switch to polling for loading to circumvent gateway timeout

pyproject.toml

@@ -30,6 +30,9 @@ pydantic = "^2.3.0"
 black = "^22.8.0"
 types-Flask = "^1.1.6"
 
+[tool.poetry.group.dev.dependencies]
+hypercorn = "^0.15.0"
+
 [build-system]
 requires = ["poetry-core>=1.0.0"]
 build-backend = "poetry.core.masonry.api"

src/herald/metric.py

@@ -38,7 +38,7 @@
 artifact_size = Histogram(
     "herald_artifact_size_bytes",
     "Size of artifacts",
-    buckets=[1e6, 1e7, 1e8] + [n*1e8 for n in range(2, 10)] + [1e9] + [float("inf")],
+    buckets=[1e6, 1e7, 1e8] + [n * 1e8 for n in range(2, 10)] + [1e9] + [float("inf")],
 )
 artifact_size_rejected = Counter(
     "herald_artifact_size_rejected_total",

src/herald/templates/loading.html

@@ -5,11 +5,6 @@
 {%- endblock %}
 
 {% block content %}
-			<div
-			hx-get="{{ url_for('view', owner=owner, repo=repo, artifact_id=artifact_id, file=file) }}"
-			hx-trigger="load"
-	 		hx-swap="innerHTML"
-			>
 			<div class="level">
 				<div class="level-left">
 					<div class="level-item">
@@ -29,5 +24,10 @@
 						<img src="{{ url_for('static', filename='spinner.svg') }}" alt="Loading" />
 					</div>
 				</div>
+
+			<div hx-get="{{ url_for('view_poll', owner=owner, repo=repo, artifact_id=artifact_id) }}"
+					hx-trigger="load"
+					hx-swap="outerHTML" style="display:none;">
 			</div>
+
 	{% endblock %}

src/herald/web.py

@@ -225,41 +225,9 @@ async def async_get_file() -> Tuple[IO[bytes], str]:
                     to_png=to_png,
                 )
 
-            @stream_with_context
-            async def reload_response():
-                yield await render_template(
-                    "loading.html",
-                    file=file,
-                    repo=f"{owner}/{repo}",
-                    artifact_id=artifact_id,
-                )
-                try:
-                    await async_get_file()
-                except Exception as e:
-                    details: str | None = None
-                    if isinstance(e, github.ArtifactExpired):
-                        details = f"Artifact #{artifact_id} has expired on GitHub"
-                        artifact_expired[artifact_id] = True
-                    if isinstance(e, github.ArtifactTooLarge):
-                        details = f"Artifact #{artifact_id} is too large to download"
-                    elif isinstance(e, fs.errors.ResourceNotFound):
-                        details = f"File {file} not found in artifact"
-
-                    message = json.dumps(
-                        await render_template(
-                            "error.html",
-                            file=file,
-                            repo=f"{owner}/{repo}",
-                            artifact_id=artifact_id,
-                            details=details,
-                        )
-                    )
-                    yield f"""
-                        <script>
-                        document.getElementById("content").innerHTML = {message}
-                        </script>"""
-                    return
-                yield "<script>" + "window.location.reload()" + "</script>"
+            is_cached = gh.is_artifact_cached(f"{owner}/{repo}", artifact_id)
+
+            is_htmx = "HX-Request" in request.headers
 
             # Assumption: curl etc will `Accept` *anything*
             is_browser: bool = request.headers.get("Accept", "*/*") != "*/*"
@@ -268,24 +236,24 @@ async def reload_response():
                     "We think this is a browser request based on Accept header %s",
                     request.headers.get("Accept"),
                 )
-            if (
-                gh.is_artifact_cached(f"{owner}/{repo}", artifact_id)
-                or not is_browser
-                or not config.ENABLE_LOADING_PAGE
-            ):
+            if is_cached or not is_browser or not config.ENABLE_LOADING_PAGE:
                 logger.debug("File is cached, call and return immediately")
                 buf, mime = await async_get_file()
                 response = await make_response(buf.read())
                 response.headers["Content-Type"] = mime
                 response.headers["Cache-Control"] = "max-age=31536000"
                 response.headers["Etag"] = exp_etag
 
-                if "HX-Request" in request.headers:
-                    response.headers["HX-Redirect"] = request.url
+                response.headers["HX-Redirect"] = request.url
+                response.headers["HX-Redirect"] = request.url
 
             else:
                 logger.debug("File is not cached")
 
+                if is_browser:
+                    logger.debug("Starting artifact download in the background")
+                    app.add_background_task(async_get_file)
+
                 return redirect(
                     url_for(
                         "loading",
@@ -313,6 +281,39 @@ async def reload_response():
         #  except fs.errors.ResourceNotFound:
         #  abort(404)
 
+    @app.route("/poll/<owner>/<repo>/<int:artifact_id>")
+    @app.route("/view/<owner>/<repo>/<int:artifact_id>/<path:file>")
+    async def view_poll(owner: str, repo: str, artifact_id: int, path: str = ""):
+        is_cached = gh.is_artifact_cached(f"{owner}/{repo}", artifact_id)
+        logger.debug(
+            "Polling for %s/%s #%d => %s, is cached: %s",
+            owner,
+            repo,
+            artifact_id,
+            path,
+            is_cached,
+        )
+
+        if is_cached:
+            return (
+                "",
+                200,
+                {
+                    "HX-Redirect": url_for(
+                        "view",
+                        owner=owner,
+                        repo=repo,
+                        artifact_id=artifact_id,
+                        file=path,
+                    )
+                },
+            )
+        poll_url = url_for("view_poll", owner=owner, repo=repo, artifact_id=artifact_id)
+
+        return f"""
+            <div hx-get="{poll_url}" hx-trigger="load delay:2s" hx-swap="outerHTML" style="display:none;"></div>
+            """
+
     @app.route("/loading/<owner>/<repo>/<int:artifact_id>/")
     @app.route("/loading/<owner>/<repo>/<int:artifact_id>/<path:file>")
     async def loading(owner: str, repo: str, artifact_id: int, file: str = ""):