fixes all plot types reprojections by lazarusA · Pull Request #691 · EarthyScience/Browzarr

lazarusA · 2026-07-18T21:06:03Z

closes #690

Our test case goes from

+proj=lcc +lat_1=38.5 +lat_2=38.5 +lat_0=38.5 +lon_0=-97.5 +x_0=0 +y_0=0 +R=6371229 +units=m +no_defs

to

+proj=longlat +R=6371229 +no_defs +units=degrees

A lot of the changes to other files were done to not rely on hard coded numbers and use a more generic approach to determine values. Plus some changes that were not done/updated when doing #674, yeah, mostly things missed in that PR. Plus, support for point-cloud and re-projections into the sphere plot type.

Also, analytics indeed doesn't work on reprojected data. We should address that in a different PR.

Some screenshots.

Sphere

FlatMap

Point clouds

Volume

Blocks

gemini-code-assist

Code Review

This pull request introduces reprojection support across multiple visualization components (including PointCloud, Sphere, SphereBlocks, and CountryBorders) by integrating a remap texture and handling coordinate flipping. The review feedback highlights several critical issues, such as potential runtime crashes and division-by-zero errors when coordinate ranges or shape dimensions are zero. Additionally, it points out a Three.js-specific issue where setting a define to false still defines the macro in GLSL, recommending that these defines be conditionally omitted or deleted instead.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

… from data retrieval

…ssing dimensions in PointCloud component

lazarusA · 2026-07-19T09:27:35Z

Done with all suggestions 👍🏻 !

…rDev into la/fix_reprojections

TheJeran · 2026-07-19T09:59:53Z

Do you have the test cases datasets remote somewhere?

Also, no issues at the moment. But we will have to change the logic for the sphere. Because it doesn't always make sense to show the reprojected data.

Also feels like the point-cloud could be simplified somehow. But since reprojection is basically brand new data. I think what you have done is at the moment the best option I can think of. This also would help free up some data if we just decided to stick with sampling the main texture array for the non-reprojected data as well. Can remove the data textureData global.

I just got two things I will comment in the code

lazarusA · 2026-07-19T10:02:43Z

See issue #690 for the issue and example data.

Also, I do like the deformed reprojection on the sphere.

TheJeran · 2026-07-19T10:01:02Z

+      defines: {
+        REPROJECT: remapTexture ? true : false,
+        FLIP_Y: flipY ? true : false,
+        IS_2D: is2D,


Is this, (and the subsequent shader counterpart) redundant? I don't think ou can have pointcloud with 2D data.

on 2d point cloud data, we should support that. So, I will come back to that and enable it.

TheJeran · 2026-07-19T10:03:50Z

    newAxisDimArrays[yIdx] = yTicks;

    const newAxisDimUnits = [...axisDimUnits];
+    const targetUnits = (crsCheck.oProj as any)?.units || 'degrees';


This is wrong. It can't default to degrees

yes, but it needs something empty then. Because otherwise I think it will failed for some cases. Empty should be fine, ''.

TheJeran · 2026-07-19T10:07:51Z

+    if (plotType === 'point-cloud') {
+        targetWidth = width;
+        targetHeight = height;
+        xTicks = linspace(minX, maxX, targetWidth);
+        yTicks = flipY ? linspace(maxY, minY, targetHeight) : linspace(minY, maxY, targetHeight);
+        data = new Uint16Array(targetWidth * targetHeight * 4);
+
+        const xDiff = Math.abs(maxX - minX);
+        const yDiff = Math.abs(maxY - minY);
+
+        for (let j = 0; j < targetHeight; j++) {
+            const lat = yArray[j];
+            for (let i = 0; i < targetWidth; i++) {
+                const lon = xArray[i];
+                const [px, py] = proj.forward([lon, lat]);
+                const valid = (isFinite(px) && isFinite(py)) ? 1 : 0;
+
+                const u = xDiff > 0 ? (px - minX) / xDiff : 0;
+                const v = yDiff > 0 ? (py - minY) / yDiff : 0;
+
+                const idx = (j * targetWidth + i) * 4;
+                data[idx]     = THREE.DataUtils.toHalfFloat(u);  
+                data[idx + 1] = THREE.DataUtils.toHalfFloat(v);
+                data[idx + 2] = THREE.DataUtils.toHalfFloat(valid);
+            }
+        }
+    } else {
+        targetWidth = Math.ceil(adjustedResolution * aspectRatio);
+        targetHeight = adjustedResolution;
+        xTicks = linspace(minX, maxX, targetWidth);
+        yTicks = flipY ? linspace(maxY, minY, targetHeight) : linspace(minY, maxY, targetHeight);
+
+        // Detect if coordinate axes are descending
+        const isXDescending = xArray.length > 1 ? xArray[0] > xArray[xArray.length - 1] : false;
+        const isYDescending = yArray.length > 1 ? yArray[0] > yArray[yArray.length - 1] : false;
+
+        data = new Uint16Array(targetWidth * targetHeight * 4);
+        const xRangeDiff = xMax - xMin;
+        const yRangeDiff = yMax - yMin;
+        for (let j = 0; j < targetHeight; j++) {
+            for (let i = 0; i < targetWidth; i++) {
+                const [lon, lat, valid] = safeInverse(proj, [xTicks[i], yTicks[j]]);
+                const u = xRangeDiff > 0 ? (isXDescending ? (xMax - lon) / xRangeDiff : (lon - xMin) / xRangeDiff) : 0;
+                const v = yRangeDiff > 0 ? (isYDescending ? (yMax - lat) / yRangeDiff : (lat - yMin) / yRangeDiff) : 0;
+
+                // Check boundary bounds to avoid displaying clamped blocks outside the dataset area
+                const inBounds = lon >= xMin && lon <= xMax && lat >= yMin && lat <= yMax;
+                const validVal = (valid === 1 && inBounds) ? 1 : 0;
+
+                const idx = (j * targetWidth + i) * 4;
+                data[idx]     = THREE.DataUtils.toHalfFloat(u);  
+                data[idx + 1] = THREE.DataUtils.toHalfFloat(v);
+                data[idx + 2] = THREE.DataUtils.toHalfFloat(validVal);
+            }  
+        }


Can you explain the use cases where this is necessary? When does the normal case not work? I'm not convinced this isn't redundant or can't just be trivially undid by flipping the resulting texture.

first part is to do reprojection directly on point data, second on your resampling texture strategy. Or? what else do you mean? Also, keeping track of flip or not flip is necessary.

So the point-cloud doesn't increase in resolution?

point-cloud should do native points. No need for resolution. What we see is what we get.

Hmm... okay. I think I will come back to this. Something in my brain says there's some way to simplify it but I'm to tired to discover if thats true. But otherwise I'm content to merge.

Let's open an issue to keep track of this then.

This reverts commit c07b0c4.

lazarusA · 2026-07-19T10:49:59Z

once you are done with your changes and satisfied let me know so that I can pull and test before merging. Thanks for taking a look so fast.

lazarusA · 2026-07-19T11:51:20Z

Something went south. I will track down what was it, and revert the minimal needed lines.

Edit:
Sorry, it only works reliable if we splat things. I will go back to those.

…ss plot components

…sary definitions

…h improved define handling

…leanup

lazarusA · 2026-07-19T13:03:39Z

fallback to degrees allows for

instead of

I agree that this does not apply to all projections, but if not information is provided about the dest projection units having a default seems like the reasonable thing to do. In any case, we need a proper helper function that infers the right units from the +proj string entry. Will open an issue and independent PR to fix that. #693

TheJeran · 2026-07-19T13:26:26Z

The degrees thing is on the user to properly add the degrees in the proj string. +units=degrees .

But since all other units are massive. You can just add a check that if there are no units it checks if the new axis fits inside the lat/lon domains. if the max/min values are larger than 360 or less than -180 its almost certainly not going to be degrees. unless its a tiny square with lon/lat offsets.

I also thing lon/lat projections have a name in the proj object indicating that it will be degrees.

TheJeran · 2026-07-19T13:27:38Z

+                    const oldTextures = textures;
+                    setTimeout(() => {
+                        oldTextures.forEach((tex) => {
+                            tex.dispose();
+                            if (tex.source) (tex.source as any).data = null;
+                        });
+                    }, 0);
+                    setTextures(null);


Can you reproduce this?

I really don't like adding extra code unless there's a specific reproducible error and I haven't had any issues with this since it's implementation

yes, I can. I did test it. Without it, I will always see a red type error on console, I don't like those. This does work.

…ojectionTexture, user is responsible for it.

fixes all plot types reprojections

164c827

lazarusA requested a review from TheJeran July 18, 2026 21:06

gemini-code-assist Bot reviewed Jul 18, 2026

View reviewed changes

lazarusA added 5 commits July 18, 2026 23:13

?

f7826ba

checks

5894c7a

reproject points not textures

88ef526

remove setStatus calls from GetCurrentArray to decouple state updates…

3c5d82c

… from data retrieval

fix: prevent runtime errors by validating image existence before acce…

90a3c3b

…ssing dimensions in PointCloud component

TheJeran added 2 commits July 19, 2026 11:57

added pnmp global install for gribberish and changed define

e051172

Merge branch 'la/fix_reprojections' of github.com:EarthyScience/ViZar…

34de9d0

…rDev into la/fix_reprojections

TheJeran added 3 commits July 19, 2026 12:11

returned defines to normal

5f4b078

ifdef --> if

c07b0c4

remove degrees

c849410

TheJeran reviewed Jul 19, 2026

View reviewed changes

Revert "ifdef --> if"

6248cd6

This reverts commit c07b0c4.

lazarusA added 4 commits July 19, 2026 14:11

refactor: conditionally set IS_FLAT and REPROJECT shader defines acro…

2de8c18

…ss plot components

conditionally include REPROJECT in DataCube uniforms to avoid unneces…

135cf55

…sary definitions

update projection shader logic and standardize coordinate mapping wit…

66ad216

…h improved define handling

fix: defer texture disposal to prevent race conditions during state c…

afeaec7

…leanup

TheJeran reviewed Jul 19, 2026

View reviewed changes

fix: remove fallback to degrees when target units are undefined in Pr…

930f7d2

…ojectionTexture, user is responsible for it.

lazarusA merged commit f6e2602 into main Jul 19, 2026
6 checks passed

Uh oh!

Conversation

lazarusA commented Jul 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Sphere

FlatMap

Point clouds

Volume

Blocks

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

lazarusA commented Jul 19, 2026

Uh oh!

TheJeran commented Jul 19, 2026

Uh oh!

lazarusA commented Jul 19, 2026

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!

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!

lazarusA commented Jul 19, 2026

Uh oh!

lazarusA commented Jul 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lazarusA commented Jul 19, 2026

Uh oh!

TheJeran commented Jul 19, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

lazarusA commented Jul 18, 2026 •

edited

Loading

lazarusA commented Jul 19, 2026 •

edited

Loading