Add support for TypedDict about crosshair HOT 11 CLOSED

I see what you mean. I don't think the error is that bad to read, and neither is the SmtObject one. And I also realize that my code is a bit of an "outlier" when it comes to Python usage, I suppose. Much reflection and assumptions about internal attributes. I just wanted to try crosshair and that's what I've been working on.
I guess fixing the remaining error comes down to telling crosshair/python what types I really expect in this function, instead of simply Any. I tried that before, even, but it was pretty difficult.

from crosshair.

I'd at least like to get this exception fixed and have the rest of my code analyzed "in peace".

from crosshair.

Votes help! I'll investigate this one more during the week.

Would love more bugs for other issues you find - thank you for trying it out!

from crosshair.

Thanks for this amazing project. I'll be trying this out some more over the next weeks and give feedback. I'm totally fine with stuff not working, by the way. It's just that I let CrossHair loose on my whole code basically, and it kept on spitting out these error messages so I wan't able to parse what's happening.

from crosshair.

Thanks :)
Added some pretty basic support just now. Would you give it a shot and let me know how it goes? (you can install the HEAD version with pip install git+https://github.com/pschanely/crosshair)

from crosshair.

Okay, I tried this with my code, and it still gives errors. I'm not sure if they are the same errors as before though. The code is already self-contained and reeeelatively small, so I attached it to this comment. Does this help?
python_schema.txt

from crosshair.

Very much helps, thanks! I think there is also something going with forward type references; I see NameError: name 'ConfigAST' is not defined in this one. I think I know what's up & should have a fix in the morning.

from crosshair.

Alright! 3eb1ddb should do it. Give it one more try?

Of course, all this just makes CrossHair able to ingest your module; you'll want to add pre/post conditions for it to do something useful.

For instance, if you wanted to harden your typed_dict_to_schema against a wider variety of possible arguments, you could put a trivial '''post:True''' condition in there to have it find inputs that could break it. (CrossHair does very little with Any typed arguments, but can find some things)

from crosshair.

Okay, I've tried running the code with post: True and get the following error now:

I found an exception while running your function.
python_schema.py:146:
|        return SchemaBool()
|    if t == float:
|        return SchemaFloat()
>    if "__origin__" in t.__dict__:
|        if t.__origin__ == list:
|            return SchemaList(typed_dict_to_schema(t.__args__[0]))
|        if t.__origin__ == Union:

AttributeError: 'object' object has no attribute '__dict__'
when calling typed_dict_to_schema(t = <object object at 0x7efc3089f600>)

I realize that my code is somewhat gnarly, but this error seems interesting nontheless. Shouldn't there always be a __dict__ for Python objects?

Again thanks for trying to get this to work.

from crosshair.

Interestingly, not all python objects have a __dict__!

>>> object().__dict__
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'object' object has no attribute '__dict__'

In particular, I think native classes don't. It's also the case for things like strings and ints.
The repr for object instances isn't very great. I think I'd like to make the CrossHair error here say
when calling typed_dict_to_schema(t = object())
rather than
when calling typed_dict_to_schema(t = <object object at 0x7efc3089f600>)

Just filed #34.

Now, if you work around that issue, you'll likely run into another that is also confusing. A hint I can offer up early is that when CrossHair is emulating a value of a yet-unknown type, that thing will be of type "SmtObject." I'll need to think a little harder about whether I can improve the error there.

from crosshair.

Closing this issue as we've got basic TypedDict support implemented.
But I'd love to hear more about your experience in the form of more bugs, a github discussion, or messages in the gitter chat!

from crosshair.