Add support of attribute tree when available.

[onlinux]

When devices have attributes defined as attr1.attr2.attr3 etc... you can define the attribute as attr1.attr2.attr3 etc

Ex: for RPI monitor :

type: custom:mini-graph-card entities:

• entity: sensor.rpi_raspi64_rpi_monitor_raspi64 attribute: cpu.load_1min_prcnt name: Load Raspi64
• entity: sensor.rpi_monitor_raspi65 attribute: cpu.load_1min_prcnt name: Load Raspi65 unit: '%'

This will prevent the need to create numerous virtual sensors for each sub-attribute you wish to access.

[onlinux]

[akloeckner]

Thanks for the suggestion! I like the feature and it seems to be a quite local change.

However, I feel, there is still a bit of redundancy in the code that should be reduced:

• The logic is spelled out two times. I'd suggest a function for it then.
• A regular attribute is actually an edge case of your new logic, so the else branch should be avoided.
• I'm not sure: is this double-check for an object needed? Why not simply check typeof?
• The logic also feels a bit "literal": there are many return statements and if blocks that could be written in a more compact form.

How about adding a function like this and then just re-use it:

  function getObjectAttr (obj, path) {
      let res = obj;

      path.split('.').forEach(function (key) {
         res = typeof res === 'object' && res ? res[key] : undefined;
      });

      return res;
   }

We might have to adjust the style to the mini-graph-card habits a bit.

Also, there is some housekeeping to be done:

• The feature should be documented. A short hint in the options' table should be enough.
• The screenshot and the changes to package-lock.json can probably be ignored. (My setup also changes the lock file all the time, but you don't use any new dependency.)
• Your PR should target the dev branch.

What do you think?

[ildar170975]

This functionality is very useful. Are these usecases possible?

1. Attribute is a dictionary. A sub attribute is called as “dictattr[“subattr”] - in addition to “dictattr.subattr”.
2. Attribute is a list. An element is called is “listattr[2]”.

Can be tested with a weather entity.

[akloeckner]

Are these usecases possible?

I would keep it simple for a start:

• [] notation will be more difficult to parse and does not give an advantage over . notation.
• listattr[2] will thus likely not be possible, but listattr.2 could be added in a next step, once we have this feature in place. It will just be a matter of converting the string '2' to the number 2...

I suggest to get this merged as simple as possible first. And then extend.