Improved model detection

[CloCkWeRX]

In #904 we talked a little about flags to help detect the right model for a LUT.

Annoyingly, with companies like Arlec, or Mirabella Genio; it's all Tuya under the hood; and their product numbers, model numbers vs what is presented through the various APIs all vary.

https://www.mirabellagenio.com.au/product-range/mirabella-genio-wi-fi-dimmable-5w-g95-filament-led-bulb/
Packaging, globe itself feature identifiers like:
Mirabella Genio Wi-Fi Dimmable 5W G95 ES CCT Filament LED Bulb
Item # I002747

Presented through the office Tuya integration, and viewing Devices > Device Diagnostics:

"data": {
    ...
    "product_id": "v5ixexiafmjifkfr",
    "product_name": "Genio Smart WIFI bulb G95",
}

Via localtuya, with the cloud integration turned on:

  "data": {
    "device_config": {
      ....
      "device_id": "bfad661ea9e8cc77b9gola",
      "product_key": "key8u54q9*****w",
      "model": "Genio Smart WIFI bulb G95"
    },
    "device_cloud_info": {
      ...
      "product_id": "v5ixexiafmjifkfr",
      "product_name": "Genio Smart WIFI bulb G95",
      ...

While ideally this would be consistent, and everyone would just populate "model" with an agreed identifier... it's not.
I imagine a variety of other bulbs

Possible solutions

Directory structure - messy, likely to cause conflicts, etc.

An array of aliases? - this is better, as you get multiple possible matches; but if it's just an array of strings trying to map against model name; you can't look up other attributes

{
    "name": "Arlec Smart R80 9.5W 806lm CCT Bulb",
    "aliases": [
       'xyz',
      'GLD115HA'
    ],
    "standby_power": 0.58,
    "supported_modes": [
        "lut"
    ]
}

An array of alias regexps? - also good, allows wildcards, etc, possibly a bit of a pain encoding in JSON. Built in support in python; fairly well understood. Doesn't know about other attributes to look at though.

{
    "name": "Arlec Smart R80 9.5W 806lm CCT Bulb",
    "aliases": [
       '/xyz*/',
      '/GLD\d+HA/i'
    ],
    "standby_power": 0.58,
    "supported_modes": [
        "lut"
    ]
}

An array of JSONPaths? - More complicated; requires an external library, but https://pypi.org/project/jsonpath-ng/ does seem to support both paths and regex matchers:

query = [x.value for x in parser.parse('$[?(@.Worker =~ ".*880099.*")]').find(json_data)]
print(json.dumps(query))

{
    "name": "Arlec Smart R80 9.5W 806lm CCT Bulb",
    "aliases": [
       '@.product_id =~ /xyz\d+/',
       '@.product_name == /Arlec Smart R80 9.5W 806lm/',
    ],
    "standby_power": 0.58,
    "supported_modes": [
        "lut"
    ]
}

How would the configuration flow work?

For each model.json, where !exact match, present a list of suggested matches based on .aliases.

[bramstroker]

The reason why the model aliases are not in model.json yet is because I wanted to eliminate unnecessary I/O as much as possible. To know all possible model aliases I have to load all model.json files before the setup procedure starts. So that's why aliases are here currently: https://github.com/bramstroker/homeassistant-powercalc/blob/master/custom_components/powercalc/aliases.py

I agree it will be much cleaner when we can just add these aliases in the model.json files to have everything about the model together. I think I will revisit this in the nearby future as it makes a lot of sense to have this included in model.json. Maybe I will create an github action to write all the model aliases to a python dictionary to keep the zero IO approach, as the aliases only change when a PR is submitted.

Currently you can already add more than 1 alias in aliases.py for a given model. However it is only an exact match.
Will see what I can do to support regexes here or even json path. External libraries can be included, but I don't want them to be too much bloated. Also this needs to be fast as every light (and in the future more than lights) needs to be matched against every possible alias.
So a lot of things to consider and test before this can be implemented.

[bramstroker]

I did a lot of work to refactor and cleanup some code regarding to the power profile library and added support to add aliases to the model.json files. See #942
This is the first iteration, we can have a look into more extensive matching (regex, json path) after this has been released.

Need to do some more work on it, but I think I might wrap it up today.

Would you like to take on the task of moving the model aliases from aliases.py to their respective model.json files?

[CloCkWeRX]

Sure. About to head out for dinner, but will take a look either later tonight or early tomorrow.