[Plugin] AttributeInspector 1.1.1 – 2014-05-08

Aerilius

A lot of the ideas are things that I had found and added myself, others did I integrate now. Sorry that I still haven't finished/published the update because I want to be sure the quality is right.
(Some extension stores have longer update cycles, so when I discover an overseen issue, I can't publish a quickfix.)

As for searching within the selected entities by attribute, I had thought of that idea, but decided not to get distracted but focus first on the main function of reading/editing attributes. One could certainly add a lot of ideas, but to make a "good" user interface one needs to carefully think about user workflows and how a feature affects the plugin's complexity (of UI and code). I'm not yet sure if search is something that I should better leave for thomthom's Selection Toys.

Checking an entity for a "direct" property is on thing, but speed would also be much more affected for attributes, when you consider every entity can have multiple dictionaries, with each mutiple attributes/values. And Christina found that attribute dictionaries can have their own nested attribute dictionaries…

jiminy-billy-bob

Have you considered browsing through the read/write_default stuff?

Aerilius

A "registry inspector" is - I believe - not possible with thr API. We can only read if we know the registry key to read.

jiminy-billy-bob

Ok!

Dan Rathbun

.. and currently it is very dangerous as (on PC,) the keys written are mixed in among the application's keys.

We can only ASK coders to prefix their keys similar to how they should prefix their plugin directories.
ex: "Plugin_AE_AttributeInspector"

Aerilius

I updated the plugin:
• definitions (component/group/image definitions)
• more data types: points and vectors
• optimized column widths

inteloide

Hello Arilius !

Your plugin is very useful for me thanks !

Just missing an "update" button, in order to avoid closing an openning the window to refresh and it will be perfect !
Thank you !
Inteloide

jiminy-billy-bob

@inteloide said:

Just missing an "update" button, in order to avoid closing an openning the window to refresh and it will be perfect !

+1

jiminy-billy-bob

I found a conflict with TIG's 2DTools.

This plugin adds an attribute dictionnary to the model called "2Dtools", with a key "z" and a value of "0.0" in my case.
This causes the AttributeInspector to not load anything for the model, and the ruby console outputs this:

AE;;AttributeInspector;;Dialog Error for callback 'get_all'; 
comparison of Length with true failed
C;/Users/Marjorie/AppData/Roaming/SketchUp/SketchUp 2014/SketchUp/Plugins/ae_AttributeInspector/AttributeInspector.rb;505;in `=='
C;/Users/Marjorie/AppData/Roaming/SketchUp/SketchUp 2014/SketchUp/Plugins/ae_AttributeInspector/AttributeInspector.rb;505;in `block (2 levels) in get_attributes'
Blablabla...

(I was able to find the conflicting dictionnary using Eneroth's Attribute Editor)

Dan Rathbun

That is the Trimble SketchUpAPI team's fault. They did not properly override the == method in the Length subclass.

It (the overridden method,) should have an unless clause that invokes super if the argument is not a Numeric subclass.

like:


  def ==(arg)
    unless arg.is_a?(Numeric)
      super
    else
      # check if arg is within SketchUp's
      # tolerance of this Length instance
    end
  end

TIG

So any plugin which saves an attribute as a Length will return an error if the test on it is not sufficiently robust ?
So make a test like:
unless object.get_attribute(key, value, nil).class == NilClass ... else next ### end
So now if the key's value is NOT nil you do something...
BUT if it is nil you skip it ###...

But as I read this plugin's code its line #505 uses:
type = (value == true || value == false) ? "Boolean" : value.class.to_s
That is where the failure happens.
The Length is checked for true/false.
If it's recast as something more like:
type = value.class.to_s; type = "Boolean" if type =~ /TrueClass|FalseClass/
Then it'll work with Lengths too...
You could also add a trap for NilClass and perhaps skip that key/value pair ?
; type = "Nil" if type =~ /NilClass/
OR take it as equivalent to FalseClass ??

Aerilius

Thanks for pointing this out!

I didn't think about that I could get an error when comparing things. I assumed something that doesn't compare would return "false". The ruby way would maybe be to do feature testing:
type = (value.is_a?(Comparable) && (value == true || value == false) ? "Boolean" : value.class.to_s
Which also fails due to SketchUp's bugged API method (Length is a Comparable).
This should be better (and without string comparison). Or shouln't ruby have a "Boolean" super class of which TrueClass and FalseClass are subclasses?
type = (value.is_a?(TrueClass) || value.is_a?(FalseClass)) ? "Boolean" : value.class.to_s

@unknownuser said:

You could also add a trap for NilClass and perhaps skip that key/value pair ?

I wanted to list all existing attributes. An attribute can persist with a value "nil" without being deleted/garbage collected.

Dan Rathbun

Every object, of every class inherits a .nil?() instance method (from Object.)
So:

val = thing.get_attribute(key, value, nil)
unless val.nil?
  # do something
else
  # recover gracefully
end

Dan Rathbun

BTW... you can currently use the triple equal method with Length class, so this will work:
true === length_value || false === length_value

Dan Rathbun

@aerilius said:

Or shouln't ruby have a "Boolean" super class of which TrueClass and FalseClass are subclasses?
type = (value.is_a?(TrueClass) || value.is_a?(FalseClass)) ? "Boolean" : value.class.to_s

I always wondered why not ?

One thing I have done in the past is create a boolean?() query method for Object (or something else like Comparable.
Now this is frowned upon, and I never released it, just suggested, or had it in my test scripts.


def boolean?()
  is_a?(TrueClass) || is_a?(FalseClass)
end unless method(;boolean?)

AND / OR, ... create a Boolean mixin module, and then mix it into only TrueClass and FalseClass. The mixin module may not actually need any additional functionality, at all ?

Then the kind_of? (alias is_a?,) methods will return the appropriate true or false using a Boolean class argument, for any, and ALL objects.
The === will also work for all objects, which makes case statements work:
Boolean === value

Whichever one (or both,) ya'll want, would need to be adopted as an API extension, by the community. (And we'd need to convince the Trimble team to add it.)

tt_su

The Length class is a bit of a special case. It's trying to be a sub-class of Float - but the Ruby API doesn't really let you do this. So there's some hackery in the background to make it work. I'll file an issue to see if we can avoid the argument error.

Dan Rathbun

Thanks TT.

Just returning false would be preferable, instead of raising an ArgumentError (when really TypeError occurs.)

tt_su

Yea, there's a few annoying inconsistency where the correct error isn't raised.