Separate config section definition from bot setup #2597
Labels
Core/Plugin Handling
Feature
Long-term Planning
Things that need to happen at some point in the future, but need to NOT happen soon.
Low Priority
Milestone
Plugin authors must currently remember to
define_section()
in two plugin hooks:setup()
, so the plugin's configuration is registered when Sopel runsconfigure()
, so the plugin's configuration is registered when the interactive config wizard runsThis is a minor inconvenience, true, but it leads directly to a larger problem: Plugin configuration definitions aren't available outside of these two scenarios. If the bot isn't in the config wizard, and isn't running normally, option validation is unavailable.
Implementing command-line
sopel-config set
functionality (as I started doing a while ago in #1986) becomes tricky, therefore, if we want the CLI to be safe and reject invalid values (such as options not available in aChoiceAttribute
). Plugins'StaticSection
definitions have to be available to the CLI, but running a plugin'ssetup()
can come with side effects.I'm sure we can just
setup()
plugins before attempting to handle a hypotheticalset
orunset
CLI action, and things would be fine most of the time. I think we can do better long-term, though, and that's what this issue is about: planning for the long term.In some minor release (hopefully within 8.x), we can add a new way for plugin authors to register their config sections that isn't tied to the plugin's
setup()
orconfigure()
hooks. It will have to be implemented such that the relevant submodules can do (pseudocode):But assuming we can devise a sensible mechanism, the new and old ways of defining a config section can coexist until the next major Sopel version when we'd remove
Config.define_section()
from the API.Considerations
We are theoretically trying to avoid reserving new top-level object names. The simplest way to do so is making any
StaticSection
subclass auto-loaded, but the catch is that aStaticSection
on its own doesn't have a name; the name used in .cfg files comes from the call toConfig.define_section()
.However, that leaves us with a natural fallback point: Auto-register any
StaticSection
subclass that has been given a name as part of its definition (using a yet-undetermined mechanism), and leaveConfig.define_section()
to register and name existing subclasses (until the plugin author updates to the new style).I admit that I haven't had time to think this specific implementation idea all the way through, but it's also just fine to get a gut reaction from the other maintainers when we start on 8.1 development and figure out the path forward together.
The text was updated successfully, but these errors were encountered: