fedora-infra / datanommer

Put all the messages in the postgres
GNU General Public License v3.0
12 stars 25 forks source link

Datanommer does not store the `agent_name` #1309

Closed abompard closed 1 month ago

abompard commented 2 months ago

The message's agent (in the agent_name property) is not store specifically by datanommer, only as part of the usernames. It would be interesting to filter on the agent name only, see https://github.com/fedora-infra/fedbadges/issues/100

abompard commented 2 months ago

Obviously this will require a schema change, and it will increase the DB size (data + index). This should be a column in the messages table, because there's only one agent_name, and joins to a hypertable are expensive in timescaledb. The extract-users command should be updated to take it into account. The agent and not_agent arguments should be added to the Message.grep() Also, datagrepper should be updated to add the new filters