ErrorTree must be return full path to error object

[python-jsonschema]

ErrorTree must be return full path to error object

[Julian]

Hi.

Not sure I'm understanding any of this, sorry.

ErrorTrees are essentially just nested dicts. The error tree in your example is essentially equivalent to the dict {"a" : {"b" : {"c" : ["spam"]}}}, but it carries along some extra information at each node, like possible errors at a, b, and c themselves, or keeping references to actual error objects. I see you have some funky behavior there iterating over the tree, which tells me you're probably on 0.7. Specifically, seeing a 0 there is totally wrong, the 0 is in tree["A"]["B"]["C"]. This should be fixed in the most recent release (v0.8.0). You should see:

>>> for x in tree:
...     print(x)
... 
A

This is the same as you'd get if you iterated over {"a" : {"b" : {"c" : ["spam"]}}} – it's telling you which keys the dict has, or in the case of ErrorTree, which children it has (which means which children of your instance had errors, in this case all of them are in the A child). If you want to dig further, you access that child and ask it for the same thing (and can do this programmatically certainly).

If you had a nested container in Python (a tree) and you wanted to iterate over all of its children, you'd do so. ErrorTrees are these kind of objects, so you iterate over it as you would any other tree (or nested dict), i.e. via recursion or with a stack or whatever. The simplest example being, say:

>>> def walk(tree):
...     errors = tree.errors
...     if errors:
...         yield errors
...     for child in tree:
...         for errors in walk(tree[child]):
...             yield errors
...         
...     
... 
>>> list(walk(tree))
[{'minItems': ValidationError(u"['spam'] is too short", None, ())}, {'type': ValidationError(u"'spam' is not of type 'number'", None, ())}]

If you wanted to keep track of the path at each recursion you could certainly modify that to do so, however if you really wanted that, as it sounds like you do, ValidationErrors already know their path and keep it in a way like it sounds like you want, so you would just iterate over the errors from iter_errors and pull off each of the path attributes and do what you'd like with them.

Let me know if this helps (I definitely can't tell what you want with #53, but if you have more ideas I'd love to hear them, just narrowing down on where exactly you're having trouble).

[ertyz]

print(jsonschema.__version__)
for x in tree:
    print(x)

0.8.0
0
A

It strange. I try to use your code, but list(walk(tree)) return me []

Maybe the reason in my python version (CPython 3.3 for Win64)

[Julian]

Interesting.

I don't have a Windows machine, but I'll try this out on 3.3 when I get home just to see what's up.

[ertyz]

I definitely can't tell what you want with #53

In my project I have a simple task:

1. load large json datagram
2. check values in loaded json
3. if value is not valid for my terms -> then load default(empty) values
   but i need to save all other valid values in loaded json data
   Now I implement it by many blocks if-then in my code. And try to rewrite project to use jsonschema for this task.

[Julian]

Cool. So do you have something specific that'd make that easier in mind?

I'd do that the same way I showed here, do a depth first search of the tree and yield elements that don't have errors, only this time yield them from the instance, not the error tree.

[Julian]

I can't reproduce either the behavior you described for walk or the presence of 0 on 3.3. Can you pastebin a full interpreter session showing exactly what you're running perhaps?

[ertyz]

damn. it's my fault. I didn't clear test code of print(tree[0].errors). This procedure create 0, when was invoked
Thank you for helping and great work!

[Julian]

Are we good to close this and #53 or was there still something you wanted to propose?

[Julian]

Closing, feel free to open a new ticket or reopen if you have something in mind.