Realigning order of basemap frame, gridlines, and annotations/ticks for all modules #4274

PaulWessel · 2020-09-30T02:54:10Z

Description of proposed changes

See #4265 for contest. We want to add consistency to how GMT places the "basemap" on top of or below a plot within the same module (obviously, stacking modules can override any default arrangements). In general, we now place the optional gridlines before plotting data, but include images and polygons (e.g., via pscoast) in which case gridlines goes on top. We generally lay down the frame on top to ensure any lines or symbols at or near the boundary do not "eat" into the frame width, but if the module has a no-clip option like -N then we draw it first so symbols can straddle that line. Annotations and tick-marks are usually laid down before the plot but in the case of MAP_FRAME_TYPE=inside then we must do it at the end instead. There are also some exceptions for true 3-D plots.

Implementation is done via a new function gmt_set_basemap_orders which sets, for each of the discussed items, what the order of plotting should be. For instance, in pscoast it may say

gmt_set_basemap_orders (GMT, clipping ? GMT_BASEMAP_FRAME_BEFORE : GMT_BASEMAP_FRAME_AFTER, GMT_BASEMAP_GRID_AFTER, GMT_BASEMAP_ANNOT_AFTER);

where clipping options may require the frame to be drawn first instead of last, but otherwise we draw items later. Each plot module has such a call and the arguments vary. The function itself can override the settings based on some of the conditions discussed above.

After implementation, we now have ~180+ "failures" in the tests. Most of them are of these types:

Some plots that had gridlines on top now has gridlines on the bottom, avoiding cutting through symbols and bars
Some plots with polygons or bars near a border (e.g., histograms) now have a clean frame line without notches
Many base frames have subtle changes because tick marks are not plotted and are visible whereas in some cases they were covered up.

I have labeled this PR WIP because I think we need to actually look at all these to see if anything got lost in the translation that we want to preserve.

Now, annotations are always at end but gridlines may happen before plotting of data.

Starge messing with -R -J inside makes it difficult to do later

PaulWessel · 2020-09-30T20:35:57Z

I know this is a big tasks to look at those 180 PNG and PDFs but I think we need all of us to do this in case there are some subtle regressions. I've stared at these for a while and the results look good to me. Let me know of you see anything that should not change, @joa-quim and @seisman.

joa-quim · 2020-09-30T21:46:56Z

I got only 133 failures in total.

PaulWessel · 2020-09-30T21:58:12Z

That is a new one - less failures on Windows than macOS (186 right now).

joa-quim · 2020-09-30T22:00:55Z

Ex16 is partially eating the frame

PaulWessel · 2020-09-30T22:09:49Z

Yes, I need to fix the setting for grdview. It draws before but shoud be after (unless 3-D etc).

seisman · 2020-09-30T22:22:17Z

doc/script/GMT_inset.sh: Tiny changes in the coastline
ex04, ex05: Tiny changes in the title string
ex08: The z-axis at the back should be invisible
ex11: this example is messy now.
ex15: in panel (d), some pixels overlap with the frame. The old PS looks better to me.
ex16: @joa-quim already mentioned above

joa-quim · 2020-09-30T22:34:54Z

grdcut/origin. The inner pajama frame is overlapped

PaulWessel · 2020-09-30T22:41:51Z

doc/script/GMT_inset.sh: Tiny changes in the coastline
No idea why coastline is affected, but perhaps I removed a pen reset - will look

ex04, ex05: Tiny changes in the title string
Unsure about what that is

ex08: The z-axis at the back should be invisible
working on that

ex11: this example is messy now.
Yes must be a major error I missed

ex15: in panel (d), some pixels overlap with the frame. The old PS looks better to me.
Fixing that in mask

ex16: @joa-quim already mentioned above
Same, fixing in grdview

PaulWessel · 2020-09-30T22:42:26Z

grdcut/origin. The inner pajama frame is overlapped

But that was always the case. We do -B in the first command and then pile on. This needs moving the -B to the last call.

PaulWessel · 2020-09-30T22:47:38Z

I think we have many tests where we plot numerous overlays and often start with -Baf. To follow our plan I think these need to move to the last command. Finally, we do not yet have a solution for subplots since it will add -B settings to the first command in the panel, not last. Changing this is harder and will be pursued after this PR.

joa-quim · 2020-09-30T22:47:50Z

grdimage/grdclip

joa-quim · 2020-09-30T22:48:25Z

grdview/categorical

PaulWessel · 2020-10-11T02:50:23Z

Apologies for fixing #4181 in this PR but I wanted to be safe it did not break other things. The double frame is now gone.

seisman · 2020-10-11T03:13:30Z

: The new PS file looks great in ghostview (gv) and test passes, but if I tried open anim04.ps then Preview would show the black checkerboard at the wrong angle.

Yes, we saw it before and we thought it may be a Preview bug.

seisman · 2020-10-11T03:22:45Z

I have two more failing tests than you, but there are just tiny differences:

	596 - test/psbasemap/asymm_grid.sh (Failed)
	640 - test/pscoast/placement.sh (Failed)

PaulWessel · 2020-10-12T02:00:43Z

So OK with me updating the orig PS in this PR then?

seisman · 2020-10-12T02:03:51Z

Yes to me.

PaulWessel · 2020-10-12T05:49:47Z

Hi @seisman ane @joa-quim, I think we have reached the end of the line on this PR. This clearly improved the quality of the plots. The one thing remaining is related to subplot panels where the -B happens on the first panel but perhaps should always be last. That will require more work since we cannot know we are last until we either move to a new frame or we are done. That is a separate PR.

seisman · 2020-10-12T06:20:09Z

../../src/psrose.c:799:10: warning: unused variable 'dim' [-Wunused-variable]
                double dim[PSL_MAX_DIMS];

PaulWessel · 2020-10-12T19:04:28Z

Hi @seisman, seems a CI failure but nothing to be concerned about? @joa-quim, may be helpful if you could run all the test one more time on this branch. I think all is well now with expected result apart form the notes on subplot panel orders.

seisman · 2020-10-12T19:58:01Z

Hi @seisman, seems a CI failure but nothing to be concerned about?

Yes, just an occasional network connection issue.

joa-quim · 2020-10-12T20:29:55Z

Same number of failures (16). Some, doc/scripts/GMT_tut_11.sh, 18, etc fail because the grid is slightly different. Don't get it.

seisman · 2020-10-12T22:36:50Z

I just made a commit 784347f to trigger full tests in CI. Will know if all tests pass in CI in 2 hours.

seisman · 2020-10-13T00:07:31Z

This PR is good to me, but please check this file test/psxy/clipping5.ps.

PaulWessel · 2020-10-13T00:08:47Z

OK, soon after finishing another PR.

PaulWessel · 2020-10-13T01:55:10Z

So clipping5.sh is flagged as GMT_KNOWN_FAILURE. I think it is not improved or degraded by this PR though; its solution lies elsewhere.

seisman · 2020-10-13T02:39:26Z

So clipping5.sh is flagged as GMT_KNOWN_FAILURE. I think it is not improved or degraded by this PR though; its solution lies elsewhere.

OK, I didn't notice that. The CI looks good. One failure on Linux and 7 failures on Windows. They are all known failures.

seisman · 2020-10-13T02:41:00Z

I just disabled the full tests in e529b14. Now this PR looks good to me.

PaulWessel added 13 commits September 27, 2020 08:17

Ensure gridlines are layed down first except for images and similar

44a4fd6

Now, annotations are always at end but gridlines may happen before plotting of data.

Merge branch 'master' into gridline-revolt

d4bbf0e

More refinement of grdlines

87f270a

pssolar with gridlines at end

3a82b78

Merge branch 'master' into gridline-revolt

2db3a23

Merge branch 'master' into gridline-revolt

d8653bd

New approach

c172b0f

Update psrose.c

8d289d9

Merge branch 'master' into gridline-revolt

5cb5e19

Dont reuse n_bins for somethng else

8dc86ba

Merge branch 'master' into gridline-revolt

1ff3d18

Lay down -B before imates in psimage for now

461c4b7

Starge messing with -R -J inside makes it difficult to do later

Let -B gridlines work after image anyway.

d6353da

PaulWessel requested review from joa-quim and seisman September 30, 2020 02:54

PaulWessel self-assigned this Sep 30, 2020

Merge branch 'master' into gridline-revolt

8cc07bf

Merge branch 'master' into gridline-revolt

84d2310

PaulWessel added 2 commits October 10, 2020 15:52

Merge branch 'master' into gridline-revolt

a3bd21c

Fix double frame reported

8c00432

PaulWessel added 2 commits October 11, 2020 17:10

Merge branch 'master' into gridline-revolt

7805261

Update tinye PS changes

768193e

Update psrose.c

f2767b4

PaulWessel mentioned this pull request Oct 12, 2020

Double frame on grdview perspective plot #4181

Closed

Temporarily enable full CI tests

784347f

GenericMappingTools deleted a comment from codecov-io Oct 12, 2020

Disable full tests

e529b14

seisman approved these changes Oct 13, 2020

View reviewed changes

PaulWessel merged commit fb62cfb into master Oct 13, 2020

PaulWessel deleted the gridline-revolt branch October 13, 2020 02:58

seisman mentioned this pull request Oct 15, 2020

Baseline images need updates due to the recent updates of earth relief data GenericMappingTools/pygmt#451

Closed

maxrjones mentioned this pull request Jun 3, 2021

The -i for memory datasets via matrix used the wrong parameter #5289

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Realigning order of basemap frame, gridlines, and annotations/ticks for all modules #4274

Realigning order of basemap frame, gridlines, and annotations/ticks for all modules #4274

PaulWessel commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

joa-quim commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

joa-quim commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

seisman commented Sep 30, 2020

joa-quim commented Sep 30, 2020 •

edited

Loading

PaulWessel commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

joa-quim commented Sep 30, 2020

joa-quim commented Sep 30, 2020

PaulWessel commented Oct 11, 2020

seisman commented Oct 11, 2020

seisman commented Oct 11, 2020

PaulWessel commented Oct 12, 2020

seisman commented Oct 12, 2020

PaulWessel commented Oct 12, 2020

seisman commented Oct 12, 2020

PaulWessel commented Oct 12, 2020

seisman commented Oct 12, 2020

joa-quim commented Oct 12, 2020

seisman commented Oct 12, 2020 •

edited

Loading

seisman commented Oct 13, 2020

PaulWessel commented Oct 13, 2020

PaulWessel commented Oct 13, 2020

seisman commented Oct 13, 2020

seisman commented Oct 13, 2020

Realigning order of basemap frame, gridlines, and annotations/ticks for all modules #4274

Realigning order of basemap frame, gridlines, and annotations/ticks for all modules #4274

Conversation

PaulWessel commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

joa-quim commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

joa-quim commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

seisman commented Sep 30, 2020

joa-quim commented Sep 30, 2020 • edited Loading

PaulWessel commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

PaulWessel commented Sep 30, 2020

joa-quim commented Sep 30, 2020

joa-quim commented Sep 30, 2020

PaulWessel commented Oct 11, 2020

seisman commented Oct 11, 2020

seisman commented Oct 11, 2020

PaulWessel commented Oct 12, 2020

seisman commented Oct 12, 2020

PaulWessel commented Oct 12, 2020

seisman commented Oct 12, 2020

PaulWessel commented Oct 12, 2020

seisman commented Oct 12, 2020

joa-quim commented Oct 12, 2020

seisman commented Oct 12, 2020 • edited Loading

seisman commented Oct 13, 2020

PaulWessel commented Oct 13, 2020

PaulWessel commented Oct 13, 2020

seisman commented Oct 13, 2020

seisman commented Oct 13, 2020

joa-quim commented Sep 30, 2020 •

edited

Loading

seisman commented Oct 12, 2020 •

edited

Loading