1
0
Fork 0
mirror of https://we.phorge.it/source/phorge.git synced 2025-01-31 00:48:21 +01:00
Commit graph

3 commits

Author SHA1 Message Date
epriestley
d5995d574d Formalize targets (users and channel) into objects
Summary:
Make users/channels/rooms into objects, so we can later sort out stuff like Campfire user IDs, Phabricator vs chat accounts, etc.

The only change here is that I removed output buffering from the macro handler. We should move throttling/buffering to adapters instead and have it apply globally.

Test Plan: Ran IRC and Campfire bots and interacted with them.

Reviewers: indiefan

Reviewed By: indiefan

CC: aran

Differential Revision: https://secure.phabricator.com/D4924
2013-02-14 05:13:38 -08:00
indiefan
431e2bee6e First (rough) pass at campfire protocol adapter for bot.
Summary:
Decided the best approach for refactoring the message/command stuff would be to actually start implementing the campfire adapter to get a better idea of what the abstractions should look like. It feels awkward and unwieldy trying to maintain the irc command interface (notice the message instantiation in the `processReadBuffer()` method. However, i'm still not clear what the best approach is without requiring a re-write of nearly all the existing handlers and defining essentially a custom dsl on top of irc's.

I suppose given that alternative, implementing to irc's dsl doesn't sound all that bad. Just feels like poor coupling.

Also, I know that there is some http stuff in libphutil's futures library, but the https future is shit and I need to do some custom curlopt stuff I wasn't sure how to do with that. But if you think this should be refactored, let me know.

I tested this with the ObjectHandler (messages with DXXX initiate the bot to respond with the title/link just as with irc), but beyond that, I haven't tried any of the other handlers, so if there are complications you think i'm going to run into, just let me know (this is one of the reasons for requesting review early on).

Also, this diff is against my last one, even though that hasn't been merged down yet. It was starting to get large and I'd prefer to keep to two conversations separate.

Fixing some lint issues.

Test Plan: Ran the bot with the Object Handler in campfire and observed it behaving properly.

Reviewers: epriestley

Reviewed By: epriestley

CC: aran, Korvin

Maniphest Tasks: T2462

Differential Revision: https://secure.phabricator.com/D4830
2013-02-07 06:34:06 -08:00
indiefan
574bc3ba31 First pass at decoupling Phabricator bot behavior from the protocol it's running on, this pulls the connection, reading, and writing functionalities out of the bot itself and into the adapter.
Summary:
Ugh, just wrote out a huge message, only to lose it with a fat-fingered ctrl-c. Le sigh.

First pass at decoupling the bot from the protocol. Noticeably absent is the command/message coupling. After this design pass I'll give that a go. Could use some advice, thinking that handlers should only create messages (which can be public or private) and not open ended, undefined 'commands'. The problem being that there needs to be some consistant api if we want handlers to be protocol agnostic. Perhaps that's a pipedream, what are your thoughts?

Secondly, a few notes, design review requests on the changes i did make:
 # Config. For now i'm passing config through to the adapter. This was mainly to remain backwards compatible on the config. I was thinking it should probably be namespaced into it's own subobject though to distinguish the adapter config from the bot config.
 # Adapter selection. This flavor is the one-bot-daemon, config specified protocol version. The upside is that in the future they won't have to run different daemons for this stuff, just have different config, and the door is open for multiple protocol adapters down the road if need be. The downside is that I had to rename the daemon (non-backwards compatible change) and there will need to be some sort of runtime evaluation for instatiation of the adapter. For now I just have a crude switch, but I was thinking of just taking the string they supply as the class name (ala `try { new $clasName(); } catch...`) so as to allow for homegrown adapters, but I wasn't sure how such runtime magic would go over. Also, an alternative would be to make the PhabricatorBot class a non-abstract non-final base class and have the adapters be accompanied by a bot class that just defines their adapter as a property. The upside of which is backwards compatibility (welcome back PhabricatorIRCBot) and perhaps a little bit clearer plugin path for homegrowners.
 # Logging. You'll notice I commented out two very important logging lines in the irc adapter. This isn't intended to remain commented out, but I'm not sure what the best way is to get logging at this layer. I'm wary of just composing the daemon back down into the adapter (bi-directional object composition makes my skin crawl), but something needs to happen, obviously. Advice?

That's it. After the feedback on the above, you can either merge down, or wait until i finish the command/message refactor if you don't think the diff will grow too large. Up to you, this all functions as is.

Test Plan: Ran an irc bot, connected, read input, and wrote output including handler integration.

Reviewers: epriestley

Reviewed By: epriestley

CC: aran, Korvin

Maniphest Tasks: T2462

Differential Revision: https://secure.phabricator.com/D4757
2013-02-05 18:46:54 -08:00
Renamed from src/infrastructure/daemon/irc/PhabricatorIRCMessage.php (Browse further)