit repositories. Args: url: main URL of the CGit instance, i.e. url of the index of published git repositories on this instance. instance: Name of cgit instance. Defaults to url's network location if unset. base_git_url: Optional base git url which allows the origin url computations. """ super().__init__( scheduler=scheduler, url=url, instance=instance, credentials=credentials, ) self.session = requests.Session() self.session.headers.update( {"Accept": "application/html", "User-Agent": USER_AGENT} ) self.base_git_url = base_git_url @throttling_retry(before_sleep=before_sleep_log(logger, logging.DEBUG)) def _get_and_parse(self, url: str) -> BeautifulSoup: """Get the given url and parse the retrieved HTML using BeautifulSoup""" response = self.session.get(url) response.raise_for_status() return BeautifulSoup(response.text, features="html.parser") def get_pages(self) -> Iterator[Repositories]: """Generate git 'project' URLs found on the current CGit server The last_update date is retrieved on the list of repo page to avoid to compute it on the repository details which only give a date per branch """ next_page: Optional[str] = self.url while next_page: bs_idx = self._get_and_parse(next_page) page_results = [] for tr in bs_idx.find("div", {"class": "content"}).find_all( "tr", {"class": ""} ): repository_link = tr.find("a")["href"] repo_url = None git_url = None base_url = urljoin(self.url, repository_link).strip("/") if self.base_git_url: # mapping provided # computing git url git_url = base_url.replace(self.url, self.base_git_url) else: # we compute the git detailed page url from which we will retrieve # the git url (cf. self.get_origins_from_page) repo_url = base_url span = tr.find("span", {"class": re.compile("age-")}) last_updated_date = span.get("title") if span else None page_results.append( { "url": repo_url, "git_url": git_url, "last_updated_date": last_updated_date, } ) yield page_results try: pager = bs_idx.find("ul", {"class": "pager"}) current_page = pager.find("a", {"class": "current"}) if current_page: next_page = current_page.parent.next_sibling.a["href"] next_page = urljoin(self.url, next_page) except (AttributeError, KeyError): # no pager, or no next page next_page = None def get_origins_from_page( self, repositories: Repositories ) -> Iterator[ListedOrigin]: """Convert a page of cgit repositories into a list of ListedOrigins.""" assert self.lister_obj.id is not None for repo in repositories: origin_url = repo["git_url"] or self._get_origin_from_repository_url( repo["url"] ) if origin_url is None: continue yield ListedOrigin( lister_id=self.lister_obj.id, url=origin_url, visit_type="git", last_update=_parse_last_updated_date(repo), ) def _get_origin_from_repository_url(self, repository_url: str) -> Optional[str]: """Extract the git url from the repository page""" try: bs = self._get_and_parse(repository_url) except HTTPError as e: logger.warning( "Unexpected HTTP status code %s on %s", e.response.status_code, e.response.url, ) return None + # check if we are on the summary tab, if not, go to this tab + tab = bs.find("table", {"class": "tabs"}) + if tab: + summary_a = tab.find("a", string="summary") + if summary_a: + summary_url = urljoin(repository_url, summary_a["href"]).strip("/") + + if summary_url != repository_url: + logger.debug( + "%s : Active tab is not the summary, trying to load the summary page", + repository_url, + ) + return self._get_origin_from_repository_url(summary_url) + else: + logger.debug("No summary tab found on %s", repository_url) + # origin urls are listed on the repository page # TODO check if forcing https is better or not ? # # # urls = [x["href"] for x in bs.find_all("a", {"rel": "vcs-git"})] if not urls: + logger.debug("No git urls found on %s", repository_url) return None # look for the http/https url, if any, and use it as origin_url for url in urls: if urlparse(url).scheme in ("http", "https"): origin_url = url break else: # otherwise, choose the first one origin_url = urls[0] return origin_url def _parse_last_updated_date(repository: Dict[str, Any]) -> Optional[datetime]: """Parse the last updated date""" date = repository.get("last_updated_date") if not date: return None parsed_date = None for date_format in ("%Y-%m-%d %H:%M:%S %z", "%Y-%m-%d %H:%M:%S (%Z)"): try: parsed_date = datetime.strptime(date, date_format) # force UTC to avoid naive datetime if not parsed_date.tzinfo: parsed_date = parsed_date.replace(tzinfo=timezone.utc) break except Exception: pass if not parsed_date: logger.warning( "Could not parse %s last_updated date: %s", repository["url"], date, ) return parsed_date diff --git a/swh/lister/cgit/tests/data/https_git.acdw.net/Readme.md b/swh/lister/cgit/tests/data/https_git.acdw.net/Readme.md new file mode 100644 index 0000000..0b02a73 --- /dev/null +++ b/swh/lister/cgit/tests/data/https_git.acdw.net/Readme.md @@ -0,0 +1 @@ +These files are a partial dump of http://git.savannah.gnu.org/cgit diff --git a/swh/lister/cgit/tests/data/https_git.acdw.net/cgit b/swh/lister/cgit/tests/data/https_git.acdw.net/cgit new file mode 100644 index 0000000..a3da859 --- /dev/null +++ b/swh/lister/cgit/tests/data/https_git.acdw.net/cgit @@ -0,0 +1,40 @@ + + + +friendware by acdw + + + + + +
+ + + + +
+index
+ + +
+
+ + + + + +
NameDescriptionIdleLinks
sfeedMy sfeed scripts +28 min.summary log tree
fooNon existing repository +28 min.summary log tree
+
+© 2022 C. Duckworth. +generated by +cgit. +comments, issues, and patches welcome at +< +git at acdw dot net +>. +
+
+ + diff --git a/swh/lister/cgit/tests/data/https_git.acdw.net/foo b/swh/lister/cgit/tests/data/https_git.acdw.net/foo new file mode 100644 index 0000000..c6560a4 --- /dev/null +++ b/swh/lister/cgit/tests/data/https_git.acdw.net/foo @@ -0,0 +1,33 @@ + + + + + + + + + +
+ + + + +
+index
+ + +
+
No repositories found
+
+
+© 2022 C. Duckworth. +generated by +cgit. +comments, issues, and patches welcome at +< +git at acdw dot net +>. +
+
+ + diff --git a/swh/lister/cgit/tests/data/https_git.acdw.net/foo_summary b/swh/lister/cgit/tests/data/https_git.acdw.net/foo_summary new file mode 100644 index 0000000..c6560a4 --- /dev/null +++ b/swh/lister/cgit/tests/data/https_git.acdw.net/foo_summary @@ -0,0 +1,33 @@ + + + + + + + + + +
+ + + + +
+index
+ + +
+
No repositories found
+
+
+© 2022 C. Duckworth. +generated by +cgit. +comments, issues, and patches welcome at +< +git at acdw dot net +>. +
+
+ + diff --git a/swh/lister/cgit/tests/data/https_git.acdw.net/sfeed b/swh/lister/cgit/tests/data/https_git.acdw.net/sfeed new file mode 100644 index 0000000..d0d01ad --- /dev/null +++ b/swh/lister/cgit/tests/data/https_git.acdw.net/sfeed @@ -0,0 +1,49 @@ + + + +sfeed - My sfeed scripts + + + + + + + + +
+ + + + +
+about summary refs log tree commit diff stats
+ + + +
+
+

sfeed

+

Turns out, sfeed is cool! You can see what this repo generates at https://acdw.casa/planet/.

+
+
+© 2022 C. Duckworth. +generated by +cgit. +comments, issues, and patches welcome at +< +git at acdw dot net +>. +
+
+ + diff --git a/swh/lister/cgit/tests/data/https_git.acdw.net/sfeed_summary b/swh/lister/cgit/tests/data/https_git.acdw.net/sfeed_summary new file mode 100644 index 0000000..b71e1cf --- /dev/null +++ b/swh/lister/cgit/tests/data/https_git.acdw.net/sfeed_summary @@ -0,0 +1,63 @@ + + + +sfeed - My sfeed scripts + + + + + + + + +
+ + + + +
+about summary refs log tree commit diff stats
+ + + +
+
+
+ + + + + + + + + + + + + + + +
BranchCommit messageAuthorAge
mainAdd APODCase Duckworth38 min.
 
 
AgeCommit messageAuthor
38 min.Add APOD HEAD mainCase Duckworth
4 daysChange fresh item colorsCase Duckworth
4 daysIndentationCase Duckworth
5 daysAdd Tab CompletionCase Duckworth
5 daysAdd Lonnie JohnsonCase Duckworth
7 daysAdd miniature calendar; metafilterCase Duckworth
9 daysAdd active listeningCase Duckworth
10 daysAdd tilde.town blogCase Duckworth
12 daysAdd zsergeCase Duckworth
12 daysRemove duplicateCase Duckworth
[...]
 
Clone
https://git.acdw.net/sfeed
+
+© 2022 C. Duckworth. +generated by +cgit. +comments, issues, and patches welcome at +< +git at acdw dot net +>. +
+
+ + diff --git a/swh/lister/cgit/tests/test_lister.py b/swh/lister/cgit/tests/test_lister.py index f996333..9b5c0c3 100644 --- a/swh/lister/cgit/tests/test_lister.py +++ b/swh/lister/cgit/tests/test_lister.py @@ -1,267 +1,280 @@ # Copyright (C) 2019-2021 The Software Heritage developers # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information from datetime import datetime, timedelta, timezone import os from typing import List import pytest from swh.core.pytest_plugin import requests_mock_datadir_factory from swh.lister import __version__ from swh.lister.cgit.lister import CGitLister, _parse_last_updated_date from swh.lister.pattern import ListerStats def test_lister_cgit_get_pages_one_page(requests_mock_datadir, swh_scheduler): url = "https://git.savannah.gnu.org/cgit/" lister_cgit = CGitLister(swh_scheduler, url=url) repos: List[List[str]] = list(lister_cgit.get_pages()) flattened_repos = sum(repos, []) assert len(flattened_repos) == 977 assert flattened_repos[0]["url"] == "https://git.savannah.gnu.org/cgit/elisp-es.git" # note the url below is NOT a subpath of /cgit/ assert ( flattened_repos[-1]["url"] == "https://git.savannah.gnu.org/path/to/yetris.git" ) # noqa # note the url below is NOT on the same server assert flattened_repos[-2]["url"] == "http://example.org/cgit/xstarcastle.git" def test_lister_cgit_get_pages_with_pages(requests_mock_datadir, swh_scheduler): url = "https://git.tizen/cgit/" lister_cgit = CGitLister(swh_scheduler, url=url) repos: List[List[str]] = list(lister_cgit.get_pages()) flattened_repos = sum(repos, []) # we should have 16 repos (listed on 3 pages) assert len(repos) == 3 assert len(flattened_repos) == 16 def test_lister_cgit_run_with_page(requests_mock_datadir, swh_scheduler): """cgit lister supports pagination""" url = "https://git.tizen/cgit/" lister_cgit = CGitLister(swh_scheduler, url=url) stats = lister_cgit.run() expected_nb_origins = 16 assert stats == ListerStats(pages=3, origins=expected_nb_origins) # test page parsing scheduler_origins = swh_scheduler.get_listed_origins( lister_cgit.lister_obj.id ).results assert len(scheduler_origins) == expected_nb_origins # test listed repositories for listed_origin in scheduler_origins: assert listed_origin.visit_type == "git" assert listed_origin.url.startswith("https://git.tizen") # test user agent content assert len(requests_mock_datadir.request_history) != 0 for request in requests_mock_datadir.request_history: assert "User-Agent" in request.headers user_agent = request.headers["User-Agent"] assert "Software Heritage Lister" in user_agent assert __version__ in user_agent def test_lister_cgit_run_populates_last_update(requests_mock_datadir, swh_scheduler): """cgit lister returns last updated date""" url = "https://git.tizen/cgit" urls_without_date = [ f"https://git.tizen.org/cgit/{suffix_url}" for suffix_url in [ "All-Projects", "All-Users", "Lock-Projects", ] ] lister_cgit = CGitLister(swh_scheduler, url=url) stats = lister_cgit.run() expected_nb_origins = 16 assert stats == ListerStats(pages=3, origins=expected_nb_origins) # test page parsing scheduler_origins = swh_scheduler.get_listed_origins( lister_cgit.lister_obj.id ).results assert len(scheduler_origins) == expected_nb_origins # test listed repositories for listed_origin in scheduler_origins: if listed_origin.url in urls_without_date: assert listed_origin.last_update is None else: assert listed_origin.last_update is not None @pytest.mark.parametrize( "date_str,expected_date", [ ({}, None), ("unexpected date", None), ("2020-0140-10 10:10:10 (GMT)", None), ( "2020-01-10 10:10:10 (GMT)", datetime( year=2020, month=1, day=10, hour=10, minute=10, second=10, tzinfo=timezone.utc, ), ), ( "2019-08-04 05:10:41 +0100", datetime( year=2019, month=8, day=4, hour=5, minute=10, second=41, tzinfo=timezone(timedelta(hours=1)), ), ), ], ) def test_lister_cgit_date_parsing(date_str, expected_date): """test cgit lister date parsing""" repository = {"url": "url", "last_updated_date": date_str} assert _parse_last_updated_date(repository) == expected_date requests_mock_datadir_missing_url = requests_mock_datadir_factory( ignore_urls=[ "https://git.tizen/cgit/adaptation/ap_samsung/audio-hal-e4x12", ] ) def test_lister_cgit_get_origin_from_repo_failing( requests_mock_datadir_missing_url, swh_scheduler ): url = "https://git.tizen/cgit/" lister_cgit = CGitLister(swh_scheduler, url=url) stats = lister_cgit.run() expected_nb_origins = 15 assert stats == ListerStats(pages=3, origins=expected_nb_origins) @pytest.mark.parametrize( "credentials, expected_credentials", [ (None, []), ({"key": "value"}, []), ( {"cgit": {"tizen": [{"username": "user", "password": "pass"}]}}, [{"username": "user", "password": "pass"}], ), ], ) def test_lister_cgit_instantiation_with_credentials( credentials, expected_credentials, swh_scheduler ): url = "https://git.tizen/cgit/" lister = CGitLister( swh_scheduler, url=url, instance="tizen", credentials=credentials ) # Credentials are allowed in constructor assert lister.credentials == expected_credentials def test_lister_cgit_from_configfile(swh_scheduler_config, mocker): load_from_envvar = mocker.patch("swh.lister.pattern.load_from_envvar") load_from_envvar.return_value = { "scheduler": {"cls": "local", **swh_scheduler_config}, "url": "https://git.tizen/cgit/", "instance": "tizen", "credentials": {}, } lister = CGitLister.from_configfile() assert lister.scheduler is not None assert lister.credentials is not None @pytest.mark.parametrize( "url,base_git_url,expected_nb_origins", [ ("https://git.eclipse.org/c", "https://eclipse.org/r", 5), ("https://git.baserock.org/cgit/", "https://git.baserock.org/git/", 3), ("https://jff.email/cgit/", "git://jff.email/opt/git/", 6), ], ) def test_lister_cgit_with_base_git_url( url, base_git_url, expected_nb_origins, requests_mock_datadir, swh_scheduler ): """With base git url provided, listed urls should be the computed origin urls""" lister_cgit = CGitLister( swh_scheduler, url=url, base_git_url=base_git_url, ) stats = lister_cgit.run() assert stats == ListerStats(pages=1, origins=expected_nb_origins) # test page parsing scheduler_origins = swh_scheduler.get_listed_origins( lister_cgit.lister_obj.id ).results assert len(scheduler_origins) == expected_nb_origins # test listed repositories for listed_origin in scheduler_origins: assert listed_origin.visit_type == "git" assert listed_origin.url.startswith(base_git_url) assert ( listed_origin.url.startswith(url) is False ), f"url should be mapped to {base_git_url}" def test_lister_cgit_get_pages_with_pages_and_retry( requests_mock_datadir, requests_mock, datadir, mocker, swh_scheduler ): url = "https://git.tizen/cgit/" with open(os.path.join(datadir, "https_git.tizen/cgit,ofs=50"), "rb") as page: requests_mock.get( f"{url}?ofs=50", [ {"content": None, "status_code": 429}, {"content": None, "status_code": 429}, {"content": page.read(), "status_code": 200}, ], ) lister_cgit = CGitLister(swh_scheduler, url=url) mocker.patch.object(lister_cgit._get_and_parse.retry, "sleep") repos: List[List[str]] = list(lister_cgit.get_pages()) flattened_repos = sum(repos, []) # we should have 16 repos (listed on 3 pages) assert len(repos) == 3 assert len(flattened_repos) == 16 + + +def test_lister_cgit_summary_not_default(requests_mock_datadir, swh_scheduler): + """cgit lister returns git url when the default repository tab is not the summary""" + + url = "https://git.acdw.net/cgit" + + lister_cgit = CGitLister(swh_scheduler, url=url) + + stats = lister_cgit.run() + + expected_nb_origins = 1 + assert stats == ListerStats(pages=1, origins=expected_nb_origins)