Add functionality to compare DataFrames #34
Conversation
engarde/checks.py
Outdated
| __all__ = ['is_monotonic', 'is_shape', 'none_missing', 'unique_index', 'within_n_std', | ||
| 'within_range', 'within_set', 'has_dtypes', | ||
| 'verify', 'verify_all', 'verify_any'] | ||
| def is_same_as(df, df_to_compare): |
There was a problem hiding this comment.
Can you accept **kwargs here and pass them through to the tm.assert_frame_equal? That can be useful for relaxing the equality requirements.
There was a problem hiding this comment.
How should I acknowledge kwargs in the docstring?
There was a problem hiding this comment.
Something like
**kwargs : dict
keyword arguments passed through to ``assert_frame_equal``
engarde/checks.py
Outdated
|
|
||
| """ | ||
| try: | ||
| tm.assert_frame_equal(df, df_to_compare) |
There was a problem hiding this comment.
You'll have to pass through **kwargs here to tm.assert_frame_equal :) This lets you do things like
df.pipe(is_same_as, other, check_names=False)
engarde/checks.py
Outdated
| try: | ||
| tm.assert_frame_equal(df, df_to_compare) | ||
| except AssertionError as e: | ||
| raise AssertionError("DataFrames are not equal") |
There was a problem hiding this comment.
It'd be nice to include the original assertion error for why they aren't equal here. I know in python3 it's just
raise AseertionError("DataFrames are not equal") from ebut I don't know about py2. I think leave it for now.
Are you using python3 or 2? I'm tempted to just make the next release python3 only.
engarde/decorators.py
Outdated
| def is_same_as(df_to_compare, **kwargs): | ||
| def decorate(func): | ||
| @wraps(func) | ||
| def wrapper(*args, **kwargs): |
There was a problem hiding this comment.
Ugh, this is a bit complicated. Can you rename the outer is_same_as **kwargs to **assert_kwargs and then pass those to ck.is_same_as on line 170? Then the wrapper **kwargs are for whatever func is being decorated (like you have now).
|
I got the other errors fixed with tests around **kwargs functionality to ensure that less strict DataFrames match (float vs int). I dug into re-raising exceptions and found this post on StackOverflow. Reading thru the referenced blogs (1, 2), we can provide better tracebacks using a three-argument raise statement for python2. Personally, I'm using Python3 (#Teamfstrings). I do not have much experience in developing for both. Would it be difficult to implement two different exception re-raising workflows? How do other libraries handle this? There are definitely good reasons to keep testing libraries like this compatible with both py2 and py3. I can help out if you decide to go this route. For now, I amended my last commit to fix **kwargs usage, added better tests around that functionality, and kept the |
|
This looks good. I'll figure out the py2 vs. py3 stuff later. Hopefully will do a release in a couple weeks. |
|
I'm listening to Hynek's talk from PyCon while I'm working. He mentioned the This should fix all exception chaining issues. I'll update my code when I get some time this weekend. |
Added functionality to compare DataFrames by leveraging
pandas.util.testingassert helpers