[susi]

This pr fixes the -g flag so that it clamps any values < 1.0 to 0.

For -a it does the same, and also uses log(fabs(min)) when calculating the interval instead of assuming 0.

r.colors map=censusblk_median_age col=plasma

r.colors map=censusblk_median_age col=plasma -g

r.colors map=censusblk_median_age col=plasma -a

[veroandreo]

Thanks @susi ! Would you mind updating/including a minimal explanation in the manual page?

[HuidaeCho]

@susi Please check discussions in #1481, which tries to address the same issue.

[HuidaeCho]

Related PR #1507

[HuidaeCho]

I think we have to discuss how to address any questions raised in #1481 and #1507 rather than creating a new PR that conflicts with the earlier #1481 and duplicating the effort. This color issue should be addressed in both the display and legend.

[susi]

Oh thanks @HuidaeCho I was not aware of any other PRs in progress, since there were not linked to the issue 😕 It wasn't my intention to override/conflict with existing work. Is there anything I can do to help with #1481? Would the stuff I did for this PR be applicable for #1481? If so I can rebase it on top of that.

[susi]

Thanks @susi ! Would you mind updating/including a minimal explanation in the manual page?

Sure once we agree we want any of this 😝

[HuidaeCho]

Oh thanks @HuidaeCho I was not aware of any other PRs in progress, since there were not linked to the issue 😕 It wasn't my intention to override/conflict with existing work. Is there anything I can do to help with #1481? Would the stuff I did for this PR be applicable for #1481? If so I can rebase it on top of that.

I understand. The link is in the middle of the long discussion, so it's hard to find ;-).

The current #1481 uses eps=1.0, which in effect does the same thing as your PR. Additionally, it tries to address non-positive rasters. I know, log negative doesn't make sense, but at least we want to display a single-color raster or throw a fatal error. The main thing is both the display (#1481) and legend (#1507) need to use the same colors for consistency.

Please read/test both PRs and let me know what you think.

[susi]

Oh thanks @HuidaeCho I was not aware of any other PRs in progress, since there were not linked to the issue confused It wasn't my intention to override/conflict with existing work. Is there anything I can do to help with #1481? Would the stuff I did for this PR be applicable for #1481? If so I can rebase it on top of that.

I understand. The link is in the middle of the long discussion, so it's hard to find ;-).

Ah I was talking about the Linked Pull Requests, also there was no-one assigned to the bug, so it looked like there had been conversations that had just stalled 🤔

The current #1481 uses eps=1.0, which in effect does the same thing as your PR. Additionally, it tries to address non-positive
rasters. I know, log negative doesn't make sense, but at least we want to display a single-color raster or throw a fatal error. The
main thing is both the display (#1481) and legend (#1507) need to use the same colors for consistency.

As you know, the way r.colors works, is that it uses the raster library functions like Rast_log_colors or Rast_abs_log_colors to set the color map, and then saves the color file with Rast_write_colors. Then d.legend also uses the raster library to read the same color map. So it seems to me to be a user error to run r.colors -g and d.legend -l And expect the color ramp to match the colors of the map. It means that you are essentially running log10 of whatever r.colors set (like log if you run with -g). 🤔

Please read/test both PRs and let me know what you think.

Sure, happy to! I'll leave comments on those PRs 🙂

[HuidaeCho]

Please try this:

r.mapcalc ex=tmp="censusblk_median_age/85.599998/2" # to create a raster in [0, 0.5]
r.colors map=tmp col=plasma -g

In this case, lmin=1e-15 and lmax=-0.693147, and generates a color rule file between 0 and 0.99309249543703693, which is not its data range between 0 and 0.5.

% 0 0.99309249543703693
nv:255
*:255
0:13:8:135 0.99309249543703693:20:7:137
0.98623270449336009:27:6:141 0.99309249543703693:20:7:137
0.97942029758692783:33:6:143 0.98623270449336009:27:6:141
...
0.50697973989501455:243:239:39 0.51050606285359668:245:235:39
0.50347777502835933:241:244:38 0.50697973989501455:243:239:39
0.5:240:249:33 0.50347777502835933:241:244:38

Note that the value jumps to 0.99309 and slowly declines to 0.5.

[susi]

Closing this as #1937 does it better :)