Can't Convert 'Nothing' to 'Missing'

General Usage
1

There has been a bit of discussion on here about ‘nothing’ vs ‘missing’ and why both are needed. I don’t really have an opinion on this as long as I can easily convert, which I can’t.

I tried the suggestion on Github [missing vs nothing #1854] which was marked as the solution:

replace( df.A , nothing => missing)

and it didn’t complain but didn’t work (‘nothings’ still there).

I wondered why the ! wasn’t needed and also why I didn’t need to broadcast, but attempting either of those led to an error.

Does anyone have any ideas?

Thanks

2

replace sans ! returns a copy of the collection with the given replacement, but doesn’t modify in place. To modify in place, use replace!. So if you do e.g.

using DataFrames
df = DataFrame(x = [1, nothing, 3])
replace(df.x, nothing => missing)

Then the last line will just return [1, missing, 3] without modifying your DataFrame. Note that replacing nothing with missing will tend to require a new array type, and so replace! will throw an error. (E.g. an array with element type Union{Int64, Nothing} can’t hold values of type Missing.) Therefore, you’ll probably want to change dataframe columns by df.A = replace(df.A, nothing => missing).

3 Likes
3

Hi Thanks, That is what I thought. But when I added the ! it threw an error:

julia>replace(pt[:Age],nothing=>missing);

julia> replace!(pt[:Age],nothing=>missing);
Warning: getindex(df::DataFrame, col_ind::ColumnIndex) is deprecated, use df[!, col_ind] instead.
caller = top-level scope at REPL[722]:1
@ Core REPL[722]:1
ERROR: MethodError: Cannot convert an object of type Missing to an object of type Int64

Thanks again
PS I edited out the first Warning about using pt[:Age] which is not relevant.

4

The array element type cannot accommodate missing. You can use replace, replacing the entire column, eg

df.col = replace(df.col , nothing => missing)
2 Likes
5

I see. Great. Thanks [worked]