Group E vscode halves: Signal dropdowns, Alignment editing, mpt paste and shared-name ids, on core v1.40.0

package.json

@@ -236,7 +236,7 @@
   },
   "dependencies": {
     "@lit/context": "^1.1.6",
-    "data-explorer-core": "github:mathworks/data-explorer-core#v1.39.0",
+    "data-explorer-core": "github:mathworks/data-explorer-core#v1.40.0",
     "fast-xml-parser": "^5.8.0",
     "fflate": "^0.8.3",
     "highlight.js": "^11.11.1",

src/host/SlddModel.ts

@@ -232,16 +232,27 @@ export function findNode(uriString: string, nodeId: string): any | null {
   // model in DataModel under a prefixed srcId but NOT in this module's `cache`,
   // so gating on `cache.get(uriString)` here would wrongly drop their selections
   // (the Property Inspector would never render). Try the registry unconditionally.
-  // PRECONDITION (untested) for the `: null` arm: the pinned data-explorer-core
-  // always exports findNodeById on DataModel. The feature-detect exists because
-  // core is a git-pinned dependency bumped independently of this repo, so an older
-  // pin must degrade to the cache fallback below rather than throw on every
-  // selection; the `try` covers the same risk for a throwing implementation.
+  // The pinned data-explorer-core always exports findNodeById on DataModel. The
+  // feature-detect exists because core is a git-pinned dependency bumped
+  // independently of this repo, so an older pin must degrade to the cache fallback
+  // below rather than throw on every selection (test/findNodeSharedName.test.ts);
+  // the `try` covers the same risk for a throwing implementation.
+  //
+  // A null from the registry is FINAL when the registry holds this uri's source. Core
+  // answers null for an id that names two nodes (F1-50: two same-named entries in one
+  // section, and every row inside them), because it cannot tell which one a row meant.
+  // Every adder here registers under the uri and caches the very tree it registered, so
+  // flattening the cache would find the FIRST twin, the wrong-entry pick core refuses,
+  // and a cell edit, a child delete, a copy, a drag and the Property Inspector all act
+  // on what this returns. So the cache answers only what the registry cannot: an old
+  // pin without findNodeById, a lookup that threw, or a source it no longer holds.
   try {
-    const viaRegistry = (DataModel as any).findNodeById
-      ? (DataModel as any).findNodeById(nodeId)
-      : null;
-    if (viaRegistry) return viaRegistry;
+    const session = DataModel as any;
+    if (typeof session.findNodeById === 'function') {
+      const viaRegistry = session.findNodeById(nodeId);
+      if (viaRegistry) return viaRegistry;
+      if (session.hasDataSource?.(uriString)) return null;
+    }
   } catch {
     /* fall through */
   }

src/host/parseWarnings.ts

@@ -33,6 +33,17 @@ import type { ParseWarning } from 'data-explorer-core';
 /** The `code`s that describe the whole source rather than one piece of it. */
 const SOURCE_LEVEL = new Set(['source-empty', 'source-unreadable']);
 
+/**
+ * The `code` core's session adds at open for siblings that share a name, and so one node
+ * id (F1-50). Nothing was lost: both are read and both are shown, but neither can be edited
+ * or deleted by id until the file names them apart. So it is never counted as a part that
+ * could not be read, and never a reason to refuse the file.
+ */
+const SHARED_NAME = 'duplicate-id';
+
+/** The headline when the only thing to say is that some names are shared. */
+const SHARED_NAME_HEADLINE = 'Some items in this file share a name, so they cannot be edited or deleted here.';
+
 /**
  * How many individual messages the banner spells out. A `.mat` with a truncated
  * record tail can report a warning per variable, and a banner tall enough to hold
@@ -91,8 +102,10 @@ export function sourceWarnings(node: unknown): ParseWarning[] {
  *
  * The headline is a source-level warning's own message when there is one, because
  * core writes those as a whole sentence about the whole file ("…so this project
- * reads as empty") and that is the larger fact; otherwise it counts the parts. The
- * details are core's messages verbatim — each already names what it lost, so the
+ * reads as empty") and that is the larger fact; otherwise it counts the parts that
+ * could not be read. A shared name (`duplicate-id`) is not one of those: it gets a
+ * headline of its own when it is all there is, and otherwise only a line below. The
+ * details are core's messages verbatim — each already names what it is about, so the
  * `part` field is used to tell two warnings apart here and never rendered.
  */
 export function warningBanner(warnings: ParseWarning[] | undefined): WarningBanner | undefined {
@@ -115,11 +128,14 @@ export function warningBanner(warnings: ParseWarning[] | undefined): WarningBann
 
   const sourceLevel = unique.find((w) => SOURCE_LEVEL.has(w.code));
   const rest = unique.filter((w) => w !== sourceLevel);
+  const lost = rest.filter((w) => w.code !== SHARED_NAME).length;
   const headline = sourceLevel
     ? sourceLevel.message
-    : rest.length === 1
-      ? 'One part of this file could not be read, so what you see below is incomplete.'
-      : `${rest.length} parts of this file could not be read, so what you see below is incomplete.`;
+    : lost === 0
+      ? SHARED_NAME_HEADLINE
+      : lost === 1
+        ? 'One part of this file could not be read, so what you see below is incomplete.'
+        : `${lost} parts of this file could not be read, so what you see below is incomplete.`;
 
   const details = rest.slice(0, MAX_BANNER_DETAILS).map((w) => w.message);
   const hidden = rest.length - details.length;

src/host/rowBuilder.ts

@@ -36,10 +36,13 @@ export const COLUMN_LABELS: Record<string, string> = {
 // columns (dimensions/complexity/dimensionsMode → Data Object;
 // storageClass/headerFile/alignment → Code Generation) all get their picker
 // header here. Read-only for most classes, but not by definition: on a
-// Simulink.BusElement, complexity and dimensionsMode emit an editable select over
-// MATLAB's own enum, which the generic cell path renders and edits like any other
-// object cell. Labels still come from the schema (schemaColumnLabels), which
-// remains the single source of truth for a column's display name.
+// Simulink.BusElement, a Simulink.ValueType and a Simulink.Signal, complexity and
+// dimensionsMode emit an editable select over MATLAB's own enum for that class,
+// and on a Simulink.Parameter or Simulink.Signal whose StorageClass is not 'Auto',
+// alignment emits an editable text cell. The generic cell path renders and edits
+// each like any other object cell.
+// Labels still come from the schema (schemaColumnLabels), which remains the
+// single source of truth for a column's display name.
 export const COLUMN_GROUPS: Record<string, string> = {
   Min: 'Data Object', Max: 'Data Object', Unit: 'Data Object',
   dimensions: 'Data Object', complexity: 'Data Object', dimensionsMode: 'Data Object',

src/host/sectionRules.ts

@@ -4,9 +4,17 @@
 // needs to PREDICT a drop locally (dropDecision) without a host round-trip on
 // every dragover: the section name, its human label, whether it is derived
 // (architectural), and the classes it accepts. It is posted to each webview
-// alongside its rows, and mirrors SectionNode's allow-list + the section→
-// isderived mapping so the webview's prediction matches what the host paste
-// would actually do.
+// alongside its rows, and mirrors the section→isderived mapping and the classes
+// SectionNode.allowsType admits, so the webview's prediction matches what the
+// host paste would actually do.
+//
+// `allowedTypes` feeds only that prediction: the drop cursor and the Paste item.
+// It is what a paste ADMITS (getAdmittedTypes), not what may be CREATED
+// (getAllowedTypes). The two differ by the registered subclasses: a paste or drop
+// of an mpt.Parameter or mpt.Signal the dictionary already holds lands in Design
+// Data, but none can be created there, so the Add gallery keeps reading
+// getAllowedTypes. A core without getAdmittedTypes admits exactly its creation
+// list, which is the fallback.
 import { getSectionMetadata } from 'data-explorer-core';
 
 export interface SectionRule {
@@ -22,6 +30,11 @@ export function sectionRules(model: any): SectionRule[] {
     sectionName: s.name,
     sectionLabel: s.displayName ?? s.name,
     isDerived: getSectionMetadata(s.name).isderived === '1',
-    allowedTypes: typeof s.getAllowedTypes === 'function' ? s.getAllowedTypes() : [],
+    allowedTypes:
+      typeof s.getAdmittedTypes === 'function'
+        ? s.getAdmittedTypes()
+        : typeof s.getAllowedTypes === 'function'
+          ? s.getAllowedTypes()
+          : [],
   }));
 }