Compute vector scaling correctly #1988

aashish24 · 2016-05-23T18:44:02Z

VTK does clamping and not remapping of input range to desired
range which leads to undesirable visual effects.

VTK does clamping and not remapping of input range to desired range which leads to undesirable visual effects.

aashish24 · 2016-05-23T19:55:57Z

Ref: https://github.com/UV-CDAT/uvcdat-testdata/pull/135/files

aashish24 · 2016-05-23T19:56:17Z

@danlipsa @doutriaux1 please review

aashish24 · 2016-05-23T21:35:24Z

Ref #1977

danlipsa · 2016-05-23T21:56:13Z

Packages/vcs/vcs/vcsvtk/vectorpipeline.py

-        # [min, max] --> [0, 1].
-        glyphFilter.ClampingOn()
-        glyphFilter.SetRange(0.01, 1.0)
-


It seems the original problem was triggered by this code. By clamping everything to 1, you end-up with vectors of the same size. I'll post an image generated with this code commented out.

danlipsa · 2016-05-23T21:58:12Z

The scaling factor is not quite right, but scaling does happen. This is test_vcs_basic_vector.png

doutriaux1 · 2016-05-23T22:42:40Z

I love the flies swarming look to this!

aashish24 · 2016-05-24T03:35:38Z

@danlipsa yes, all of these problems are fixed in my PR. Please have a look at the new baselines and my commit message on the source of the problem.

danlipsa · 2016-05-24T14:33:06Z

@aashish24 Why do you need to compute another scalar? You could use ScaleByVector isn't it?

danlipsa · 2016-05-24T15:47:50Z

@aashish24 @doutriaux1 After a discussion with @aashish24 this LGTM as long as users know/are OK with a vector magnitude range remapping to hard-coded [0.1:1]. In the future we'll add options to vcs to show data as is, clamp it low and high, use linear rescalling (what @aashish24 does), use logarithmic rescalling.

aashish24 · 2016-05-24T15:49:27Z

@danlipsa and I had a good conversation about it and I pointed out that there are few API parameters are missing at this point. I would like to propose those changes post 2.6 release. The options will be

An API to enable / disable clamping and option to set the clamping range, default to OFF
An API to enable / disable re-mapping vector magnitude range (logarithmic, linear etc.), by default to OFF

For now, this branch takes care of the vector appearing of the same size because of clamping and keeping the old look and feel for most part.

aashish24 · 2016-05-24T15:55:45Z

Forgot to mention that I am okay with taking out the linear range mapping from this branch (or have it not exposed) too, although that means that I would need to look into the uniform scaling factor more closely.

danlipsa · 2016-05-24T16:02:15Z

Indeed this was a good discussion - we clarified the trade-offs for each approach and a path for fixing this properly. With @doutriaux1 input we can make a choice for the 2.6 release. I am also fine with either path we choose.

doutriaux1 · 2016-05-24T16:17:21Z

@aashish24 while you're at it, we need to add the legend for vectors. use template.legend.x1/y1 for the location. The legend should be arrow of constant length with a number showing the magnitude value for this length. Does this make sense?

aashish24 · 2016-05-24T17:42:35Z

@doutriaux1 @danlipsa I am adding few options to the graphic method, I will push my changes today.

doutriaux1 · 2016-05-24T20:33:00Z

@aashish24 thanks! Let me know when I can pull the trigger on a build.

aashish24 · 2016-05-25T12:41:34Z

So I added an option to the vector graphics method that could be used in the script like shown below:

import cdms2
import vcs
data = cdms2.open("./../build_debug/install/share/uvcdat/sample_data/clt.nc")
v = data['v']
u = data['u']
x = vcs.init()
gv = vcs.createvector()
gv.scaletype = 'linear'
x.plot(u, v, gv)
x.interact()

You have five options to chose from (I am open for name suggestions)
('off', 'constant', 'normalize', 'linear', 'constantNNormalize', 'constantNLinear')
off --> Do not apply any scaling (scale factor or otherwise), just show the vector magnitude
constant --> Apply constant scale factor that is gm.scale only
normalize --> Apply scale by computing the maximum vector magnitude and normalize all other vectors to it
linear --> Apply scale computing using the user provided range (using scalerange) and map vector magnitudes to that range.

I have attached images for all the options below for comparison:

aashish24 · 2016-05-26T02:23:42Z

@doutriaux1 this branch is ready for another review. Please let me know if you have better names for the options.

doutriaux1 · 2016-05-26T14:23:49Z

Packages/vcs/vcs/vector.py

@@ -528,6 +535,26 @@ def _setalignment(self, value):
        self._alignment = value
    alignment = property(_getalignment, _setalignment)

+


@aashish24 I would rather the set/get function to be in VCS_Validation functions, just in case we ever need to re-use it for something else and have here

scalerange = VCS_validation_functions.scalerange

see colormap above for example.

doutriaux1 · 2016-05-26T14:32:25Z

@aashish24 I made a few comments, building locally.

doutriaux1 · 2016-05-27T18:11:33Z

@aashish24 see: https://open.cdash.org/viewTest.php?onlyfailed&buildid=4385643

     79 - test_vcs_background_mode_rotate (Failed)
    354 - test_vcs_isoline_width_stipple (Failed)
    413 - test_vcs_vectors_scale_options (Failed)
    453 - test_vcs_line_patterns (Failed)
    524 - dv3d_vector_test (Failed)
    525 - dv3d_slider_test (Failed)
    526 - Hovmoller_volume_test (Failed)
    527 - dv3d_volume_test (Failed)
    528 - dv3d_constituents_test (Failed)
    529 - dv3d_surface_test (Failed)
    530 - dv3d_hovmoller_test (Failed)
    588 - diags_test_02 (Failed)
    589 - diags_test_03 (Failed)
    590 - diags_test_04 (Failed)
    591 - diags_test_41 (Failed)
    592 - diags_test_05 (Failed)
    593 - diags_test_06 (Failed)
    594 - diags_test_07 (Failed)
    595 - diags_test_08 (Failed)
    596 - diags_test_09 (Failed)
    597 - diags_test_10 (Failed)
    598 - diags_test_11 (Failed)
    599 - diags_test_12 (Failed)
    600 - diags_test_13 (Failed)
    601 - diags_test_15 (Failed)
    603 - basemap_verify_import (Failed)

aashish24 · 2016-05-28T00:39:52Z

thanks @doutriaux1. test_vcs_vectors_scale_options is failing because I didn't have bg=1 but all other ones are failing not because of my branch.

doutriaux1 · 2016-05-28T04:48:01Z

@aashish24 how come the test suite got dirtied then? I though your new test suite had everything cleaned up? Beside some of these test do not fail on @danlipsa branch.

aashish24 · 2016-05-28T08:50:44Z

@doutriaux1 I added the vector method to the dump json test and made sure that it is correct. Please have a look at the branch now.

aashish24 · 2016-05-28T18:44:46Z

@doutriaux1 I don't know as some of these are not failing on my machine. I am doing a clean build of CDAT just to be sure. The DV3D tests are failing because of some key error. Some of the diagnostic tests are failing because of color map issues. Two of them might be failing because i made the bg=1 default if not specified.

aashish24 · 2016-05-28T19:08:16Z

On master, I have these tests failing:

The following tests FAILED:
354 - test_vcs_isoline_width_stipple (Failed)
452 - test_vcs_line_patterns (Failed)
523 - dv3d_vector_test (Failed)
524 - dv3d_slider_test (Failed)
525 - Hovmoller_volume_test (Failed)
526 - dv3d_volume_test (Failed)
527 - dv3d_constituents_test (Failed)
528 - dv3d_surface_test (Failed)
529 - dv3d_hovmoller_test (Failed)
575 - CDMS_Test_multiple_formats (Failed)
587 - diags_test_02 (Failed)
588 - diags_test_03 (Failed)
589 - diags_test_04 (Failed)
590 - diags_test_41 (Failed)
591 - diags_test_05 (Failed)
592 - diags_test_06 (Failed)
593 - diags_test_07 (Failed)
594 - diags_test_08 (Failed)
595 - diags_test_09 (Failed)
596 - diags_test_10 (Failed)
597 - diags_test_11 (Failed)
598 - diags_test_12 (Failed)
599 - diags_test_13 (Failed)
600 - diags_test_15 (Failed)

danlipsa · 2016-05-31T14:09:09Z

@aashish24 switching to the system python from the brew one (and rebuilding uvcdat) fixed the dv3d_ tests for me. What are the generated files for the stipple tests?

aashish24 · 2016-05-31T14:40:29Z

@danlipsa let's file another bug issue and carry discussion on failing tests there. I will post information there.

aashish24 · 2016-05-31T14:43:43Z

@danlipsa please see this: #1995

Also, can we merge this branch (I don't see any major objections)?

danlipsa · 2016-05-31T14:52:34Z

@aashish24 fine with me.

aashish24 · 2016-05-31T14:55:40Z

thanks @danlipsa. @doutriaux1 if you have any other concern, I will make sure to address them quickly.

doutriaux1 · 2016-05-31T14:59:09Z

rebuilding from scratch, will let you know.

doutriaux1 · 2016-05-31T14:59:44Z

@aashish24! Really! How quicker could I have been!

aashish24 force-pushed the fix_vector_scale branch from 7836cd2 to 2095654 Compare May 23, 2016 18:44

Compute vector scaling correctly

1129e5b

VTK does clamping and not remapping of input range to desired range which leads to undesirable visual effects.

aashish24 force-pushed the fix_vector_scale branch from 2095654 to 1129e5b Compare May 23, 2016 18:47

Set default to 1 for range

9589f12

aashish24 changed the title ~~[WIP] Compute vector scaling correctly~~ Compute vector scaling correctly May 23, 2016

Using descriptive variable names

9648e96

Fixed flake8 test

2a4089e

danlipsa reviewed May 23, 2016
View reviewed changes

aashish24 force-pushed the fix_vector_scale branch 5 times, most recently from 742681d to 12dbc57 Compare May 25, 2016 12:33

aashish24 added 2 commits May 25, 2016 08:52

Added scale options to vector graphic method

2513a0c

Added check for valid range

e1d77e4

Fixed using wrong baseline image

b69d09d

doutriaux1 reviewed May 26, 2016
View reviewed changes

aashish24 added 2 commits May 27, 2016 13:12

Fixed flake8

c30239c

Updated method name for consistency

9a2d02d

aashish24 added 3 commits May 28, 2016 03:52

Made background rendering default for testing

a7f5b86

Using consistent naming scheme

b59d84e

Fixed another flake8 issue

c8befa3

aashish24 force-pushed the fix_vector_scale branch from 5efb441 to c8befa3 Compare May 28, 2016 08:29

Using more generic name for testing

e3f7ede

aashish24 force-pushed the fix_vector_scale branch from e645cf3 to e3f7ede Compare May 28, 2016 08:47

Added vector method for testing

8c05ac4

aashish24 merged commit 8a8c1e9 into master May 31, 2016

aashish24 deleted the fix_vector_scale branch May 31, 2016 14:55

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Compute vector scaling correctly #1988

Compute vector scaling correctly #1988

aashish24 commented May 23, 2016

aashish24 commented May 23, 2016

aashish24 commented May 23, 2016

aashish24 commented May 23, 2016

danlipsa May 23, 2016

danlipsa commented May 23, 2016 •

edited

Loading

doutriaux1 commented May 23, 2016

aashish24 commented May 24, 2016

danlipsa commented May 24, 2016

danlipsa commented May 24, 2016

aashish24 commented May 24, 2016

aashish24 commented May 24, 2016 •

edited

Loading

danlipsa commented May 24, 2016

doutriaux1 commented May 24, 2016

aashish24 commented May 24, 2016

doutriaux1 commented May 24, 2016

aashish24 commented May 25, 2016 •

edited

Loading

aashish24 commented May 26, 2016

doutriaux1 May 26, 2016

doutriaux1 commented May 26, 2016

doutriaux1 commented May 27, 2016

aashish24 commented May 28, 2016

doutriaux1 commented May 28, 2016

aashish24 commented May 28, 2016

aashish24 commented May 28, 2016

aashish24 commented May 28, 2016

danlipsa commented May 31, 2016

aashish24 commented May 31, 2016

aashish24 commented May 31, 2016

danlipsa commented May 31, 2016

aashish24 commented May 31, 2016

doutriaux1 commented May 31, 2016

doutriaux1 commented May 31, 2016

		@@ -528,6 +535,26 @@ def _setalignment(self, value):
		self._alignment = value
		alignment = property(_getalignment, _setalignment)

Compute vector scaling correctly #1988

Compute vector scaling correctly #1988

Conversation

aashish24 commented May 23, 2016

aashish24 commented May 23, 2016

aashish24 commented May 23, 2016

aashish24 commented May 23, 2016

danlipsa May 23, 2016

Choose a reason for hiding this comment

danlipsa commented May 23, 2016 • edited Loading

doutriaux1 commented May 23, 2016

aashish24 commented May 24, 2016

danlipsa commented May 24, 2016

danlipsa commented May 24, 2016

aashish24 commented May 24, 2016

aashish24 commented May 24, 2016 • edited Loading

danlipsa commented May 24, 2016

doutriaux1 commented May 24, 2016

aashish24 commented May 24, 2016

doutriaux1 commented May 24, 2016

aashish24 commented May 25, 2016 • edited Loading

aashish24 commented May 26, 2016

doutriaux1 May 26, 2016

Choose a reason for hiding this comment

doutriaux1 commented May 26, 2016

doutriaux1 commented May 27, 2016

aashish24 commented May 28, 2016

doutriaux1 commented May 28, 2016

aashish24 commented May 28, 2016

aashish24 commented May 28, 2016

aashish24 commented May 28, 2016

danlipsa commented May 31, 2016

aashish24 commented May 31, 2016

aashish24 commented May 31, 2016

danlipsa commented May 31, 2016

aashish24 commented May 31, 2016

doutriaux1 commented May 31, 2016

doutriaux1 commented May 31, 2016

danlipsa commented May 23, 2016 •

edited

Loading

aashish24 commented May 24, 2016 •

edited

Loading

aashish24 commented May 25, 2016 •

edited

Loading