Replacing values in a table
조회 수: 12 (최근 30일)
이전 댓글 표시
I have a table created from a database query using T = select(conn,selectquery); Missing database entries for numeric integer types are read as -2147483648 for long integer types (int32) and -32768 (int16) for short integer types. I would like to replace these with NaN throughout the entire table. If I use I=ismissing(T,[-2147483648 -32768]) I returns logical array with all the table indices where there are missing values as 1=true, in other words, this works. However, if I try to use T = standardizeMissing(T,[-32768 -2147483648]) or any variation, the values are not replaced with NaN as the command would suggest should happen.
댓글 수: 3
dpb
2018년 4월 11일
OK, I see Peter already identified what was going to be my second thought (that didn't come to me until after the posting, I'll admit)
채택된 답변
Peter Perkins
2018년 4월 11일
Are these actual integer types, and not just doubles with integer values in them? Because you can't put NaNs into int16 and int32. standardizeMissing completely ignores integer types:
>> T = table([1;-32768;3;4],[5;6;-2147483648;8])
T =
4×2 table
Var1 Var2
______ ___________
1 5
-32768 6
3 -2.1475e+09
4 8
>> T = standardizeMissing(T,[-32768 -2147483648])
T =
4×2 table
Var1 Var2
____ ____
1 5
NaN 6
3 NaN
4 8
>> T = table(int16([1;-32768;3;4]),int32([5;6;-2147483648;8]))
T =
4×2 table
Var1 Var2
______ ___________
1 5
-32768 6
3 -2147483648
4 8
>> T = standardizeMissing(T,[-32768 -2147483648])
T =
4×2 table
Var1 Var2
______ ___________
1 5
-32768 6
3 -2147483648
4 8
댓글 수: 2
Peter Perkins
2018년 4월 13일
I'll make a note to tweak the doc.
There is no "work-around" per se because integer types just don't have the notion of a NaN value. The thing to do is to convert to either 1) floating point, maybe single if memory is an issue 2) categorical, if these are really categorical data (categorical has undefined analogous to NaN)
추가 답변 (1개)
David Fletcher
2018년 4월 11일
tab =
10×2 table
C1 C2
__ ___________
1 -32768
2 9
3 3
4 2
5 1
6 -2.1475e+09
7 7
8 9
9 1
10 2
standardizeMissing(tab,{-32768,-2147483648},'DataVariables',{'C2'})
ans =
10×2 table
C1 C2
__ ___
1 NaN
2 9
3 3
4 2
5 1
6 NaN
7 7
8 9
9 1
10 2
댓글 수: 2
David Fletcher
2018년 4월 11일
See the answer below - these weren't expressly cast to ints so would have been doubles. I still have a nasty habit of typing what I assume is an int and forgetting that Matlab defaults to a double type.
참고 항목
카테고리
Help Center 및 File Exchange에서 Data Type Conversion에 대해 자세히 알아보기
제품
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!