Pandas version checks

  • [X] I have checked that the issue still exists on the latest versions of the docs on main here

Location of the documentation

https://pandas.pydata.org/docs/user_guide/basics.html#dtypes

Documentation problem

Upcasting is discussed, but not downcasting.

Suggested fix for documentation

It is fine not to implement a warning when downcasting will cause data loss, but the fact that this is not implemented means the user should be advised that they assume the risk.

In short, the behavior when downcasting a column should be discussed antonymously to the upcasting section in the documentation.

Resulting from https://github.com/pandas-dev/pandas/issues/50214.

Comment From: MarcoGorelli

hi - what do you mean by "downcasting" a column?

could you please provide a simple example, like: - here's what I ran - here's what happened - here's what I expected

Comment From: mcp292

Please see the issue linked in the original post and let me know if that's what you're looking for.

Comment From: MarcoGorelli

Which downcasting are you referring to?

From https://github.com/pandas-dev/pandas/issues/50125#issuecomment-1342886489 , as far as I can tell, it's just what numpy does

In [28]: arr[2]
Out[28]: array([11.11], dtype=float16)

In [29]: arr[2].item()
Out[29]: 11.109375

And regarding the inequality, this reproduces in pure numpy, so isn't an issue in pandas

In [30]: arr = np.array(data, dtype=np.float16)

In [31]: arr[arr <= 99.99]
Out[31]:
array([  0.   ,  10.25 ,  11.11 ,   5.555,  50.56 ,  70.75 ,  99.   ,
       100.   , 100.   , 100.   ], dtype=float16)

In [32]: arr
Out[32]:
array([  0.   ,  10.25 ,  11.11 ,   5.555,  50.56 ,  70.75 ,  99.   ,
       100.   , 100.   , 100.   ], dtype=float16)

Comment From: MarcoGorelli

And regarding the inequality, this reproduces in pure numpy, so isn't an issue in pandas

As per the above, pandas did not do any downcasting here

Closing then, but thanks for the report