it is worth pointing out that octave is also not the most consistent in accepting/rejecting logicals for numeric inputs. e.g.,

>> interp1(logical(eye(5)),1:5)
ans =

   1   0   0   0   0
   0   1   0   0   0
   0   0   1   0   0
   0   0   0   1   0
   0   0   0   0   1

>> interp3(logical(cat(3, eye(4), eye(4))))
error: Invalid call to interp3.  Correct usage is:
...

interpn behaves similarly to interp3. interpft also accepts logical inputs. (well, it calls fft(x) which does any x validation, and that likely internally converts to double).

pushed https://hg.savannah.gnu.org/hgweb/octave/rev/3606d14d6317
to stable after verifying it passes a make check. it updates error messages, docstring, and BISTs in interp2, interp3, and interpn to be a bit more consistent. no intended behavior change.  I suspect interp1 input validation is deliberately minimal to reduce overhead as it's likely called rather often by other functions. no real reason to change that unless it becomes an issue.

marking as Ready for Test.

comment #6:

Ok. I manual convert logical to double.

regarding your comment #4 - yes, + double(0) does a type conversion to double.  you can do this on any input before passing to interp and it will interpolate as floating point numbers.

So, to quick fix this bug, simply add a double(0) is okay, like this:

>> result = [true false; false true]

result =

  1  0
  0  1

>> result + double(0)

ans =

   1   0
   0   1

yes, recommend changing the report summary to focus on error message / input validation.  stumbled on some other things too:

interp2(1:10)
error: index (1): out of bound 0 (dimensions are 0x0)
error: called from
    interp2 at line 232 column 11

should be a handled error that size must be >=2 in each interpolated dimension.  there may be others.  the docstring could also better articulate function input type/size requirements

For example, imread() a pbm picture will generate a logical 2-D matrix.
                                                                    

>> input_pbm=imread('input_pbm.pbm');
>> size(input_pbm)

ans =

   270   480

When interp this picture, the real problem is, logical data cannot auto convert to double data, and interp2() function only take double matrix, which is not "error: interp2: Z must be a 2-D matrix".

>> interp2(input_pbm, 2, 'nearest');

error: interp2: Z must be a 2-D matrix
error: called from
    interp2 at line 134 column 5

>> interp2(double(input_pbm), 2, 'nearest');
>>