Some RSS feeds will edit the published time of an item AFTER putting it out,
which resulted in RSS Bot sending the same article twice. We now remember the
"GUID" field for each item and de-dupe based on that. Normal timestamp
algorithm still applies.
This feels a lot better because now `OnPoll` works in a similar way to
`OnReceiveWebhook` (called on a `Service`) rather than have this strange
global-per-service-type struct.
Because apparently "Feed Reader" is the name of a thing popular enough that we
don't want to step on toes, so let's call it the name of a less popular thing.
Just need to send messages into rooms now for a first cut to be done. Notable
improvements to make:
- We currently do 1 goroutine per service. This could be bad if we have lots of these things running around.
- We do not cache the response to RSS feeds. If we have 10 independent services on the same feed URL, we will
hit the URL 10 times. This is similar to how we currently do 1 webhook/service, so it's plausible that in
the future we will want to have some kind of generic caching layer.
- We don't send messages to Matrix yet. We need a `Clients` instance but can't get at one. There's only ever
one, so I wonder if we should global it like we do with `GetServiceDB()` for ease of use?
- The polling interval is divorced from the actual feed repoll time. Ideally we would schedule the goroutine
only when we need it, rather than checking frequently, determining we have nothing to do, and going back
to sleep.
* WIP: Initial Guggy
* Finish Guggy integration
* typo & send No GIF Found when GIF string empty
* unused function
* Unused import, HTTPS API and GIFs
* APIKey -> api_key
* Use api_key on the wire, APIKey in structs