PrefixList validation tracebacks are chained

[mdickinson]

Validation exceptions from assignment to a PrefixList trait have some unnecessary chaining; this is going to clutter logs and consoles. It would be good to get rid of this.

This is probably not specific to PrefixList.

>>> from traits.api import HasTraits, PrefixList
>>> class A(HasTraits):
...     foo = PrefixList(["yes", "no"])
... 
>>> a = A()
>>> a.foo = "bogus"
Traceback (most recent call last):
  File "/Users/mdickinson/Enthought/ETS/traits/traits/trait_type.py", line 355, in _value_for
    return self.value_for(value)
  File "/Users/mdickinson/Enthought/ETS/traits/traits/trait_types.py", line 2629, in value_for
    raise TraitError(
traits.trait_errors.TraitError: The value of a PrefixList trait must be 'yes' or 'no' (or any unique prefix), but a value of 'bogus' <class 'str'> was specified.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/mdickinson/Enthought/ETS/traits/traits/trait_type.py", line 357, in _value_for
    self.error(object, name, value)
  File "/Users/mdickinson/Enthought/ETS/traits/traits/base_trait_handler.py", line 74, in error
    raise TraitError(
traits.trait_errors.TraitError: The 'foo' trait of an A instance must be 'yes' or 'no' (or any unique prefix), but a value of 'bogus' <class 'str'> was specified.

Assigning to the 6.1 milestone to see if there's anything we can reasonably do about this before the release. If not, it can be bumped.

[kitchoi]

One way to fix is to do raise ... from None in

traits/traits/base_trait_handler.py

Lines 57 to 76 in 94a3dca

	def error(self, object, name, value):
	"""Raises a TraitError exception.
	This method is called by the validate() method when an assigned value
	is not valid. Raising a TraitError exception either notifies the user
	of the problem, or, in the case of compound traits, provides a chance
	for another trait handler to handle to validate the value.
	Parameters
	----------
	object : object
	The object whose attribute is being assigned.
	name : str
	The name of the attribute being assigned.
	value : object
	The proposed new value for the attribute.
	"""
	raise TraitError(
	object, name, self.full_info(object, name, value), value
	)

But this is perhaps too general and removed from the context where the exception came from. There are probably situations where the chained traceback is helpful. We could add a flag in BaseTraitHandler.error to optionally raise ... from None I suppose. Does this worth the extra code branch?

[mdickinson]

Bumping to 6.2.0

[mdickinson]

"Fixed" in #1563 for PrefixList and PrefixMap, by replacing their use of value_for with plain old validate.

More generally, any TraitType subclass that uses the value_for method for its validation will have this issue. PrefixList and PrefixMap are the only TraitType subclasses in the current main branch that use value_for. Given that the symptoms are mild and that the prevalence of the issue seems tiny, it doesn't seem worth fixing this more generally.

Relevant code:

• traits/traits/trait_type.py

  Lines 155 to 161 in 0e74444

  	``value_for ( self, value )``
  	As another alternative to implementing the ``validate`` method, you
  	can instead implement the ``value_for`` method, which receives only
  	the ``value`` being assigned. It should return the validated form of
  	``value`` if it is valid, or raise a ``TraitError`` if the value is not
  	valid.

• traits/traits/trait_type.py

  Lines 402 to 404 in 0e74444

  	validate = getattr(self, "value_for", None)
  	if validate is not None:
  	validate = self._value_for

[mdickinson]

This was a duplicate of #955, and was fixed in #1564.