pandas handling of integers and NaNs

pandas converts integer columns with nan values to float64 to accommodate those values.

If the results of a SQL query that contains, say, min() or max() operations, and count() operations, are read into pandas, there is a good chance they will be read in as different data types therefore.

• min() or max() (or many other operations) can result in SQL null values. These will be nan values when the data is read into pandas, meaning the column will be held as a float64
• count() never returns NULL, meaning the column will be held as an integer (int64)