Fix extraction push race and churn; pin deps; bump actions

.github/workflows/extract-pdf-text.yaml

@@ -7,20 +7,30 @@ on:
       - "corpus/**/*.pdf"
       - "corpus/MANIFEST.yaml"
       - "scripts/extract_pdf_text.py"
+      - "scripts/requirements.txt"
   workflow_dispatch: {}
 
 permissions:
   contents: write
 
+# Merges that land close together would otherwise race to push to main and
+# all but one would fail. Queue runs instead; each one checks out the latest
+# main, so the last run always covers everything merged before it.
+concurrency:
+  group: extract-pdf-text
+  cancel-in-progress: false
+
 jobs:
   extract:
     runs-on: ubuntu-latest
     steps:
       - name: Check out repo
-        uses: actions/checkout@v4
+        uses: actions/checkout@v5
+        with:
+          ref: main
 
       - name: Set up Python
-        uses: actions/setup-python@v5
+        uses: actions/setup-python@v6
         with:
           python-version: "3.11"
 
@@ -39,5 +49,11 @@ jobs:
             echo "No changes to commit."
           else
             git commit -m "Extract PDF text for search index ($(date -u +%Y-%m-%d))"
-            git push
+            # A push can still lose to a merge that lands mid-run; rebase onto
+            # it and retry. _extracted/ is only written here, so no conflicts.
+            for i in 1 2 3; do
+              git push && exit 0
+              git pull --rebase origin main
+            done
+            exit 1
           fi

.github/workflows/process-inbox.yaml

@@ -24,7 +24,7 @@ jobs:
       # be written to, so they get a dry run and the report in the job summary.
       SAME_REPO: ${{ github.event.pull_request.head.repo.full_name == github.repository }}
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v5
         with:
           ref: ${{ github.event.pull_request.head.ref }}
           repository: ${{ github.event.pull_request.head.repo.full_name }}
@@ -34,7 +34,7 @@ jobs:
           # processed commit.
           token: ${{ secrets.CORPUS_BOT_TOKEN || github.token }}
 
-      - uses: actions/setup-python@v5
+      - uses: actions/setup-python@v6
         with:
           python-version: "3.11"
 

.github/workflows/validate-corpus.yaml

@@ -6,18 +6,19 @@ on:
       - "corpus/**"
       - "inbox/**"
       - "scripts/validate_corpus.py"
+      - "scripts/requirements.txt"
 
 jobs:
   validate:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v6
         with:
           python-version: "3.11"
 
       - name: Install dependencies
-        run: pip install pyyaml
+        run: pip install -r scripts/requirements.txt
 
       - name: Validate corpus manifest
         run: python scripts/validate_corpus.py --repo-root .

scripts/extract_pdf_text.py

@@ -63,10 +63,20 @@ def extract_one(repo_root: str, pdf_path: str, resource_id: str, license_: str)
         f"\n"
     )
 
+    body = "\n\n".join(pages) + "\n"
+
+    # Leave unchanged files alone; otherwise every run rewrites every file
+    # just to bump the `extracted:` date in the header.
+    if os.path.isfile(out_abs):
+        with open(out_abs, "r", encoding="utf-8") as f:
+            old = f.read()
+        if old.startswith(header.split("# extracted:")[0]) and old.endswith("\n\n" + body):
+            print(f"[extract_pdf_text] unchanged {out_rel}")
+            return False
+
     with open(out_abs, "w", encoding="utf-8") as f:
         f.write(header)
-        f.write("\n\n".join(pages))
-        f.write("\n")
+        f.write(body)
 
     print(f"[extract_pdf_text] wrote {out_rel} ({len(pages)} pages)")
     return True
@@ -111,7 +121,7 @@ def main() -> None:
             if extract_one(repo_root, pdf_path, r["id"], r.get("license", "UNKNOWN")):
                 wrote += 1
 
-    print(f"[extract_pdf_text] done: {wrote} file(s) extracted")
+    print(f"[extract_pdf_text] done: {wrote} file(s) written")
 
 
 if __name__ == "__main__":

scripts/requirements.txt

@@ -1,2 +1,2 @@
-pyyaml
-pypdf
+pyyaml==6.0.3
+pypdf==6.21.4