fix: tributaries now end at the confluence hex, not one hex before it

[countercheck]

What changed

_split_at_confluences in hydrology.py was trimming lower-flow tributaries with path[:cut] where cut = i — the index of the first claimed (confluence) hex. This stopped the tributary at path[i-1], one hex before the confluence, so the tributary's endpoint was never on the main trunk's polyline. The result was tributaries that appeared to dead-end on flat inland terrain.

Root cause: off-by-one in the cut index.

Fix: cut = i + 1 — the confluence hex is now included as the tributary's final point. It's already part of the trunk path, so the two SVG polylines share a vertex and visually connect. The hex is double-drawn in SVG but invisible at normal stroke widths.

Scale of impact on the sample world (128×128): 60 out of 115 rivers were ending on interior flat-land hexes before this fix.

Files changed

File	Change
worldgen/stages/hydrology.py	cut = i → cut = i + 1; updated docstring
tests/test_hydrology.py	Updated test_no_shared_hexes_between_rivers to permit confluence hexes (tributary endpoints) to appear in two rivers — that sharing is now intentional

Reviewer notes

• test_rivers_reach_ocean passes without logic changes. With the fix, a tributary mouth IS the confluence hex; the test's at_confluence check correctly identifies the next downstream trunk hex as a qualifying neighbor.
• test_no_shared_hexes_between_rivers now excludes tributary endpoints from the "shared hex" check. Any hex shared between rivers that is not a tributary endpoint still fails the assertion, so genuine trunk duplication is still caught.
• The yaml import error in test_cli.py / test_config.py was a pre-existing venv gap (pyyaml not installed); installing it unblocked the pre-commit hook. All 210 tests pass.

[Copilot]

Pull request overview

This PR fixes an off-by-one error in hydrology river splitting so tributaries terminate on the confluence hex (shared with the higher-flow trunk), preventing tributaries from visually “dead-ending” one hex short in rendered outputs.

Changes:

• Include the confluence hex when trimming tributary paths in _split_at_confluences (cut = i + 1).
• Update hydrology docstring/comments to reflect confluence-inclusive trimming semantics.
• Adjust test_no_shared_hexes_between_rivers to allow intentional sharing at confluence endpoints.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
worldgen/stages/hydrology.py	Fixes tributary trimming to include the confluence hex so tributaries visually connect to trunks.
tests/test_hydrology.py	Updates shared-hex assertion logic to permit confluence endpoint sharing by design.

Comments suppressed due to low confidence (1)

worldgen/stages/hydrology.py:751

• The comment about recalculating flow_volume refers to the “last exclusive land hex”, but trimmed paths can now intentionally include the confluence hex (a land hex that is shared with the trunk). Please update this comment to match the new trimming semantics (e.g., last land hex in the trimmed path).

                cut = i + 1  # include confluence hex so tributary visually meets the trunk
                break
        trimmed = path[:cut]
        if len(trimmed) >= 2:
            # Recalculate flow_volume from the last exclusive land hex in the trimmed path.

---

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

[countercheck]

@copilot apply changes based on the comments in this thread

[Copilot]

@copilot apply changes based on the comments in this thread

Applied the review-thread updates in commit 97dca06: _split_at_confluences docs/comments now explicitly state that only intersecting tributaries are trimmed to include the confluence, while non-intersecting trunk rivers keep full paths. No UI changes in this PR, so there’s no UI screenshot to provide.