When streaming API is disconnected, poll home/notifications

[Gargron]

• Display slightly different empty home timeline message if user is following others (see Empty home timeline after period of inactivity #2619)
• Cull notifications to 20 items when over 40 get added in real-time, just like happens in other columns
• Run manage:translations

[ykzts]

I think it would be better to create methods like renderEmptyMessage and renderEmptyMessageHasFollows. Execution of React.createElement takes some time.

const emptyMessage = hasFollows ? this.renderEmptyMessage() : this.renderEmptyMessageHasFollows();

[Gargron]

Unless that is moved out of the class I don't see how it would improve performance, since renderEmptyMessage() would also execute React.createElement live?

[ykzts]

Yes. But, https://github.com/tootsuite/mastodon/pull/2776/files#diff-c520240100d5372b6dfb9299ff13e7deR24 is always running.

[Gargron]

Ah! I see what you mean. This could also be an if/else then. Fair enough

[ykzts]

I agree.

let emptyMessage;

if (hasFollws) {
  emptyMessage = ...;
} else {
  emptyMessage = ...;
}

It is also good to write like the above.

[ykzts]

note: duplicated with https://github.com/tootsuite/mastodon/pull/2776/files#diff-f2c27f858abca0ea3fa97f9400fca112L84

[Gargron]

This is all a result of yarn run manage:translations, which sorts JSON locale files alphabetically, etc.

[ykzts]

Yes. This is a comment for when other people saw it.

[nightpool]

Is it common practice to add the english translations to the specific locale files? Can't we use a default message instead, to avoid polluting the files?

(or is it good because it calls out what needs to be changed, I guess?)

[szbalint]

Do I understand it right that the fallback would happen silently from the streaming api to polling?

If so, would it make sense to configuration-gate this? Otherwise it'd be quite hard to notice if the streaming api is unexpectedly down as opposed to running on a hosting environment where it's not supported.

[Gargron]

@szbalint You would see WS connection failing in console logs. What kind of configuration-gate are you thinking of?

[szbalint]

@Gargron I mean for this feature from a maintainability/ops perspective:

• an instance either runs the WS streaming service in which case it's better if that fails visibly when the service is down, as falling back to notifications would mask any issues until someone checks the console
• or an instance can't/won't run WS based streaming in which case there shouldn't be connection attempts to WS/defined endpoint in the configuration file.

It looks to me that those two usecases aren't overlapping so it'd make sense for me if instance admins can decide in the .env.production config or somewhere else whether they run the streaming service or not and if not that's when they'd get polling as a fallback.