497

[kennethshsu]

• Fixed a calculation bug in onlevel factors not matching CAS example #497
• Introduced approximation_grain in chainladder.ParallelogramOLF()
• Added two test cases as discussed in onlevel factors not matching CAS example #497

[kennethshsu]

@jbogaardt, are you familiar with these errors? I looked at it in detail and I don't think I understand what is going on...

=========================== short test summary info ============================
ERROR chainladder/methods/tests/test_predict.py - ValueError: Fancy indexing on origin/development is not supported.
ERROR chainladder/workflow/tests/test_predict.py - ValueError: Fancy indexing on origin/development is not supported.
!!!!!!!!!!!!!!!!!!! Interrupted: 2 errors during collection !!!!!!!!!!!!!!!!!!!!

Is this line causing the error? Why?

[jbogaardt]

I'm not sure the specifics on the error from this PR, but the error stems from non-contiguous slices out of origin/development.

Here is a bare-bones example that is not supported by the package where we slice an arbitrary list of origins that are out of sequence:

import chainladder as cl
clrd = cl.load_sample('clrd')
clrd.iloc[:, :, [2, 4, 3]]

[kennethshsu]

Hmm... What's weird is that these tests are passing locally... Is there an issue with the way that tests are set up on GitHub?

chainladder/methods/tests/test_predict.py:

% pytest test_predict.py
================================================ test session starts ================================================
platform darwin -- Python 3.11.3, pytest-7.3.1, pluggy-1.0.0
rootdir: /Users/kennethhsu/Documents/GitHub/chainladder-python
plugins: xdist-2.5.0, cov-4.0.0, forked-1.3.0, anyio-3.7.0
collected 13 items                                                                                                  

test_predict.py .............                                                                                 [100%]

================================================= warnings summary ==================================================
chainladder/methods/tests/test_predict.py::test_misaligned_index[normal_run]
chainladder/methods/tests/test_predict.py::test_misaligned_index[normal_run]
  /Users/kennethhsu/Documents/GitHub/chainladder-python/chainladder/core/base.py:250: UserWarning: The argument 'infer_datetime_format' is deprecated and will be removed in a future version. A strict version of it is now the default, see https://pandas.pydata.org/pdeps/0004-consistent-to-datetime-parsing.html. You can safely remove this argument.
    arr = dict(zip(datetime_arg, pd.to_datetime(**item)))

chainladder/methods/tests/test_predict.py: 17 warnings
  /Users/kennethhsu/opt/anaconda3/envs/cl_dev/lib/python3.11/site-packages/sparse/_coo/core.py:261: DeprecationWarning: coords should be an ndarray. This will raise a ValueError in the future.
    warnings.warn(

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
========================================== 13 passed, 19 warnings in 9.69s ==========================================

Same thing with chainladder/workflow/tests/test_predict.py. Passing locally.

% pytest test_predict.py
================================================ test session starts ================================================
platform darwin -- Python 3.11.3, pytest-7.3.1, pluggy-1.0.0
rootdir: /Users/kennethhsu/Documents/GitHub/chainladder-python
plugins: xdist-2.5.0, cov-4.0.0, forked-1.3.0, anyio-3.7.0
collected 1 item                                                                                                    

test_predict.py .                                                                                             [100%]

================================================= 1 passed in 0.18s =================================================

[jbogaardt]

It may be differences in dependencies. Github grabs the latest packages. It is using pandas==2.2.1 .

My suspicion is there are some differences in these versions that is causing it to fail. In issue #494, we have this problem with pandas>2.2.0. I think you just made the same issue visible in our unit tests. I've committed a fix since the pandas team gave a suggestion and also signaled they are unlikely to change the behavior. It seemed to clear those errors, but now has more pointed suggestions for the on-leveling updates.

FAILED chainladder/utils/tests/test_utilities.py::test_non_vertical_line - assert 0.1746221283369899 < 0.0
FAILED chainladder/utils/tests/test_utilities.py::test_vertical_line - UnboundLocalError: cannot access local variable 'cum_avg_rate_non_leaps' where it is not associated with a value

[kennethshsu]

@jbogaardt how did you end up fixing the remaining test errors?

=========================== short test summary info ============================
FAILED chainladder/core/tests/test_display.py::test_empty_triangle - AttributeError: 'Triangle' object has no attribute 'values'
FAILED chainladder/core/tests/test_display.py::test_to_frame[normal_run] - ValueError: The truth value of a DataFrame is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().
FAILED chainladder/core/tests/test_display.py::test_to_frame[sparse_only_run] - ValueError: The truth value of a DataFrame is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().
===== 3 failed, 603 passed, 11 xfailed, 6055 warnings in 283.53s (0:04:43) =====

I was going to take a look at those. Or are those not resolved yet?

[jbogaardt]

they aren't resolved, but they are display so I assumed they were related to package dependencies and not the PR itself. Looking closer, I see you have he following in display.py:

        try:
            self.values
        except:
            print("Triangle is empty")
            return

I suspect this is causing it. Can you let me know your intent with this logic?

[kennethshsu]

Ahh, I think that's from another branch to address #491, it's not quite done... I think I messed up my version control while working on the two branches. Let me try to resolve this.