Introduce shaped arg to execute_mdx_dataframe

[MariusWirtz]

Inspired by
Kevin-Dekker@8ab9901#diff-5242c6c19161de7f529a4e5b0d0b33ca9c15f69be65fa5409881b3d8310ff655R2799-R2802

execute_mdx_dataframe, execute_view_datafame can be run with shaped=True to retrieve a similar but not fully equivalent result to execute_mdx_dataframe_shaped.

This allows retrieving a shaped data frame while specifying options that are not available in execute_mdx_dataframe_shaped. One example is the usage of use_iterative_json=True and use_blob=True which optimizes memory usage during string to json / dict conversion prior to the creation of the data frame.

Fixes #888

[Cubewise-gtejeda]

The memory footprint was much lower when trying this, confirming previous tests.
Not sure if this is just a formatting issue, but when printing out the dataframe the formats of the numbers look different:

use_iterative_json=False:

use_iterative_json=True:

The shape and order seem to be identical, which is also expected.
However, there is an additional string appearing when using use_iterative_json=True:

For existing queries, I am worried that this might break backward compatibility. Could there be an option to suppress that string?

[MariusWirtz]

Thanks for the review! Good catch on the index column. I will make sure we get rid of this incorrect column label.

Regarding the formatting, this is kinda expected. The data frame is built differently and now the type is inherited. In the old execute_view_dataframe_shaped function the type would always be a string. The new approach is more intelligent IMO.

The scientific notation is just a pandas thing.
You should be able to switch to normal numbers like this

import pandas as pd
pd.options.display.float_format = '{:.5f}'.format

[Cubewise-gtejeda]

Thanks for the update.
The datatype inheritance - although a good thing - might break compatibility with existing reports in Power BI, since users currently would change the datatype in the query. I am not sure if this would cause an error in the query, but cannot test this in the connector until the release.

Could this be a parameter as well?

Thoughts?

[MariusWirtz]

The datatype inheritance - although a good thing - might break compatibility with existing reports in Power BI, since users currently would change the datatype in the query. I am not sure if this would cause an error in the query, but cannot test this in the connector until the release.

It wouldn't break any existing applications, because it only applies when use_blob=True or use_iterative_json=True is passed.
I think we are on track to release to PyPI this week.

I can't reproduce the strange column header. Please post the code that produces the data frame.

[MariusWirtz]

The unwanted column header has been removed.