)]}'
{"etc/ceilometer/pipeline.yaml":[{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"c950eee5ce0afa34bb054c1467cf1e5a13cb56cb","unresolved":false,"context_lines":[{"line_number":19,"context_line":"                  name: \"cpu_util\""},{"line_number":20,"context_line":"                  unit: \"%\""},{"line_number":21,"context_line":"                  type: \"gauge\""},{"line_number":22,"context_line":"                  scale: \"100.0 * (10**9 / resource_metadata.get(\u0027cpu_number\u0027, 1))\""},{"line_number":23,"context_line":"              replace: False"},{"line_number":24,"context_line":"    publishers:"},{"line_number":25,"context_line":"        - rpc://"}],"source_content_type":"text/x-yaml","patch_set":4,"id":"AAAAP3%2F%2FsZ0%3D","line":22,"updated":"2013-07-12 21:30:50.000000000","message":"I\u0027m still worried about this resource_metadata.get() call. Should deployers have to know about our data model? Can\u0027t we make \"cpu_number\" a valid value in the namespace? If we want it to be namespaced, we could use metadata.cpu_number and give them some other way to provide a default value.\n\nFor example, if the namespace for the eval() call was a defaultdict() that returned None, then they could do something like:\n\n  100.0 * (10**9 / (resource_metadata.cpu_number or 1))\n\nAlternately, we could let them specify default values in the transformer configuration:\n\n defaults:\n    resource_metadata.cpu_number: 1\n    scale: 100.0 * (10**9 / resource_metadata.cpu_number)\n\nAlthough with that form we have no protection against divide-by-zero, so I think I like the first version better.","commit_id":"d4c01842cc2a0ae21f59e89a69dd6b7ff3cb1780"}]}