plot(df) and plot_correlation(df) fail when data has 'list' columns

[jnwang]

When running plot(df) and plot_correlation(df) on the following dataframe, since the author column is a list, both plot and plot_correlation failed.

For plot(), the reported error is TypeError: unhashable type: 'list'

For plot_correlation(df), the reported error AssertionError: No numerical columns found

[brandonlockhart]

This has been fixed for plot()

[Waterpine]

@jnwang I think the reason is that plot_correlation() function only supports for numerical data. It is impossible for us to calculate a correlation value between list or categorical data. Thus, I believe that we could solve this problem in the next version. Thanks!

[jnwang]

Can you please check how df.corr() handles this? It would be better if we can make plot_correlation() consistent with df.corr()?

[Waterpine]

@jnwang @dovahcrow df.corr() also does not handles this. There is nothing output when we use the df.corr() function.

[jnwang]

It looks like that df.corr() automatically ignores non-numerical columns. Can we do the same rather than raise an error?

https://www.geeksforgeeks.org/python-pandas-dataframe-corr/amp/

[Waterpine]

@jnwang @dovahcrow @jinglinpeng I have fixed this bug. The plot_correlation() function only supports for numerical data. Plot_correlation() will ignore non-numerical data and output nothing when the input data is not numerical. If you have any questions, please let me know as soon as possible. Thanks! #57

[dovahcrow]

@jnwang why empty output is good? I feel like a hard error is more explicit than an implicit empty return, and explicit is better than implicit.

[jnwang]

We should output an empty plot if all cols are non-numerical. I think we did it for plot_correlation(k).

[Waterpine]

@jnwang @dovahcrow I agree with young. However, an empty plot is also a good option.

[dovahcrow]

Let do empty and can make it error later if users are complaining.

[Waterpine]

Okay, I have fixed this bug. Thanks! #57