Feedback on using ltk.Model

[Neon22]

MVP Feedback:

In trying to migrate to using the MVP (ltk.Model class) I have come across some issues.
I will lay them out here in the hopes that it might help in future development.

• Using ltk 0.2.20 and [email protected] with pyscript 2024.11.1 and pyodide

1. To get around the potential memory leak in widget() I have successfully been using a global var. The MVP approach looks like it may supersede the need for the recently added widget()

2. I am also using a global var to make easy ref to the following reactive component. This is working well.
   I am also using a global to deal with rippling updates, if I update two or more reactive variables. This is working well.

I.e.

g_RC_values = None  # Hold reactive Model
g_inhibit_update = False  # prevent rippling updates

class Reactive_component(ltk.Model):
    """
    These sets of triplet variables update each other. Doing unit conversions
    """
    #  Length
    Len_measure: str = "250yd"
    Len_units: str = "yd"
    Len_altlabel: str = "229m"
    #
    Sk_wraps: str = "20wpi"
    Sk_units: str = "wpi"
    Sk_altlabel: str = "7.9wpcm"

    def changed(self, name, value):
        """
        Called whenever a UI element is changed by user.
        Also invoked for internal changes
        - g_inhibit_update is used to prevent needless rippling
        """
        global g_inhibit_update
        print("Change requested:",name,value)
        if not g_inhibit_update:
            print("Changing:",name,value)
            g_inhibit_update = True  # inhibit rippling calls while we update
            # Skein Length
            if name in ["Len_measure", "Len_units"]:
                vars = ["Len_measure", "Len_units", "Len_altlabel", ["250","yd","m"]]
                if name == "Len_measure":
                    self.update_primary(value, *vars)
                elif name == "Len_units":
                    self.recalc_units(value, *vars)
            # Skein units
            if name in ["Sk_wraps", "Sk_units"]:
                vars = ["Sk_wraps", "Sk_units", "Sk_altlabel", ["20","wpi","wpcm"]]
                if name == "Sk_wraps":
                    self.update_primary(value, *vars)
                elif name == "Sk_units":
                    self.recalc_units(value, *vars)
            # reset ripple updates
            g_inhibit_update = False
        else:
            print(" -Skipping change")

if __name__ == "__main__":
    g_RC_values = Reactive_component()
    w = Widget(g_RC_values)
    widget = w.create()
    widget.appendTo(ltk.window.document.body)

3. autocreation of instance variables:

• The reason that I am using the names of the variables rather than the variables themselves is because of a side effect of ltk.Model using class variables.
  Specifically:
• I can set a class var just fine with self.Len_measure = "foo"
• but if I pass self.Len_measure through a function header and try to assign it, then an instance variable is created and the class var is not updated.
  E.g.

# in changed()
self.update_myvar(self.Len_measure)

def update_myvar(self, myvar):
   myvar = "Foo"
   # The iv is created and updated but the classvar is unchanged

So to get around this I send the name of the classvar through (as in changed() above and do a more complicated operation like so:

def update_primary(self, value, primary_var, unit_var, alt_var,
                      defaults=["250","yd","m"]):
       """
       Update the primary, units, alts on basis of input.
       value can have units or not. Use as supplied and update alt.
       Use defaults if missing or error
       - Have to use self.__setattr__("varname", value) because its class var !
       """
       num, unit = parse_units(value)
       if num and num > -1:
           # got a valid measure but maybe no units
           if not unit:
               # use on-screen units
               units = self.__dict__[unit_var]
               self.__setattr__(primary_var, f"{format_nice(num)}{units}")
               if units in metric:
                   num = num * convert_factors[units]
               self.__setattr__(alt_var, convert_imp(num, unit_swaps[units], self.Len_units=="in"))
           else:  # unit from measure
               if unit not in convert_factors.keys():
                   unit = defaults[1] #"yd"
               self.__setattr__(unit_var, unit)
               usnum,_ = parse_to_US(value)
               self.__setattr__(primary_var, convert_imp(usnum, unit))
               self.__setattr__(alt_var, convert_imp(usnum, unit_swaps[unit]))
       else:  # not valid so set default
           self.__setattr__(primary_var, defaults[0]+defaults[1])
           self.__setattr__(unit_var, defaults[1])
           self.__setattr__(alt_var, convert_imp(int(defaults[0]), defaults[2]))

In short:
I am using self.__setattr__(primary_var, newvalue) instead of primary_var = newvalue because its a classvar.

So I am ok with my solution and its as neat as I can work out how to make it (by using the expanding *vars parameter).

But I wonder if there is another way to architect it so it works differently.
Working example here:

• https://pyscript.com/@neon22/mvp-testing/latest?files=main.py

Cheers...

[Neon22]

Additional related information:
Can't use += kinds of operators on ModelAttributes

When changing a ModelAttribute we can do this:

class Reactive_component(ltk.Model):
    Drawin = 12

def changed(self, name, value):
   self.Drawin = self.Drawin + 2  # works
   self.Drawin.set_value(self.Drawin.get_value() + 2)  # also works
   # but
   self.Drawin += 2  # fails

Error message is:

Error calling function from JavaScript
  <closure <function set_model_value at 0x542a0> at 542c0, n_closed=2
     <cell 539b0
        <Checkbox object at 539b0>>
    <cell 53290 False> >:
 can't convert NoneType to int

This may all be classvar vs instancevar problems and not so easily solved.

[Neon22]

Also I have been caught out a few times as ModelAttribute has a __str__ function which means it prints as a string which sometimes works and sometimes doesn't if I am not careful when I use it. So been trying to manually add a lot of str() or f"{}" when I mean to get its value. Maybe we could do with an __repr__ but not an __str__ ??

[Neon22]

I have not succeeded in making a component style object using the Model class.
Because its a class variable I have to make a new class for each instance of a component. This is impractical.

A good solution would be to move to instance variables. I wonder if this is at all possible?