Background and foreground colors for categorical CPTs

[seisman]

Description of the problem

As per the documentation:

For categorical data, background color or foreground color do not apply.

The actual behavior is inconsistent with the documentation.

Full script that generated the error

gmt begin map png

cat > types.cpt << EOF
0   yellow   ;desert
1   blue     ;forest
2   red      ;iceland
B   white
F   black
N   gray
EOF

gmt plot -R0/12/0/6 -Jx1c -Baf -BWSen -Sc0.8c -W1p -Ctypes.cpt << EOF
1  3  -1
2  3  0
3  3  1
4  3  1.5
5  3  2
6  3  3
7  3  3.1
8 3  NaN
EOF

gmt colorbar -Ctypes.cpt -DJBC+e+n
gmt end show

Actual outcome

Expected outcome

The figure doesn't match the documentation:

1. It seems background and foreground colors (white and black) still works for categorical CPTs with numeric keys. If yes, we need to document it better, otherwise it's a bug.
2. If YES for point 1, then here is another bug. The color of the circle with z=3 should have a foreground color (black), but now it's red. If I change z=3 to z=3.1, it's black.
3. z=1.5 is in blue, but for categorical data, z=1.5 is an undefined value.

System information

• Operating system: macOS
• GMT version (gmt --version): 6.2.0_2cce034_2020.07.28

[stale[bot]]

This issue has been automatically marked as stale because it has not had activity in the last 90 days. It will be closed if no further activity occurs within 7 days. Thank you for your contributions.

[PaulWessel]

I do not think we have checks that might apply to categorical data. We should check that the input z value exactly matches the entries in the CPT and not the z = 1.5 which should give an error. People who want z from 1-2 to give a constant color should use an discrete CPT, not a categorical one. So we should start with implementing that test. Likewise I think back and foreground does not make too much sense either but I suppose for numerical categories we could allow that. For key-based categorically we clearly should not allow it. I will see what checks I can add.

[PaulWessel]

Shall we distinguish between numerical categorical CPTs and key-based (e.g. text) categorical CPTs? I see the CPT has an unsigned int categorical that is used as a bit flag and bits 1 and 2 are set but I see no use of the bits, just that it is nonzero. I am pretty sure that if you have categories, then any value not among the listed categories should return NaN and there cannot be a fore and background color. It should not matter that the keys happen to be numbers or strings.

[seisman]

I am pretty sure that if you have categories, then any value not among the listed categories should return NaN and there cannot be a fore and background color. It should not matter that the keys happen to be numbers or strings.

Yes, it makes more sense to me.

[seisman]

Closed #4407.