Fix S-57 area spike artifacts from merged interior holes

[philliphoff]

Fixes #470

Problem

Viewing S-57 (NOAA ENC) datasets showed large straight "spike" segments cutting across otherwise-smooth sea/land area curves — most visible in line/outline rendering (area fills hid them as zero-width slivers). Reproduced with NOAA US3WA01M (Strait of Juan de Fuca).

Root cause

This repo renders S-57 by translating each cell in-memory to an S101Document (S57ToS101Translator) and reusing the S-101 pipeline. In TranslateAreaSpatial, every consecutive interior (USAG=2) boundary edge of an area was grouped into a single interior composite curve. S-57 lists all of an area's interior edges consecutively regardless of which hole they belong to, so multi-hole areas had all their holes merged into one ring. Flattening that ring to coordinates jumped from the end of one hole to the start of the next, producing the long "spike" segments.

Empirically, on US3WA01M: 71 of 134 interior composites merged more than one disjoint hole (e.g. DepthArea merging 4 holes with a 1.32° jump; QualityOfBathymetricData merging 151 chains into one ring).

Fix

TranslateAreaSpatial now chains area boundary edges into contiguous rings by shared begin/end node identity (reversing individual edges as needed), emitting one composite/ring per hole. Applied to exterior edges too for ordering robustness. Two small helpers were added: ChainEdgesIntoRings and EdgeNode.

Verification

• Node-identity chaining splits the 134 merged interior composites into 1864 rings, all cleanly closed (0 open).
• Worst interior-ring jump 1.32° → eliminated; remaining long segments are all legitimate rectilinear meta-coverage boundaries (ZOC boxes, TSS lanes, magnetic-variation/administration zones).
• Confirmed visually in the viewer (smooth depth contours after the fix).
• Two new xunit regression tests (two-hole area → 2 separate interior rings; no cross-hole spike). Confirmed they fail without the fix (spike = 8.485E-05°, the predicted jump) and pass with it.
• All 282 S-57 tests pass; dotnet format whitespace / style IDE0005 clean.

Files changed

• src/EncDotNet.S100.Datasets.S57/S57ToS101Translator.cs — the fix
• tests/EncDotNet.S100.Datasets.S57.Tests/S57ToS101TranslatorTests.cs — regression tests
• src/EncDotNet.S100.Datasets.S57/README.md — documented ring-chaining behavior

[github-actions]

Performance Gate

✅ PASSED — no regressions.

Threshold: 10.0%, MAD multiplier (k): 3.0, retry-zone mult: 2.0×

Scenario summary

Scenario	Status	Δ median (%)	z (Δ/MAD)	Base median (ms)	Samples (b/c)
exchange-set-open	✅ pass	-0.2	-0.05	1.59	5/5
s101-portray-cold	✅ pass	-0.9	-0.18	0.27	5/5
s101-portray-warm	✅ pass	-6.8	-0.20	0.24	5/5
s101-real-cold	✅ pass	-0.6	-0.09	0.28	5/5
s101-real-warm	✅ pass	+21.4	+0.91	0.25	5/5
s101-render-warm	✅ pass	-11.0	-1.13	0.23	5/5
s102-coverage	✅ pass	-5.4	-0.82	0.17	5/5
s102-coverage-open	✅ pass	0.0	0.00	2.83	20/20
s102-coverage-render-large	✅ pass	+6.4	+0.75	0.22	5/5
s102-real-warm	✅ pass	+178.8	+5.33	0.33	5/5
s111-real-warm	✅ pass	-8.7	-1.44	0.33	5/5
s124-vector	✅ pass	-8.9	-1.33	0.21	5/5
s201-vector	✅ pass	-19.2	-2.17	0.23	5/5

exchange-set-open

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	1.59	1.59
Baseline MAD (ms)	0.07	—
Δ median	—	-0.2%
z (Δ/MAD)	—	-0.05

Spans (sum of all iterations)

Span	Baseline (ms)	Candidate (ms)	Delta	Status
s100.asset.read	4.35	4.32	-0.8%	▫️
s100.exchangeset.parse	46.42	46.67	+0.5%	▫️

Metrics

Metric	Baseline	Candidate	Delta	Status
s100.asset.read.duration	0.16	0.20	+24.5%	❌

s101-portray-cold

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.27	0.26
Baseline MAD (ms)	0.01	—
Δ median	—	-0.9%
z (Δ/MAD)	—	-0.18

s101-portray-warm

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.24	0.23
Baseline MAD (ms)	0.08	—
Δ median	—	-6.8%
z (Δ/MAD)	—	-0.20

s101-real-cold

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.28	0.28
Baseline MAD (ms)	0.02	—
Δ median	—	-0.6%
z (Δ/MAD)	—	-0.09

s101-real-warm

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.25	0.31
Baseline MAD (ms)	0.06	—
Δ median	—	+21.4%
z (Δ/MAD)	—	+0.91

s101-render-warm

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.23	0.20
Baseline MAD (ms)	0.02	—
Δ median	—	-11.0%
z (Δ/MAD)	—	-1.13

s102-coverage

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.17	0.16
Baseline MAD (ms)	0.01	—
Δ median	—	-5.4%
z (Δ/MAD)	—	-0.82

s102-coverage-open

Iteration statistics

Stat	Baseline	Candidate
Samples	20	20
Median (ms)	2.83	2.83
Baseline MAD (ms)	0.06	—
Δ median	—	0.0%
z (Δ/MAD)	—	0.00

Spans (sum of all iterations)

Span	Baseline (ms)	Candidate (ms)	Delta	Status
s100.dataset.open	450.95	447.83	-0.7%	▫️
s100.hdf5.dataset.read	162.91	162.24	-0.4%	▫️
s100.hdf5.file.open	18.57	18.39	-1.0%	▫️
s100.hdf5.open	18.47	18.53	+0.3%	▫️

Metrics

Metric	Baseline	Candidate	Delta	Status
s100.hdf5.read.bytes	36456.00	36456.00	+0.0%	▫️
s100.hdf5.read.duration	26.43	26.35	-0.3%	▫️
s100.hdf5.read.duration	31.42	31.30	-0.4%	▫️
s100.hdf5.read.duration	11.18	11.13	-0.4%	▫️

s102-coverage-render-large

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.22	0.23
Baseline MAD (ms)	0.02	—
Δ median	—	+6.4%
z (Δ/MAD)	—	+0.75

s102-real-warm

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.33	0.91
Baseline MAD (ms)	0.11	—
Δ median	—	+178.8%
z (Δ/MAD)	—	+5.33

s111-real-warm

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.33	0.30
Baseline MAD (ms)	0.02	—
Δ median	—	-8.7%
z (Δ/MAD)	—	-1.44

s124-vector

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.21	0.19
Baseline MAD (ms)	0.01	—
Δ median	—	-8.9%
z (Δ/MAD)	—	-1.33

s201-vector

Iteration statistics

Stat	Baseline	Candidate
Samples	5	5
Median (ms)	0.23	0.19
Baseline MAD (ms)	0.02	—
Δ median	—	-19.2%
z (Δ/MAD)	—	-2.17

---

Generated by EncDotNet.S100.PerfReport gate command

[Copilot]

Pull request overview

Fixes S-57 area “spike” artifacts caused by incorrectly merging multiple interior holes into a single boundary when translating S-57 areas into S-101 surface topology.

Changes:

• Updates TranslateAreaSpatial to chain boundary edges into contiguous rings by shared begin/end node identity, emitting one ring per hole (and improving exterior ordering robustness).
• Adds regression tests covering a two-hole area translation and asserting the absence of cross-hole “spike” segments.
• Documents the ring-chaining behavior and rationale in the S-57 dataset README.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
src/EncDotNet.S100.Datasets.S57/S57ToS101Translator.cs	Rebuilds area boundary rings by node contiguity and introduces helpers for edge chaining and node lookup.
tests/EncDotNet.S100.Datasets.S57.Tests/S57ToS101TranslatorTests.cs	Adds regression tests to ensure multi-hole areas produce separate interior rings and no cross-hole spikes.
src/EncDotNet.S100.Datasets.S57/README.md	Documents why and how area rings are reassembled to avoid spike artifacts.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.