Currently we pump in the twitter keys through a command line argument. This is fine but it doesn't work so well for automated deployment to the server. It would be good if we could migrate these into environment variables perhaps. We still need a good way of getting these onto the server though, something I've not done much of in the past. Any ideas for managing environment variables on EC2 would be smashing.

We scrape the site once every minute from start up. This works fine but if two updates are posted between a scrape then the app will only tweet one. It would be good to get a more robust way of checking the incidents that we scrape. That way we can be sure we can get all the updates. It's also important we don't repeat ourselves though!

It would make testing and dev much easier if there was a toggle to just print to stdout and ignore the twitter reporter / other reporters.

Something like a -d flag perhaps?

Gun Violence Archive would appreciate a shout out in the tweets. Preferably something like the example below, we need to be aware of the 140 character limit though.

This would be good for a beginner, it could simply be added in the TwitterReporter for now. Alternatively we could add it to the Incident class and pass it though.

Example they suggested:

Source: GVA

It looks like quite a few suicides are coming through on the scraper, it would be a good idea to have some sort of content filter if possible that attempts to block the posting of suicide reports. They don't represent the issue correctly and so it would be useful to not add them to the posts.

This is the number 1 back lash we seem to be getting right now so it's worth fixing in the next few days.

Not too fussed about 100% coverage, that would be overkill for something so small, but it would be good to get a few tests going in the following areas.

• - Scraper class could do with some coverage, we'll probably need to pull out the code that makes the web request to make this easier to test.
• - IncidentStore needs a bit of testing. It interacts with the file system a lot so this will probably be more integration tests. The code is so thin that splitting it out would be pointless.
• - Incident class, it's just a PORO for holding data but we've got some read only attributes so we may as well test those.

Raygun.io have given me a free account to use for this bot on twitter. It would be good to add raygun as an optional error reporter that you can toggle on start up, this would mean we're not forcing it if someone doesn't have it.

Either way we need to get some basic raygun in there to allow error tracking easily!

They also want some credits in the readme which are sitting on an email I have but that's something I can easily do after the fact.