fix: Arc mark does not stack for `quantitative` definitions when different fields are used for `theta` and `radius` [draft] by hydrosquall · Pull Request #9512 · vega/vega-lite

hydrosquall · 2025-01-20T16:06:20Z

Motivation

When using different marks for theta vs radius using an arc mark with stacks, change graph rendering
Fixes Radial chart radius field unexpectedly changes arc stack behavior #7957

Changes

Do not include groupBy field when the dimension field is using a quantitative scale. This ensures that stacking is applied correctly.
I've made this change only impact polar coordinate channels to keep the impact minimal, since this customization is specific to the arc mark.
Try the VL Spec from the original bug report
- preview
- You can also check with the colorBy case
- For comparison this shape is cool, and shows we can still get something useful when the stack is applied on radius rather than theta

Testing

API design: does this approach make sense for how stacks should apply for arc marks?
See unit test suite : Let me know if additional cases should be added to either the image snapshots or the unit test suite.
Compare with similar cartesian plot

cloudflare-workers-and-pages · 2025-01-20T16:15:37Z

Deploying vega-lite with Cloudflare Pages

Latest commit:	`dd36450`
Status:	✅ Deploy successful!
Preview URL:	https://93a330b1.vega-lite.pages.dev
Branch Preview URL:	https://cameron-yick-fix-compile-pie.vega-lite.pages.dev

View logs

hydrosquall · 2025-01-20T16:19:52Z

src/stack.ts

+
+    if (!hasSameDimensionAndStackedField && (isPolar ? dimensionChannel === fieldChannel || encoding['color'] : true)) {
      // avoid grouping by the stacked field
+      // TKTK: find out why


If this works here, we may need a similar change for the code block under dimensionOffsetChannel .

hydrosquall · 2025-01-20T16:28:52Z

src/stack.ts

    const isPolar = isPolarPositionChannel(fieldChannel) || isPolarPositionChannel(dimensionChannel);

-    if (!hasSameDimensionAndStackedField && (isPolar ? dimensionChannel === fieldChannel || encoding['color'] : true)) {
+    if (!hasSameDimensionAndStackedField && (isPolar ? dimensionChannel === fieldChannel : true)) {


Before I removed the color encoding check, the graph rendered like this:

src/stack.ts

kanitw · 2025-02-18T16:29:00Z

src/stack.ts

+    const isPolar = isPolarPositionChannel(fieldChannel) || isPolarPositionChannel(dimensionChannel);
+    const shouldAddPolarGroupBy = !isUnbinnedQuantitative(dimensionDef);
+
+    if (isPolar ? shouldAddPolarGroupBy : !hasSameDimensionAndStackedField) {


I wonder if we really need isPolar ? shouldAddPolarGroupBy : clause? If we strictly check the field above, would that be sufficient?

By strictly check, do you mean we'd check if dimensionDef and stackedFieldDef have the same

field (the current check)

bin

aggregator

timeUnit

for both polar and cartesian graphs?

Currently if I just remove the isPolar ? shouldAddPolarGroupBy clause and use the existing check, we'd introduce this bug

Made a PoC commit to see if this is what you had in mind

8b010a6

we'd introduce this bug

can't find this link

Fixed link: https://github.com/vega/vega-lite/blob/231ed17c2b37811701a89029ab1524750abdd80b/examples/compiled/arc_quantitative_stacked_theta.png

I found it interesting to see how many of the visual tests break if I removed the check that only applies this change to polar encodings.

#9557

kanitw

Thanks for the PR. I think the test and regression case look good, but I wonder if the condition added can be more precise/simpler.

hydrosquall · 2025-03-06T19:13:07Z

Shrinking the example down to see if how the issue looks in both coordinate systems on current prod editor (not this branch)

cartesian version of the “bug” (having a common x1) here (pic 1)
polar version of the bug (having a common theta1) here (pic 2)

Cartesian	Polar

Note in the cartesian case, the stack transform did not get created when compiled to Vega (this is because Rect is not supported in the STACKABLE_MARKS constant).

Whereas in the polar case, the stack transform gets created, but has a groupBy of value .

…el and stackChannel are same

…hannels are matching

hydrosquall · 2025-05-01T16:17:30Z

I’ve been thinking more about why the polarCoordinates check was needed in this PR (and revisiting the grammar of graphics book) as I have tried a few variations on trying to get the intended behavior without checking for polar coordinates, and I noticed something that makes me feel more OK about having a special condition for polar.

We write that arc is analogous to rect
However, rect does not support stacking (even if forced), only bar does! (according to STACKABLE_MARKS)
Therefore, arc is more like bar than rect . If we want arc to have attributes of both rect and bar, we are going to have to add special logic for it in more places given that in cartesian coordinates, we aren't treating rect and bar the same way when it comes to stacking.

If we want stacking to be supported by arc that doesn’t have to care about whether we are in cartesian or polar, we have the choices of

Allow stacking to be forced in rect too (currently not supported)
Leave rect alone, and just allow forcing stacking via arc (as I do in the current PR) or
Introduce a new mark (radial_bar?) which is to arc as bar is to rect . We leave the current arc behavior alone, and we only give the new behavior of enabling force stacking to radial_bar.
Something else?

domoritz · 2025-07-03T11:49:44Z

@kanitw @hydrosquall what's left here to merge this?

hydrosquall · 2025-07-03T23:50:51Z

I’m hoping to get @kanitw’s review on whether this change is usable. We both wish a different solution that wasn’t polar aware might be possible, but I haven’t found a way to do it yet.

…

On Thu, Jul 3, 2025 at 7:50 AM Dominik Moritz ***@***.***> wrote: *domoritz* left a comment (vega/vega-lite#9512) <#9512 (comment)> @kanitw <https://github.com/kanitw> @hydrosquall <https://github.com/hydrosquall> what's left here to merge this? — Reply to this email directly, view it on GitHub <#9512 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACE2MM6BR2QAP6X2AD2XOST3GUKG7AVCNFSM6AAAAABVQWNFUCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTAMZRHE3TOMBQGY> . You are receiving this because you were mentioned.Message ID: ***@***.***>

domoritz · 2025-12-08T18:57:02Z

@kanitw can you do a review on this so we can merge? This just came up in Altair: #9680

hydrosquall changed the title ~~[draft]~~ [draft] Radial chart does not render as expected when different fields are used for theta and radius Jan 20, 2025

hydrosquall changed the title ~~[draft] Radial chart does not render as expected when different fields are used for theta and radius~~ fix: Radial chart does not render as expected when different fields are used for theta and radius [draft] Jan 20, 2025

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch 2 times, most recently from 0b7380a to 7f6ed6d Compare January 20, 2025 16:15

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch from 7f6ed6d to 3b1d979 Compare January 20, 2025 16:16

hydrosquall changed the title ~~fix: Radial chart does not render as expected when different fields are used for theta and radius [draft]~~ fix: Radial mark does not render as expected when different fields used for theta and radius [draft] Jan 20, 2025

hydrosquall commented Jan 20, 2025

View reviewed changes

hydrosquall mentioned this pull request Jan 20, 2025

Radial chart radius field unexpectedly changes arc stack behavior #7957

Open

5 tasks

hydrosquall self-assigned this Jan 20, 2025

hydrosquall commented Jan 20, 2025

View reviewed changes

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch 5 times, most recently from 89b647c to 369fa1e Compare January 26, 2025 02:31

hydrosquall changed the title ~~fix: Radial mark does not render as expected when different fields used for theta and radius [draft]~~ fix: Radial mark does not stack for quantitative definitions when different fields are used for theta and radius [draft] Jan 26, 2025

hydrosquall changed the title ~~fix: Radial mark does not stack for quantitative definitions when different fields are used for theta and radius [draft]~~ fix: Arc mark does not stack for quantitative definitions when different fields are used for theta and radius [draft] Jan 26, 2025

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch from a21fed9 to 905e53c Compare January 26, 2025 02:55

hydrosquall marked this pull request as ready for review January 26, 2025 03:23

hydrosquall requested a review from a team as a code owner January 26, 2025 03:23

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch from 905e53c to 6d37199 Compare January 30, 2025 05:02

kanitw reviewed Feb 18, 2025

View reviewed changes

src/stack.ts Show resolved Hide resolved

kanitw reviewed Feb 18, 2025

View reviewed changes

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch from 6d37199 to ca62a23 Compare February 19, 2025 04:02

hydrosquall mentioned this pull request Feb 19, 2025

test: show effect of checking for same dimension and stacked field in polar and cartesian fields #9537

Draft

hydrosquall requested a review from kanitw February 23, 2025 17:58

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch from ca62a23 to 2ebd1b1 Compare March 3, 2025 01:13

hydrosquall mentioned this pull request Mar 8, 2025

fix(compile): Check if groupBy check can be generalized to non-polar coordinates #9557

Draft

5 tasks

hydrosquall added 5 commits May 1, 2025 09:30

fix(stack): do not add field to a stack's groupBy unless groupByChann…

2ea24ae

…el and stackChannel are same

test: check ensure groupBy is not added for stack when radial chart c…

d071466

…hannels are matching

test(stack): limit radius of change to polar coordinate graphs

65c28e1

fix: settle stacked arc marks with a check for field type

37ab111

fix: add test case for quantitative stacked scale theta graphs

6763a4a

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch from 158366e to dd36450 Compare May 1, 2025 13:32

hydrosquall mentioned this pull request May 1, 2025

feat(dataviz): check simpler unbinned quantitative test solution #9601

Draft

5 tasks

chore: update examples [CI]

c5eae62

hydrosquall force-pushed the cameron.yick/fix-compile-pie-stack branch from d8b3ab4 to c5eae62 Compare May 1, 2025 13:40

GitHub Actions Bot added 2 commits May 1, 2025 13:43

chore: update examples [CI]

9a260ed

style: auto-formatting [CI]

f7f2600

hydrosquall mentioned this pull request Dec 6, 2025

Independent theta and radius in pie chart renders incorrectly #9680

Open

Uh oh!

Conversation

hydrosquall commented Jan 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Changes

Testing

Uh oh!

cloudflare-workers-and-pages bot commented Jan 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Deploying vega-lite with Cloudflare Pages

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kanitw left a comment

Choose a reason for hiding this comment

Uh oh!

hydrosquall commented Mar 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

hydrosquall commented May 1, 2025

Uh oh!

domoritz commented Jul 3, 2025

Uh oh!

hydrosquall commented Jul 3, 2025 via email

Uh oh!

domoritz commented Dec 8, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

hydrosquall commented Jan 20, 2025 •

edited

Loading

cloudflare-workers-and-pages bot commented Jan 20, 2025 •

edited

Loading

hydrosquall commented Mar 6, 2025 •

edited

Loading