Keeping the auto-scaling to modern mode in line with other auto-scalings sounds good to me.

Ping @Esteban82, @joa-quim and @meghanrjones. I have implemented auto-scaling for the magnetic rose as well. Please build from this branch and use this script to explore and give feedback on the scalings I chose. FYI, I set the scales so that for size 2.5i (the original script) the roses looks similar. For smaller sizes things scale down accordingly:

#!/usr/bin/env bash
# Change W for smaller compass whose attributes scale with size
W=2.5i
gmt begin mag
	gmt set GMT_THEME cookbook
# Magnetic rose with a specified declination
	gmt basemap -R-10/-2/12.8812380332/0.661018975345+r -JOc0/0/50/60/7i -B0 -X1.25i
	gmt basemap -Tmg-2/0.5+w${W}+d-14.5+t45/10/5+i0.15p,blue+p0.15p,red+l+jCM \
		--FONT_ANNOT_PRIMARY=9p,Helvetica,blue --FONT_ANNOT_SECONDARY=12p,Helvetica,red --FONT_LABEL=14p,Times-Italic,darkgreen \
		--MAP_FRAME_WIDTH=10p --COLOR_BACKGROUND=green --MAP_DEFAULT_PEN=2p,darkgreen --COLOR_BACKGROUND=darkgreen \
		--MAP_VECTOR_SHAPE=0.5 --MAP_TICK_PEN_SECONDARY=thinner,red --MAP_TICK_PEN_PRIMARY=thinner,blue
EOF
gmt end show

See gmt_plot.c line 3308 for where the scales are - feel free to experiment changing those numbers if you think they can be improved.

Note: I have not yet dealt with positions and adjustments due to text size as I have done for directional rose - that comes later.

Would auto-scaling for the intervals typically set using +t be possible? Or an option to apply MAP_ANNOT_MIN_SPACING to those?

NZ: These are tiny maps (3i) so default symol is 0.3 inches or about 0.75 cm. It is small - the original had a fixed 0.5i symbol. I can restore to 0.5i but will still fail due to (now) correct placement. This figure is included in the cylindrical doc section so we do want it to be readable. OK, have set it back to 0.5i since the script is not shown anyway.

There have been no change to which font is used, only size of that font.

NZ: These are tiny maps (3i) so default symol is 0.3 inches or about 0.75 cm. It is small - the original had a fixed 0.5i symbol. I can restore to 0.5i but will still fail due to (now) correct placement. This figure is included in the cylindrical doc section so we do want it to be readable. OK, have set it back to 0.5i since the script is not shown anyway.

The rose on the right panel looks to still be shrunk:

There have been no change to which font is used, only size of that font.

Ah, as you said it is set by FONT_TITLE which is Helvetica-Bold.

My main concern is about MAP_EMBELLISHMENT_MODE taking precedence over FONT_TITLE in this PR. I think that a user should not need to set both MAP_EMBELLISHMENT_MODE=manual and FONT_TITLE=9p for their font preference to be applied, since they may prefer auto offset with a specific font size or already have scripts in modern mode with a specific font size. For example, the relevant line in GMT_obl_nz.sh sets --FONT_TITLE=9p but this does not affect the plot in this PR.

TYpo in script, now same size.
As for FONT: yes the 9p has no effect. I see the auto affecting lots of parts and it would make little sense to micro-manage font sizes from the user. THis is an older script so I should just remove that setting (predates this PR).

As for FONT: yes the 9p has no effect. I see the auto affecting lots of parts and it would make little sense to micro-manage font sizes from the user. THis is an older script so I should just remove that setting (predates this PR).

Do you mean that you prefer the current behavior in this PR of MAP_EMBELLISHMENT_MODE overriding FONT_TITLE settings?

Yes, because I do not know how to track what was set manually. Remember, this is after the auto-scaling of fonts etc have already been set for the map. So without a set of book-keeping I would not know if I should not change the font-size because the user wanted 24p. They can change the font, but not manually set a size and still benefit from all the other auto-things. I think the vast amount of users will be pleased they dont have to think about these settings, and a few may wish to do things entirely manually.

OK, I think it's good to go after removing the now irrelevant settings from the scripts.

FYI: This PR got delayed a bit because the alignment of the N-arrow in example 4 was off. The solution is to temporarily reset the PostScript matrix projection, compute the offsets, then reinstate the projection (the -p60/30 as implemented in PostScript) as the offsets need to be measured without perspective. Now, ex04 has a good default placement and size I think. About 8 tests related to map roses, including some new tests with no PS files, fail. I think we will add these to dvc soon.