test: add antimeridian splitting tests for low npoints values by thomas-hervey · Pull Request #76 · springmeyer/arc.js

thomas-hervey · 2026-03-30T01:38:07Z

Branch to addressing partial regression #75. While testing v1.0.0-beta.1 in @turf/great-circle, I found that antimeridian splitting fails for some Pacific routes when npoints is low.

Related: #69, #71

This first commit adds tests covering three antimeridian-crossing routes (Tokyo→LAX, Auckland→LAX, Shanghai→SFO) at both npoints=100 and npoints=10. All three pass at 100 points. At 10 points, Tokyo→LAX and Shanghai→SFO return an unsplit LineString instead of a MultiLineString.

The second commit adjusts the splitting logic. It moves away from the GDAL heuristic towards an analytical approach using interpolation. This should be reviewed for precision.

The third commit addresses a minor test assertion suggestion by Copilot.

The fourth commit primarily replaces stale snapshot coordinates, and adds an additional southern-to-southern hemisphere antimeridian test.

The fifth commit cleans up GDAL attribution & license documentation since the heuristic was replaced in the second commit.

Copilot

Pull request overview

Adds Jest/TypeScript regression tests to cover antimeridian splitting behavior in GreatCircle.Arc() for several Pacific routes, with special focus on low npoints values implicated in issue #75.

Changes:

Add a new test/antimeridian.test.ts suite covering three antimeridian-crossing routes at npoints=100 and npoints=10.
Add a non-crossing route assertion (Seattle → DC) to confirm non-dateline behavior remains stable.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-30T01:40:34Z

test/antimeridian.test.ts

+  describe('with npoints=100', () => {
+    for (const { name, start, end } of PACIFIC_ROUTES) {
+      test(`${name} produces a split MultiLineString`, () => {
+        const result = new GreatCircle(start, end).Arc(100, { offset: 10 }).json();
+        expect(result.geometry.type).toBe('MultiLineString');
+        assertSplitAtAntimeridian((result.geometry as MultiLineString).coordinates);
+      });
+    }
+  });
+


This file duplicates the existing Tokyo/Auckland/Shanghai antimeridian split assertions already present in test/great-circle.test.ts for npoints=100. Consider keeping the new low-npoints coverage but avoid repeating the npoints=100 suite (or move these cases into the existing "Dateline crossing" block) to reduce duplicated fixtures and maintenance overhead.

Suggested change

describe('with npoints=100', () => {

for (const { name, start, end } of PACIFIC_ROUTES) {

test(`${name} produces a split MultiLineString`, () => {

const result = new GreatCircle(start, end).Arc(100, { offset: 10 }).json();

expect(result.geometry.type).toBe('MultiLineString');

assertSplitAtAntimeridian((result.geometry as MultiLineString).coordinates);

});

}

});

This is low stakes but worth considering as a minor clean-up task in the future. Ignoring for now.

Copilot · 2026-03-30T01:40:35Z

test/antimeridian.test.ts

+      test(`${name} splits correctly`, () => {
+        const result = new GreatCircle(start, end).Arc(10, { offset: 10 }).json();
+        expect(result.geometry.type).toBe('MultiLineString');
+        assertSplitAtAntimeridian((result.geometry as MultiLineString).coordinates);
+      });


As written, the npoints=10 tests will fail on the current GreatCircle.Arc implementation (it often returns an unsplit LineString for Tokyo→LAX and Shanghai→SFO because the split detector only triggers for longitude diffs > 360 - offset). If this PR is intended to be tests-only, mark the known-bad cases with test.failing/test.todo (or skip with a link to #75) to keep CI green; otherwise include the corresponding implementation fix in this PR.

Suggested change

test(`${name} splits correctly`, () => {

const result = new GreatCircle(start, end).Arc(10, { offset: 10 }).json();

expect(result.geometry.type).toBe('MultiLineString');

assertSplitAtAntimeridian((result.geometry as MultiLineString).coordinates);

});

if (name === 'Tokyo → LAX' || name === 'Shanghai → SFO') {

test.skip(`${name} splits correctly (known issue, see #75)`, () => {

const result = new GreatCircle(start, end).Arc(10, { offset: 10 }).json();

expect(result.geometry.type).toBe('MultiLineString');

assertSplitAtAntimeridian((result.geometry as MultiLineString).coordinates);

});

} else {

test(`${name} splits correctly`, () => {

const result = new GreatCircle(start, end).Arc(10, { offset: 10 }).json();

expect(result.geometry.type).toBe('MultiLineString');

assertSplitAtAntimeridian((result.geometry as MultiLineString).coordinates);

});

}

This was addressed in commit cc9037e. The correct tests are passing.

Copilot · 2026-03-30T01:40:35Z

test/antimeridian.test.ts

+function assertSplitAtAntimeridian(coords: number[][][]) {
+  const seg0 = coords[0];
+  const seg1 = coords[1];
+
+  expect(seg0).toBeDefined();
+  expect(seg1).toBeDefined();
+
+  if (!seg0 || !seg1) return; // narrow for TS
+
+  const lastOfFirst = seg0[seg0.length - 1];
+  const firstOfSecond = seg1[0];
+
+  expect(lastOfFirst).toBeDefined();
+  expect(firstOfSecond).toBeDefined();
+
+  if (!lastOfFirst || !firstOfSecond) return; // narrow for TS
+
+  // Both sides of the split must be at ±180
+  expect(Math.abs(lastOfFirst[0] ?? NaN)).toBeCloseTo(180, 1);
+  expect(Math.abs(firstOfSecond[0] ?? NaN)).toBeCloseTo(180, 1);
+
+  // Latitudes must match — no gap at the antimeridian
+  expect(lastOfFirst[1] ?? NaN).toBeCloseTo(firstOfSecond[1] ?? NaN, 3);
+}


assertSplitAtAntimeridian only checks abs(lon) ≈ 180, so it would still pass if both segment endpoints use the same sign (e.g. 180/180) or if the geometry contains >2 segments with later bad splits. To prevent false positives, assert coords.length === 2, both segments are non-empty, and that the split longitudes are exactly 180 and -180 (opposite signs) for the adjacent endpoints (matching the implementation and existing tests in test/great-circle.test.ts).

This is a minor concern, but it is addressed in a follow-up commit d3598f2.

thomas-hervey · 2026-03-30T04:34:25Z

Those copilot reviews ^ were not supposed to be automatic, but I will look through them and address as needed.

…n splitting - Remove bHasBigDiff / dfMaxSmallDiffLong / dfDateLineOffset heuristic - Bisect for exact crossing fraction via interpolate() (50 iterations) - Insert [±180, lat*] boundary points; npoints ≤ 2 keeps current behavior - Fixes issue springmeyer#75: low npoints (e.g. 10) no longer skips the split - Tighten test assertions: SPLIT_NPOINTS constant, directional ±180 checks

Addresses Copilot review comment — adds coords.length === 2 check to assertSplitAtAntimeridian to guard against false positives from 3+ segment splits.

- Routes 2-5 (antimeridian crossers): replace stale coordinate snapshots with semantic assertions (Feature type, properties pass-through, WKT contains two LINESTRING parts). Splitting correctness is owned by antimeridian.test.ts. - Southern hemisphere filter: switch to coordinate comparison (start.y < 0 || end.y < 0) and flatten MultiLineString coordinates before .some() to fix number[][][] vs number[][] traversal bug. - Add south-to-south antimeridian crossing coverage: Sydney ↔ Buenos Aires at npoints=10 and 100 in both directions. - Reformat antimeridian.test.ts to consistent 2-space indentation. - Add geographic place names to all routes for maintainer clarity.

thomas-hervey · 2026-04-09T21:58:19Z

@jgravois and @springmeyer, this PR was previously approved but some tests had failed because of the GDAL heuristic. I've updated the op to summarize the changes including moving away from this GDAL heuristic.

As a result, I think it needs more review and manual testing before I pull it out of draft. If it's hard to decipher these changes let me know and I can start over with more atomic commits in a new draft PR.

For context, I have planned two small PRs as fast-follows:

A docs/type cleanup: Fix stale DEVELOPING.md scripts and update the ArcOptions.offset JSDoc to note it's a no-op under the new analytical splitting approach;
Additional edge-case tests: Cover npoints=2/3 on antimeridian routes, prime meridian crossings, and equatorial routes. This should be a sufficient coverage matrix before we cut a v1.0.0 release and then close turf/turf#3030.

The GDAL-ported dateline splitting heuristic was removed when the analytical bisection approach was introduced. No remaining code derives from GDAL, so delete GDAL-LICENSE.md, remove it from the package.json files list, drop the file-level attribution block in great-circle.ts, and remove the GDAL references from README.md.

springmeyer · 2026-04-10T03:25:47Z

Thanks so much for the update @thomas-hervey - I'll take a closer look tomorrow.

test: add antimeridian splitting tests for low npoints values

5f58c4f

Copilot AI review requested due to automatic review settings March 30, 2026 01:38

thomas-hervey marked this pull request as draft March 30, 2026 01:38

Copilot started reviewing on behalf of thomas-hervey March 30, 2026 01:38 View session

thomas-hervey mentioned this pull request Mar 30, 2026

GreatCircle.Arc antimeridian split fails at low npoints for mid-arc crossings #75

Open

Copilot AI reviewed Mar 30, 2026

View reviewed changes

springmeyer approved these changes Mar 30, 2026

View reviewed changes

thomas-hervey added 3 commits April 9, 2026 11:16

test(antimeridian): assert exactly 2 segments at antimeridian split

d3598f2

Addresses Copilot review comment — adds coords.length === 2 check to assertSplitAtAntimeridian to guard against false positives from 3+ segment splits.

thomas-hervey requested a review from springmeyer April 9, 2026 21:58

thomas-hervey mentioned this pull request Apr 9, 2026

docs: remove deprecated offset option and GDAL attribution #77

Open

2 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

test: add antimeridian splitting tests for low npoints values#76

test: add antimeridian splitting tests for low npoints values#76
thomas-hervey wants to merge 5 commits intospringmeyer:chore/esmfrom
thomas-hervey:test/antimeridian-low-npoints

thomas-hervey commented Mar 30, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 30, 2026

Uh oh!

thomas-hervey Apr 9, 2026

Uh oh!

Copilot AI Mar 30, 2026

Uh oh!

thomas-hervey Apr 9, 2026

Uh oh!

Copilot AI Mar 30, 2026

Uh oh!

thomas-hervey Apr 9, 2026

Uh oh!

thomas-hervey commented Mar 30, 2026

Uh oh!

thomas-hervey commented Apr 9, 2026

Uh oh!

springmeyer commented Apr 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

thomas-hervey commented Mar 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Mar 30, 2026

Choose a reason for hiding this comment

Uh oh!

thomas-hervey Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 30, 2026

Choose a reason for hiding this comment

Uh oh!

thomas-hervey Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 30, 2026

Choose a reason for hiding this comment

Uh oh!

thomas-hervey Apr 9, 2026

Choose a reason for hiding this comment

Uh oh!

thomas-hervey commented Mar 30, 2026

Uh oh!

thomas-hervey commented Apr 9, 2026

Uh oh!

springmeyer commented Apr 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

thomas-hervey commented Mar 30, 2026 •

edited

Loading