Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Jun 23 2015 20:15
    gitter-badger opened #1
Stuart P. Bentley
@stuartpb
So I was originally thinking command parsing would be done with something like separating on spaces into an array of words and then just using a state machine for contexts moving forward, but now I'm wondering if it wouldn't be better to use a Parsing Expression Grammar
Stuart P. Bentley
@stuartpb
When @command asks you to clarify an ambiguous term like "everything I said", I guess it should detect if you give it a context, and you can optionally decorate that context with "I mean" and/or "everything"
Stuart P. Bentley
@stuartpb
"chown <context> to <owner>" should be a command
Stuart P. Bentley
@stuartpb
Or... chown is overcomplicated, but "chgroup" is fine
@command is designed to take commands that don't require domain knowledge for human observers in the room to understand (so not like Unix commands, except for the commands that look kind of like Unix commands like "chown" which should only be used in groups with technical familiarity with the Unix syntax)
Stuart P. Bentley
@stuartpb
And secondarily for moderators / admins to naturally be able to compose, although not perfectly (it's worth adding syntax rules to make constructions less awkward, but the aim isn't to make command able to understand anything that anybody would say to it)
The idea is that people generally learn @command uses by osmosis, imitating things virally based on what people have seen the @command developers do (and from @command's suggestions back, and maybe some blog posts explaining how one can use @command to achieve X goals)
(and by reading the source code for @command)
@command development is (like all good development, pretty much) primarily need-driven: developer wants command to be able to do something, it doesn't work, developer goes and writes a PR to make command capable of understanding the command the developer wrote, it gets merged (presuming that the command the developer wrote is designed in a way that's broad enough that it could be conceivably needed in other circumstances in the future), developer re-reuns the command successfully .
Stuart P. Bentley
@stuartpb
so like "retag the discussion we just had about rockets", that's too vague of a command / too hairy of a job to puzzle out - the human needs to do a little leg work to decide how to quantify "the discussion we just had about rockets" in query terms
but "retag messages tagged #rockets before today as #explosives", if that doesn't work, yes submit a PR to fix the aspect that's not understood by command (like if it rejects "before today" like it's not understandable)
How should "today" be implemented? User local time, switches at midnight?
Stuart P. Bentley
@stuartpb
deleted from Slack (because discussion around @command on Slack has gotten a little disproportionate and heated): the point of @command is so that if I want to retag all my messages that aren't tagged docker in the last 5 days, my options are better than either "do it manually" or "sift through an abstruse visual query-builder to find the right buttons to click to explain the relation I could just type"
imagine if Google didn't have a search bar, you had to construct your search based on clicking together a rebus based on random images Google presented for their most popular topics because "clicking is easier than typing". That is what not having @command would be like.
Stuart P. Bentley
@stuartpb
so... should admins be allowed to answer questions posed by @command to other users? I'm thinking "no", with @command noting this is the case if you try as such.