fixes issues with super->sub-class auto-cast and handles MultiInputObj coercion

[tclose]

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Summary

• fixes bugs with super->sub-class auto-casting, including when the sub-classes are part of a union. If the super class is a super class of any of the union args then it passes the check.
• adds handling for MultiInputObj
• streamlines typing error messages

Checklist

• I have added tests to cover my changes (if necessary)

[codecov]

Codecov Report

Attention: Patch coverage is 97.76536% with 4 lines in your changes are missing coverage. Please review.

Project coverage is 88.01%. Comparing base (7d84fba) to head (4c00389).
Report is 9 commits behind head on master.

❗ Current head 4c00389 differs from pull request most recent head ad28ae5

Please upload reports for the commit ad28ae5 to get more accurate results.

Files	Patch %	Lines
pydra/utils/typing.py	93.44%	4 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           master     #746       +/-   ##
===========================================
+ Coverage   15.14%   88.01%   +72.87%     
===========================================
  Files          53       53               
  Lines       15693    15817      +124     
  Branches     2788     1610     -1178     
===========================================
+ Hits         2376    13922    +11546     
+ Misses      12988     1893    -11095     
+ Partials      329        2      -327     

Flag	Coverage Δ
unittests	88.01% <97.76%> (+72.91%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[tclose]

NB: the reference hash updates required in the unittests seem be related to changes to the file-hashing process. This could be due to a change in fileformats but I can't remember making any changes that would be related.

[tclose]

@effigies I have parametrized a number of test batches, there a couple more that could probably be done but would require a bit more work so I thought I could leave them for now.

@djarecka if you are busy are you happy to merge if @effigies has been through it? It just cleans up some corner cases that were missed in the original PR

[effigies]

It looks like coerce() and check_type() have a lot of duplicated code. Does not need to be addressed now, but this is inviting desynchronization and fixing bugs in one and not the other.

Otherwise LGTM.

[effigies]

Another re-ordering of branching.

Suggested change

	except TypeError as e:
	# Special handling for MultiInputObjects (which are annoying)
	if isinstance(self.pattern, tuple) and self.pattern[0] == MultiInputObj:
	# Attempt to coerce the object into arg type of the MultiInputObj first,
	# and if that fails, try to coerce it into a list of the arg type
	inner_type_parser = copy(self)
	inner_type_parser.pattern = self.pattern[1][0]
	try:
	return [inner_type_parser.coerce(object_)]
	except TypeError:
	add_exc_note(
	e,
	"Also failed to coerce to the arg-type of the MultiInputObj "
	f"({self.pattern[1][0]})",
	)
	raise e
	else:
	raise e
	except TypeError as e:
	# Special handling for MultiInputObjects (which are annoying)
	if not isinstance(self.pattern, tuple) or self.pattern[0] != MultiInputObj:
	raise e
	# Attempt to coerce the object into arg type of the MultiInputObj first,
	# and if that fails, try to coerce it into a list of the arg type
	inner_type_parser = copy(self)
	inner_type_parser.pattern = self.pattern[1][0]
	try:
	return [inner_type_parser.coerce(object_)]
	except TypeError:
	add_exc_note(
	e,
	"Also failed to coerce to the arg-type of the MultiInputObj "
	f"({self.pattern[1][0]})",
	)
	raise e