Merge pull request #103 from sw360/martin/fix-github-tag-matching

GitHub tag matching

capycli/bom/findsources.py

@@ -11,7 +11,9 @@
 import re
 import sys
 import time
-from typing import Any, Dict, List, Tuple
+from collections.abc import Iterable
+from typing import Any, Dict, List, Set, Tuple
+from urllib.parse import parse_qs, urlparse
 
 import requests
 import semver
@@ -36,13 +38,59 @@
 class FindSources(capycli.common.script_base.ScriptBase):
     """Go through the list of SBOM items and try to determine the source code."""
 
+    class TagCache:
+        """A key task performed in this module is fetching tags from GitHub
+           and match tags to (component) versions. This task includes many
+           calls to the GitHub API, which we seek to limit by implementing
+           an internal cache and a logic to guess tags, instead of
+           performing exhaustive searches.
+        """
+        def __init__(self) -> None:
+            self.data: Dict[Tuple[str, str], Set[str]] = {}
+
+        def __getitem__(self, key: Any) -> Set[str]:
+            """Get the set of all cached tags for a key."""
+            return self.data[self._validate_key(key)]
+
+        def _validate_key(self, key: Tuple[str, str]) -> Tuple[str, str]:
+            """Ensure our keys are hashable."""
+            if len(key) != 2 or key != (str(key[0]), str(key[1])):
+                raise KeyError(f'{self.__class__.__name__} key must consist of'
+                               'a project name and a version string')
+            return key
+
+        def add(self, project: str, version: str, tag: str) -> None:
+            """Cache a tag for a specific project and version."""
+            key = self._validate_key((project, version))
+            tags = self.data.setdefault(key, set())
+            tags.add(tag)
+
+        def filter(self, project: str, version: str, data: Any) -> List[str]:
+            """Remove all cached entries from @data."""
+            if isinstance(data, str):
+                data = [data]
+            elif not isinstance(data, Iterable):
+                raise ValueError('Expecting an iterable of tags!')
+            key = self._validate_key((project, version))
+            return [item for item in data
+                    if item not in self.data.get(key, [])
+                    and len(item) > 0]
+
+        def filter_and_cache(self, project: str, version: str, data: Any) -> List[str]:
+            """Convenience method to to filtering and adding in one run."""
+            candidates = set(self.filter(project, version, data))
+            for tag in candidates:
+                self.add(project, version, tag)
+            return list(candidates)
+
     def __init__(self) -> None:
         self.verbose: bool = False
-        self.version_regex = re.compile(r"[\d+\.|_]+[\d+]")
+        self.version_regex = re.compile(r"(\d+[._])+\d+")
         self.github_project_name_regex = re.compile(r"^[a-zA-Z0-9-]+(/[a-zA-Z0-9-]+)*$")
         self.github_name: str = ""
         self.github_token: str = ""
         self.sw360_url: str = os.environ.get("SW360ServerUrl", "")
+        self.tag_cache = self.TagCache()
 
     def is_sourcefile_accessible(self, sourcefile_url: str) -> bool:
         """Check if the URL is accessible."""
@@ -70,34 +118,61 @@ def is_sourcefile_accessible(self, sourcefile_url: str) -> bool:
         return False
 
     @staticmethod
-    def github_request(url: str, username: str = "", token: str = "") -> Any:
+    def github_request(url: str, username: str = "", token: str = "",
+                       return_response: bool = False,
+                       allow_redirects: bool = True,  # default in requests
+                       ) -> Any:
         try:
             headers = {}
             if token:
                 headers["Authorization"] = "token " + token
             if username:
                 headers["Username"] = username
-            response = requests.get(url, headers=headers)
-            if not response.ok:
-                if response.status_code == 429 or \
-                        'rate limit exceeded' in response.reason or \
-                        "API rate limit exceeded" in response.json().get("message"):
-                    print(
-                        Fore.LIGHTYELLOW_EX +
-                        "      Github API rate limit exceeded - wait 60s and retry ... " +
-                        Style.RESET_ALL)
-                    time.sleep(60)
-                    return FindSources.github_request(url, username, token)
-
-            return response.json()
+            response = requests.get(url, headers=headers,
+                                    allow_redirects=allow_redirects)
+            if response.status_code == 429 \
+                    or 'rate limit exceeded' in response.reason \
+                    or 'API rate limit exceeded' in response.json().get('message', ''):
+                print(
+                    Fore.LIGHTYELLOW_EX +
+                    "      Github API rate limit exceeded - wait 60s and retry ... " +
+                    Style.RESET_ALL)
+                time.sleep(60)
+                return FindSources.github_request(url, username, token, return_response=return_response)
+            if response.json().get('message', '').startswith("Bad credentials"):
+                print_red("Invalid GitHub credential provided - aborting!")
+                sys.exit(ResultCode.RESULT_ERROR_ACCESSING_SERVICE)
+
+        except AttributeError as err:
+            # response.json() did not return a dictionary
+            if hasattr(err, 'name'):
+                name = err.name
+            else:  # Python prior to 3.10
+                name = err.args[0].split("'")[3]
+            if not name == 'get':
+                raise
+
+        except requests.exceptions.JSONDecodeError:
+            response._content = b'{}'
+
+        except requests.exceptions.ConnectionError as ex:
+            print(
+                Fore.LIGHTYELLOW_EX +
+                f"      Connection issues accessing {url} " + repr(ex) +
+                "\n      Retrying in 60 seconds!" +
+                Style.RESET_ALL)
+            time.sleep(60)
+            return FindSources.github_request(url, username, token, return_response=return_response)
 
         except Exception as ex:
             print(
                 Fore.LIGHTYELLOW_EX +
                 "      Error accessing GitHub: " + repr(ex) +
                 Style.RESET_ALL)
-
-            return {}
+            response = requests.Response()
+            response._content = \
+                b'{' + f'"exception": "{repr(ex)}"'.encode() + b'}'
+        return response if return_response else response.json()
 
     @staticmethod
     def get_repositories(name: str, language: str, username: str = "", token: str = "") -> Any:
@@ -135,27 +210,133 @@ def get_repo_name(github_url: str) -> str:
     @staticmethod
     def get_github_info(repository_url: str, username: str = "",
                         token: str = "") -> get_github_info_type:
+        """This method used to iterate through all resource pages of
+           GitHub's /tags API, collect the results, then return a huge
+           list with all results.
+           Removed because this approach does not scale well and we did
+           encounter projects with tens of thousands of tags.
         """
-        Query tag infos from GitHub.
-
-        In the good case a list of tags entries (= dictionaries) is returned.
-        In the bad case a JSON error message is returned.
+        raise NotImplementedError(
+            "Removed with introduction of get_matchting_source_tag!")
+
+    def _get_github_repo(self, github_ref: str) -> Dict[str, Any]:
+        """Fetch GitHub API object identified by @github_ref.
+           @github_ref can be a simple "<owner>/<repo>" string or any
+                       from the plethora of links that refer to a
+                       project on GitHub.
+           By using urlparse() we save ourselves a little bit of work
+           with trailing queries and fragments, but any @github_ref with
+           colons, where the first colon is not part of '://' will not
+           yield viable results,
+           e.g. 'api.github.com:443/repos/sw360/capycli'.
+        """
+        url = 'api.github.com/repos/'
+        gh_ref = urlparse(github_ref, scheme='no_scheme')
+        if gh_ref.scheme == 'no_scheme':  # interpret @github_ref as OWNER/REPO
+            url += gh_ref.path
+        elif not gh_ref.netloc.endswith('github.com'):
+            raise ValueError(f'{github_ref} is not an expected @github_ref!')
+        elif gh_ref.path.startswith('/repos'):
+            url += gh_ref.path[6:]
+        else:
+            url += gh_ref.path
+        if url.endswith('.git'):
+            url = url[0:-4]
+        url = 'https://' + url.replace('//', '/')
+        repo = {}
+        while 'tags_url' not in repo and 'github.com' in url:
+            repo = self.github_request(url, self.github_name, self.github_token)
+            url = url.rsplit('/', 1)[0]  # remove last path segment
+        if 'tags_url' not in repo:
+            raise ValueError(f"Unable to make @github_ref {github_ref} work!")
+        return repo
+
+    def _get_link_page(self, res: requests.Response, which: str = 'next') -> int:
+        """Fetch only page number from link-header."""
+        try:
+            url = urlparse(res.links[which]['url'])
+            return int(parse_qs(url.query)['page'][0])
+        except KeyError:  # GitHub gave us only one results page
+            return 1
+
+    def get_matching_source_url(self, version: Any, github_ref: str,
+                                version_prefix: Any = None
+                                ) -> str:
+        """Find a URL to download source code from GitHub. We are
+           looking for the source code in @github_ref at @version.
+
+           We expect to match @version to an existing tag in the repo
+           identified by @github_ref. We want to have the source
+           code download URL of that existing tag!
+
+           In order to perform this matching, we must retrieve the tags
+           from GitHub and then analyse them. First, we use
+           get_matching_tag(). If that doesn't yield a positive result,
+           we try to infer a tag for @version, to prevent an exhaustive
+           search over all tags.
         """
-        length_per_page = 100
-        page = 1
-        tags: List[Dict[str, Any]] = []
-        tag_url = "https://api.github.com/repos/" + repository_url + "/tags"
-        query = "?per_page=%s&page=%s" % (length_per_page, page)
-        tmp = FindSources.github_request(tag_url + query, username, token)
-        if not isinstance(tmp, list):
-            return tags
-        tags.extend(tmp)
-        while len(tmp) == length_per_page:
-            page += 1
-            query = "?per_page=%s&page=%s" % (length_per_page, page)
-            tmp = FindSources.github_request(tag_url + query, username, token)
-            tags.extend(tmp)
-        return tags
+        try:
+            repo = self._get_github_repo(github_ref)
+        except ValueError as err:
+            print_yellow("      " + str(err))
+            return ""
+
+        tags_url = repo['tags_url'] + '?per_page=100'
+        git_refs_url_tpl = repo['git_refs_url'].replace('{/sha}', '{sha}', 1)
+
+        res = self.github_request(tags_url, self.github_name,
+                                  self.github_token, return_response=True)
+        pages = self._get_link_page(res, 'last')
+        for _ in range(pages):  # we prefer this over "while True"
+            # note: in res.json() we already have the first results page
+            try:
+                tags = [tag for tag in res.json()
+                        if version_prefix is None
+                        or tag['name'].startswith(version_prefix)]
+                source_url = self.get_matching_tag(tags, version, tags_url)
+                if len(source_url) > 0:  # we found what we believe is
+                    return source_url    # the correct source_url
+
+            except (TypeError, KeyError, AttributeError):
+                # res.json() did not give us an iterable of things where
+                # 'name' is a viable index, for instance an error message
+                tags = []
+
+            new_prefixes = self.tag_cache.filter_and_cache(
+                repo['full_name'], version,  # cache key
+                [self.version_regex.split(tag['name'], 1)[0]
+                 for tag in tags
+                 if self.version_regex.search(tag['name']) is not None])
+
+            for prefix in new_prefixes:
+                url = git_refs_url_tpl.format(sha=f'/tags/{prefix}')
+                w_prefix = self.github_request(url, self.github_name,
+                                               self.github_token)
+                if isinstance(w_prefix, dict):  # exact match
+                    w_prefix = [w_prefix]
+
+                # ORDER BY tag-name-length DESC
+                by_size = sorted([(len(tag['ref']), tag) for tag in w_prefix],
+                                 key=lambda x: x[0])
+                w_prefix = [itm[1] for itm in reversed(by_size)]
+
+                transformed_for_get_matching_tags = [
+                    {'name': tag['ref'].replace('refs/tags/', '', 1),
+                     'zipball_url': tag['url'].replace(
+                        '/git/refs/tags/', '/zipball/refs/tags/', 1),
+                     } for tag in w_prefix]
+                source_url = self.get_matching_tag(
+                    transformed_for_get_matching_tags, version, tags_url)
+                if len(source_url) > 0:  # we found what we believe is
+                    return source_url    # the correct source_url
+            try:
+                url = res.links['next']['url']
+                res = self.github_request(url, self.github_name,
+                                          self.github_token, return_response=True)
+            except KeyError:  # no more result pages
+                break
+        print_yellow("      No matching tag for version " + version + " found")
+        return ""
 
     def to_semver_string(self, version: str) -> str:
         """Bring all version information to a format we can compare."""
@@ -193,8 +374,7 @@ def find_github_url(self, component: Component, use_language: bool = True) -> st
             name_match = [r for r in repositories.get("items") if component_name in r.get("name", "")]
         if len(name_match):
             for match in name_match:
-                tag_info = self.github_request(match["tags_url"], self.github_name, self.github_token)
-                source_url = self.get_matching_tag(tag_info, component.version or "", match["html_url"])
+                source_url = self.get_matching_source_url(component.version, match["tags_url"])
                 if len(name_match) == 1:
                     return source_url
                 elif source_url:
@@ -261,10 +441,7 @@ def find_golang_url(self, component: Component) -> str:
 
                 if repository_name.startswith("https://github.com/"):
                     repository_name = repository_name[len("https://github.com/"):]
-                tag_info = self.get_github_info(repository_name, self.github_name, self.github_token)
-                tag_info_checked = self.check_for_github_error(tag_info)
-                source_url = self.get_matching_tag(tag_info_checked, component_version,
-                                                   repository_name, version_prefix or "")
+                source_url = self.get_matching_source_url(component_version, repository_name, version_prefix)
 
         # component["RepositoryUrl"] = repository_name
         return source_url
@@ -284,26 +461,15 @@ def get_github_source_url(self, github_url: str, version: str) -> str:
 
         if self.verbose:
             print_text("      repo_name:", repo_name)
-
-        tag_info = self.get_github_info(repo_name, self.github_name, self.github_token)
-        tag_info_checked = self.check_for_github_error(tag_info)
-        return self.get_matching_tag(tag_info_checked, version, github_url)
+        return self.get_matching_source_url(version, repo_name)
 
     def check_for_github_error(self, tag_info: get_github_info_type) -> List[Dict[str, Any]]:
-        if isinstance(tag_info, list):
-            # assume valid answer
-            return tag_info
-
-        # check for 'rate limit exceeded' message
-        if "message" in tag_info:
-            if tag_info["message"].startswith("API rate limit exceeded"):
-                print_red("GitHub API rate limit exceeded - aborting!")
-                sys.exit(ResultCode.RESULT_ERROR_ACCESSING_SERVICE)
-            if tag_info["message"].startswith("Bad credentials"):
-                print_red("Invalid GitHub credential provided - aborting!")
-                sys.exit(ResultCode.RESULT_ERROR_ACCESSING_SERVICE)
-
-        return []
+        """This method was introduced to check the output of
+           get_github_info() for errors.
+           Removed, because get_github_info was removed.
+        """
+        raise NotImplementedError(
+            "Removed with introduction of get_matchting_source_tag!")
 
     def get_matching_tag(self, tag_info: List[Dict[str, Any]], version: str, github_url: str,
                          version_prefix: str = "") -> str:
@@ -369,7 +535,7 @@ def get_source_url_from_release(self, release_id: str) -> str:
                 if release_details:
                     source_url = release_details.get("sourceCodeDownloadurl", "")
                     if self.verbose:
-                        print("getting source url from get from sw360 for release_id " + release_id)
+                        print("    getting source url from get from sw360 for release_id " + release_id)
                     if source_url != "":
                         return source_url
                     break
@@ -468,7 +634,8 @@ def find_source_url_recursive_by_sw360(self, component: Component) -> str:
 
     @staticmethod
     def find_source_url_by_language(component: Component) -> str:
-        capycli.dependencies.javascript.GetJavascriptDependencies().try_find_component_metadata(component, "")
+        if hasattr(capycli, 'dependencies'):
+            capycli.dependencies.javascript.GetJavascriptDependencies().try_find_component_metadata(component, "")
         url = CycloneDxSupport.get_ext_ref_source_url(component)
         if isinstance(url, XsUri):
             return url._uri