SlabsIter has weak error messages

[andeaseme]

SlabsIter will throw errors related to code deep in its own functions without mentioning what inputs could be wrong. Debugging is hard for new users as shown by the following example.

I've written some code and expect it to print the contents

def item_generator():
    for i in range(3):
        yield {'index': i}

def print_item(item):
    print(item)

with SlabsIter(item=item_generator(), print_item=print_item) as sit:
    for _ in sit:
        pass

But instead I'm met with an error

Traceback (most recent call last):
  File "/Users/Andie.Phan/Projects/otosense/plunk/plunk/ap/pain/context_func.py", line 105, in <module>
    _weak_error_message()
  File "/Users/Andie.Phan/Projects/otosense/plunk/plunk/ap/pain/context_func.py", line 96, in _weak_error_message
    for _ in sit:
  File "/Users/Andie.Phan/Projects/otosense/know/know/base.py", line 341, in __iter__
    yield next(self)
  File "/Users/Andie.Phan/Projects/otosense/know/know/base.py", line 334, in __next__
    return self._call_on_scope(scope={})
  File "/Users/Andie.Phan/Projects/otosense/know/know/base.py", line 326, in _call_on_scope
    scope[name] = _call_from_dict(scope, component, self.sigs[name])
  File "/Users/Andie.Phan/Projects/otosense/know/know/base.py", line 164, in _call_from_dict
    args, kwargs = sig.args_and_kwargs_from_kwargs(
AttributeError: 'NoneType' object has no attribute 'args_and_kwargs_from_kwargs'

The problem has something to do with None object but it doesn't tell me what's wrong or where to look. There is really no hints at all.

[andeaseme]

Not a solution to the error message issue.

The way to fix the code in the example is to wrap item_generator.

with SlabsIter(item=iter(item_generator()).__next__, print_item=print_item) as sit:
    for _ in sit:
        pass

[thorwhalen]

I added a validation step in the SlabIter init so that you'll now get the precise error message

TypeError: This component is not callable: item

before you go down the layers to fail in some obscure place that has little resemblance to your context.

There was, in fact, a simple assertion in there already, but for some reason it was commented out. I hope it wasn't a good reason, because now we have validation again.

But I profited of the occasion to do a few things:

• write this issue about the general problem of getting obscure messages in significantly layered architectures like ours (or even just in well concern-separated code of any size)
• here it made more sense to do validation anyway, so instead of just uncommenting my older assert, I set up a bit "stronger" validation for you so you can expand on it.