DOC: Correct the migration guide — v1 plans do not load in v2

[RafaelPalomar]

First docs slice. Both commits fix documentation that actively misleads, rather than documentation that is merely missing.

The migration guide promised an upgrade that does not happen

Docs/migrations/v1-to-v2.md said, verbatim:

The upgrade is seamless: opening a .lrp.fcsv through the normal load path routes it to vtkMRMLResectionPlanStorageNode::ReadFcsv

That path existed during v2 development and was retired with the v1 markups subsystem. As shipped:

• qSlicerLiverResectionsReader.cxx:139 advertises only Liver resection plan (*.lrp.json); anything else is rejected with "Unsupported file extension".
• vtkMRMLResectionPlanStorageNode.h:107 states outright: legacy .lrp.fcsv files no longer load.
• vtkSlicerLiverResectionsLogic::LoadResections errors on .fcsv.

This is the worst class of documentation defect, because it concerns user data and acting on it causes loss. A v1 user reading the old text would upgrade, discard their v1 installation, and discover their resection plans are unopenable — with no converter, no batch tool, and no warning anywhere in the upgrade path.

The guide is rewritten from a resection-plan note into an actual migration guide: an at-a-glance carries-over table, what happens to v1 scenes (core nodes load, every Slicer-Liver node is dropped), the module renames, the scripting-API table for the wrapper/carrier/display split, and a pre-upgrade checklist. The warning now leads with keep a working v1 installation.

A code comment cited a test that does not exist

vtkSlicerLiverResectionsLogic.cxx:359 told a reader hitting the rejection that migration "is covered by vtkMRMLResectionPlanLegacyFcsvMigrationTest". No such test exists anywhere in the tree, and no migration exists for it to cover. Someone debugging a failed legacy load would go looking for reassurance and find nothing.

Verification

Every factual claim was checked against both branches rather than carried forward from the old text or from a draft:

• All five v1 node classes named in the removal table confirmed present on main, absent on preview.
• v1 accessor names confirmed: ResectionMargin / UncertaintyMargin (vtkGetMacro), and the v1 enums Initialization/Deformation/Completed and Flat/Curved.
• v2 state strings confirmed as Init/Planning/Confirmed.
• vtkLiverSegmentsLogic confirmed on main.
• The three v2 storage extensions confirmed from source: .lrp.json, .vta.json, .vsd.json.

One correction made during drafting: the hand-recovery snippet originally used a flat-index SetControlPoint(i, x, y, z). That signature does not exist — the real Python-wrappable entry point is SetControlPoint(row, col, x, y, z), and SetControlGrid(const double*) deliberately does not cross the wrap. Publishing the flat-index version would have handed users broken recovery code at exactly the moment they were trying to rescue data.

One correction to an earlier assumption of mine: SlicerVMTK is not a new v2 dependency — git show main:CMakeLists.txt shows it, SegmentEditorExtraEffects and ExtraMarkups were all already in v1's EXTENSION_DEPENDS. The only genuinely new hard dependency is SlicerLayerDM. The guide says so.

Reviewer's map

• Docs/migrations/v1-to-v2.md — full rewrite. The load-bearing section is ".lrp.fcsv files do not load in v2"; the rest is the surrounding context a v1 user needs.
• vtkSlicerLiverResectionsLogic.cxx — one comment, no logic.

Still to come in this epic

Release notes (Docs/releases/v2.0.0.md), the 6-stage user guide, the developer guide, and lifting Docs/index.md out of its self-declared scaffold state. Kept separate so this correction is not held up behind them.

---

Authored with assistance from Claude (Anthropic Claude Opus 5).