Add search integration

[balloob]

Description:

Our data is interconnected, making it a graph. This means that we can also search it as a graph. This integration can give you related items based on any node in the graph. Currently implemented areas, devices, entities, config entry, scene. Leaving script/automation for a future PR.

Goal of this integration will be to provide related data in the UI. Example use cases:

• Show scenes that impact a specific area
• List automations that use a specific input_boolean

Checklist:

• The code change is tested and works locally.
• Local tests pass with tox. Your PR cannot be merged unless tests pass
• There is no commented out code in this PR.
• I have followed the development checklist

If the code does not interact with devices:

• Tests have been added to verify that the new code works.

[MartinHjelmare]

Looks good.

[balloob]

I've added support for scenes and config entries.

When we find a config entry, scene, automation or script (the outputs!), we will no longer further resolve it into subtypes. Because if I query for a device, I don't want to return all devices that share a scene or config entry with the device that I am querying for.

It's still possible to pick any of those as an entry point, in which case they will be resolved. So these types are search type only.

Support for script/automation should move to a new PR to no longer further complicate this.

[balloob]

This is ready for approval/merging.

[awarecan]

The result is not a graph, it only contains all vertices, but missing the edges.

We can extend the search methods to return a tuple (vertices, edges).

self.results = defaultdict(set)  #vertices
self.edges = defaultdict(set)

# in Searcher class
def search(self, item_type, item_id):
  self.results[item_type].add(item_id)
  self._to_resolve.add((item_type, item_id))

  while self._to_resolve:
    search_type, search_id = self._to_resolve.pop()
    await getattr(self, f"_resolve_{search_type}")(search_id)

  return (self.results, self.edges)

# for example device
def _resolve_device(self, device_id) -> None:
  device_entry = self._device_reg.async_get(device_id)
  if device_entry is not None:
    if device_entry.area_id:
      self._add_or_resolve("area", device_entry.area_id)
      # add edge
      self.edges[("device", device_id)].add(("area", device_entry.area_id))
    for config_entry_id in device_entry.config_entries:
      self._add_or_resolve("config_entry", config_entry_id)
      # add edge
      self.edges[("device", device_id)].add(("config_entry", config_entry_id))

  for entity_entry in entity_registry.async_entries_for_device(
    self._entity_reg, device_id
  ):
    self._add_or_resolve("entity", entity_entry.entity_id)
    # add edge
    self.edges[("device", device_id)].add(("entity", entity_entry.entity_id))

[awarecan]

We can keep it in vertices if we return a graph

[balloob]

Well I wonder if we should return it as a graph? The user just wants to see related data. It doesn't care much about the relationship maybe?

[Jc2k]

Nice!

Will this gain other ways of searching in the future or is it just for finding related objects? (Wondering if there is a more descriptive name if its just for finding related objects).

[balloob]

@Jc2k for now it's just related items. I'll update the websocket endpoint.