Refactor ServiceWorker tests for inclusion in WPT

third_party/WebKit/LayoutTests/external/wpt/service-workers/service-worker/navigation-preload/resources/request-headers-scope.py

Index: third_party/WebKit/LayoutTests/external/wpt/service-workers/service-worker/navigation-preload/resources/request-headers-scope.py
diff --git a/third_party/WebKit/LayoutTests/external/wpt/service-workers/service-worker/navigation-preload/resources/request-headers-scope.py b/third_party/WebKit/LayoutTests/external/wpt/service-workers/service-worker/navigation-preload/resources/request-headers-scope.py
new file mode 100644
index 0000000000000000000000000000000000000000..8482ad0d762b5996830b4c553f3d453ebfa7fd90
--- /dev/null
+++ b/third_party/WebKit/LayoutTests/external/wpt/service-workers/service-worker/navigation-preload/resources/request-headers-scope.py
@@ -0,0 +1,9 @@
+import json
+
+def main(request, response):
+    normalized = dict()
+
+    for key, values in dict(request.headers).iteritems():
+        normalized[key.upper()] = values[-1]

[falken, 2017/03/07 06:25:59]

I am a little confused why values[-1] is used. I get that |headers| is a map
from |key| to a list of |values|. Do we expect |values| to have multiple items?
Why is the last item of the list used?

[mike3, 2017/03/07 16:47:03]

I was operating under the incorrect assumption that HTTP headers were
interpreted using a "last one wins" heuristic. I was hoping to keep the
particulars of the Python header parsing out of the assertions themselves, but
I suppose that it's unavoidable to have some artifact when translating from
"header fields" to "JSON object". I'll simply return `values` unaltered and
update the assertions to expect single-element array values.

[falken, 2017/03/09 04:18:10]

Makes sense, thanks for making this rigorous.

+
+    return json.dumps(normalized)