Tweak Git image parsing

dribdat/apifetch.py

@@ -15,7 +15,9 @@ from .apievents import (
     fetch_commits_gitlab,
     fetch_commits_gitea,
 )
-from .utils import sanitize_url, load_presets, load_yaml_presets
+from .utils import (
+    sanitize_url, load_presets, load_yaml_presets, fix_relative_links
+)
 from future.standard_library import install_aliases
 install_aliases()
 
@@ -154,17 +156,7 @@ def FetchGithubProject(project_url):
         readme = b64decode(readme['content']).decode('utf-8')
         # Fix relative links in text
         imgroot = "https://raw.githubusercontent.com"
-        readme = re.sub(
-            r"<img src=\"(?!http)",
-            "<img src=\"%s/%s/%s/" % (imgroot, repo_full_name, default_branch),
-            readme
-        )
-        readme = re.sub(
-            r"\!\[(.*)\]\((?!http)",
-            # TODO check why we are using \g escape here?
-            r"![\g<1>](%s/%s/%s/" % (imgroot, repo_full_name, default_branch),
-            readme
-        )
+        readme = fix_relative_links(readme, imgroot, repo_full_name, default_branch)
     return {
         'type': 'GitHub',
         'name': json['name'],

dribdat/utils.py

@@ -166,3 +166,19 @@ def load_yaml_presets(filename, by_col='name', filepath=None):
     with open(fn, mode='r') as file:
         config = load_presets(file, filename, by_col)
     return config
+
+
+def fix_relative_links(readme, imgroot, repo_full_name, default_branch):
+    """Ensures that images in Markdown are absolute."""
+    readme = re.sub(
+        r" src=\"(?!http)",
+        " src=\"%s/%s/%s/" % (imgroot, repo_full_name, default_branch),
+        readme
+    )
+    readme = re.sub(
+        r"\!\[(.*)\]\((?!http)",
+        # Pass named group to include full path in the image URL
+        "![\g<1>](%s/%s/%s/" % (imgroot, repo_full_name, default_branch),
+        readme
+    )
+    return readme

tests/test_repository.py

@@ -20,7 +20,6 @@ class TestRepository:
         assert 'dribdat' in test_obj['description']
         assert 'dribdat/dribdat' in test_obj['source_url']
 
-
     def test_datapackage_dribdat(self):
         """Test parsing a dribdat Data Package."""
         test_url = 'https://raw.githubusercontent.com/dribdat/dribdat/main/tests/mock/datapackage.json'
@@ -70,6 +69,8 @@ class TestRepository:
         assert test_obj['type'] == 'GitHub'
         assert 'commits' in test_obj
         assert len(test_obj['commits']) > 5
+        assert 'src="dribdat/static/img' not in test_obj['description']
+        assert 'src="https://raw.githubusercontent.com/dribdat/dribdat/main/dribdat/static/img' in test_obj['description']
 
     def test_github_other(self):
         """Test parsing a GitHub Markdown file."""

tests/test_sync.py

@@ -8,6 +8,7 @@ from dribdat.aggregation import (
     FetchWebProject,
     ProjectActivity,
 )
+from .utils import fix_relative_links
 from .factories import ProjectFactory
 
 from .mock.project_data import project_data
@@ -68,6 +69,7 @@ class TestSync:
         assert test_obj['source_url'] == test_url
         assert 'Guidelines' in test_obj['description']
 
+
     def test_googledoc(self):
         """Test parsing a Google Document."""
         # Handbook to Hackathons with Dribdat
@@ -76,3 +78,13 @@ class TestSync:
         assert 'description' in test_obj
         assert 'Handbook' in test_obj['description']
 
+
+    def test_fix_relative_links(self):
+        imgroot = "https://raw.githubusercontent.com"
+        repo_full_name = "dribdat/dribdat"
+        default_branch = "main"
+        readme = '![hello there](world.png) <img title="hello" src="again.jpg">'
+        readme = fix_relative_links(readme, imgroot, repo_full_name, default_branch)
+        assert imgroot in readme
+        assert not '(world.png)' in readme
+        assert not '"again.jpg"' in readme