All,
I have the following observation with gmtselect, using either an area (-R) or polygon (-F) on a file which has x y and a third column which can be mixed as a text string or number. The outcome is that sometimes the text strings are printed as NaN, and this seems to depend on the value in the first line.

The example underneath illustrates this issue. The only difference between the two tests is that I exchanged the first 2 lines, so the results should be identical. The result of test 1 gives me NaN instead of A and C, the result of test 2 is correct.

echo “Test 1”
echo “.5 .5 1” > temp
echo “.25 .25 A” >> temp
echo “1.7 .7 B” >> temp
echo “.7 .7 C” >> temp
gmtselect -R0/1/0/1 temp
Result:
0.5 0.5 1
0.25 0.25 NaN
0.7 0.7 NaN

echo “Test 2”
echo “.25 .25 A” > temp
echo “.5 .5 1” >> temp
echo “1.7 .7 B” >> temp
echo “.7 .7 C” >> temp
gmtselect -R0/1/0/1 temp
Result:
0.25 0.25 A
0.5 0.5 1
0.7 0.7 C

The bug here may be that the t in -i0-1,2t is being ignored.

• Operating system: linux
• GMT version: 6.0 and 6.2

(note: I am wondering if a similar behaviour may appear with grdtrack where one column of the input file may also have a mix)