feat(spec,cli): validate dataset measure aggregation by field semantics

packages/cli/src/utils/validate-widget-bindings.test.ts

@@ -7,6 +7,7 @@ import {
   WIDGET_MEASURE_UNKNOWN,
   CHART_FIELD_UNKNOWN,
   CHART_CONFIG_MISSING,
+  MEASURE_AGGREGATE_INCOHERENT,
 } from './validate-widget-bindings.js';
 
 /** The downstream repro from issue #1719 — dataset with a count AND a sum
@@ -320,3 +321,62 @@ describe('validateWidgetBindings (table-count-only, issue #1719)', () => {
     expect(validateWidgetBindings({ dashboards: [], datasets: [] })).toHaveLength(0);
   });
 });
+
+describe('validateWidgetBindings (measure-aggregate-incoherent — rate aggregation)', () => {
+  /** A dataset whose `probability` measure aggregates a percent field. */
+  function crmStack(aggregate: string) {
+    return {
+      objects: [{
+        name: 'opportunity',
+        fields: [
+          { name: 'amount', type: 'currency' },
+          { name: 'probability', type: 'percent' },
+          { name: 'stage', type: 'select' },
+        ],
+      }],
+      datasets: [{
+        name: 'opportunity_ds',
+        object: 'opportunity',
+        measures: [
+          { name: 'count', aggregate: 'count' },
+          { name: 'total_amount', aggregate: 'sum', field: 'amount' },
+          { name: 'win_probability', aggregate, field: 'probability' },
+        ],
+        dimensions: [{ name: 'stage', field: 'stage' }],
+      }],
+    };
+  }
+
+  it('warns when a measure SUMs a percentage field', () => {
+    const findings = validateWidgetBindings(crmStack('sum'));
+    expect(findings).toHaveLength(1);
+    expect(findings[0].severity).toBe('warning');
+    expect(findings[0].rule).toBe(MEASURE_AGGREGATE_INCOHERENT);
+    expect(findings[0].where).toContain('opportunity_ds');
+    expect(findings[0].where).toContain('win_probability');
+    expect(findings[0].path).toBe('datasets[0].measures[2]');
+    expect(findings[0].message).toContain('percent field "probability"');
+    expect(findings[0].hint).toMatch(/avg/i);
+  });
+
+  it('is clean when the percentage field is AVG’d', () => {
+    expect(validateWidgetBindings(crmStack('avg'))).toHaveLength(0);
+  });
+
+  it('also flags count_distinct of a percentage field', () => {
+    const findings = validateWidgetBindings(crmStack('count_distinct'));
+    expect(findings).toHaveLength(1);
+    expect(findings[0].rule).toBe(MEASURE_AGGREGATE_INCOHERENT);
+  });
+
+  it('does not flag SUM of a currency/amount field', () => {
+    // total_amount sums `amount` (currency) — additive, perfectly fine.
+    expect(validateWidgetBindings(crmStack('avg')).filter((f) => f.rule === MEASURE_AGGREGATE_INCOHERENT)).toHaveLength(0);
+  });
+
+  it('cannot judge — and never false-positives — without the object field types', () => {
+    const stack = crmStack('sum');
+    delete (stack as { objects?: unknown }).objects;
+    expect(validateWidgetBindings(stack)).toHaveLength(0);
+  });
+});

packages/cli/src/utils/validate-widget-bindings.ts

@@ -1,5 +1,7 @@
 // Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
 
+import { isIncoherentAggregate } from '@objectstack/spec/data';
+
 /**
  * Build-time dashboard widget binding diagnostics (issues #1719, #1721).
  *
@@ -37,6 +39,11 @@
  *   means the author wanted a per-record listing, which is not an
  *   analytics dataset at all (model it as an object-bound ListView,
  *   ADR-0017). Evaluated on the WIDGET's binding, not the dataset.
+ * - `measure-aggregate-incoherent` — a dataset measure aggregates its field
+ *   in a way that produces a meaningless number: today, SUM (or
+ *   `count_distinct`) of a `percent`/rate field, whose total routinely
+ *   exceeds 100%. Rates must AVG. Checked once per dataset (independent of
+ *   any widget) when the bound object's field types are known.
  *
  * Warnings can be deliberately suppressed per widget via
  * `suppressWarnings: ['<rule-id>']`; errors cannot — they describe a
@@ -49,6 +56,7 @@ export const WIDGET_MEASURE_UNKNOWN = 'widget-measure-unknown';
 export const CHART_FIELD_UNKNOWN = 'chart-field-unknown';
 export const CHART_CONFIG_MISSING = 'chart-config-missing';
 export const TABLE_COUNT_ONLY = 'table-count-only';
+export const MEASURE_AGGREGATE_INCOHERENT = 'measure-aggregate-incoherent';
 
 export type WidgetBindingSeverity = 'error' | 'warning';
 
@@ -159,6 +167,50 @@ export function validateWidgetBindings(stack: AnyRec): WidgetBindingFinding[] {
     if (typeof ds.name === 'string') datasets.set(ds.name, ds);
   }
 
+  // ── (0) dataset measures aggregate their field coherently ──
+  // A measure that SUMs a percentage/rate field produces a meaningless total
+  // (it can exceed 100%); rates must AVG. This is a dataset-level defect (it
+  // does not depend on any widget), so it is checked once over every dataset
+  // whose object's field types are known. Advisory — the page still renders.
+  const objectFieldTypes = new Map<string, Map<string, string>>();
+  for (const o of asArray(stack.objects)) {
+    if (typeof o.name !== 'string') continue;
+    const fm = new Map<string, string>();
+    for (const f of asArray(o.fields)) {
+      if (typeof f.name === 'string' && typeof f.type === 'string') fm.set(f.name, f.type);
+    }
+    objectFieldTypes.set(o.name, fm);
+  }
+  const datasetList = asArray(stack.datasets);
+  for (let i = 0; i < datasetList.length; i++) {
+    const ds = datasetList[i];
+    const fieldTypes = typeof ds.object === 'string' ? objectFieldTypes.get(ds.object) : undefined;
+    if (!fieldTypes) continue; // cannot judge without the object's field types
+    const dsMeasures = asArray(ds.measures);
+    for (let k = 0; k < dsMeasures.length; k++) {
+      const m = dsMeasures[k];
+      const field = typeof m.field === 'string' ? m.field : undefined;
+      const aggregate = typeof m.aggregate === 'string' ? m.aggregate : undefined;
+      if (!field || !aggregate) continue; // count(*) and underivable measures are fine
+      const ftype = fieldTypes.get(field);
+      if (ftype && isIncoherentAggregate(aggregate, ftype)) {
+        findings.push({
+          severity: 'warning',
+          rule: MEASURE_AGGREGATE_INCOHERENT,
+          where: `dataset "${typeof ds.name === 'string' ? ds.name : `(dataset ${i})`}" › measure "${typeof m.name === 'string' ? m.name : `(measure ${k})`}"`,
+          path: `datasets[${i}].measures[${k}]`,
+          message:
+            `measure "${m.name}" applies ${aggregate} to ${ftype} field "${field}" — ` +
+            `summed percentages are meaningless (they can exceed 100%).`,
+          hint:
+            `Use aggregate "avg" for percentage/rate fields (or "count" of records). ` +
+            `If a running total is genuinely intended, suppress with: ` +
+            `suppressWarnings: ['${MEASURE_AGGREGATE_INCOHERENT}'] on the measure.`,
+        });
+      }
+    }
+  }
+
   const dashboards = asArray(stack.dashboards);
   for (let i = 0; i < dashboards.length; i++) {
     const dash = dashboards[i];

packages/spec/src/data/aggregation-policy.test.ts

@@ -0,0 +1,50 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+import { describe, it, expect } from 'vitest';
+import { defaultAggregateFor, isIncoherentAggregate, MEASURE_FIELD_TYPES } from './aggregation-policy';
+
+describe('defaultAggregateFor', () => {
+  it('SUMs additive amounts', () => {
+    expect(defaultAggregateFor('currency')).toBe('sum');
+    expect(defaultAggregateFor('number')).toBe('sum');
+  });
+
+  it('AVGs rates (percent)', () => {
+    expect(defaultAggregateFor('percent')).toBe('avg');
+  });
+
+  it('defaults to sum for unknown/undefined types', () => {
+    expect(defaultAggregateFor(undefined)).toBe('sum');
+    expect(defaultAggregateFor('text')).toBe('sum');
+  });
+});
+
+describe('isIncoherentAggregate', () => {
+  it('flags SUM and count_distinct of a percentage', () => {
+    expect(isIncoherentAggregate('sum', 'percent')).toBe(true);
+    expect(isIncoherentAggregate('count_distinct', 'percent')).toBe(true);
+  });
+
+  it('allows AVG / min / max / count of a percentage', () => {
+    expect(isIncoherentAggregate('avg', 'percent')).toBe(false);
+    expect(isIncoherentAggregate('min', 'percent')).toBe(false);
+    expect(isIncoherentAggregate('max', 'percent')).toBe(false);
+    expect(isIncoherentAggregate('count', 'percent')).toBe(false);
+  });
+
+  it('allows SUM of additive amounts', () => {
+    expect(isIncoherentAggregate('sum', 'currency')).toBe(false);
+    expect(isIncoherentAggregate('sum', 'number')).toBe(false);
+  });
+
+  it('never false-positives on an unknown field type', () => {
+    expect(isIncoherentAggregate('sum', undefined)).toBe(false);
+    expect(isIncoherentAggregate('sum', 'text')).toBe(false);
+  });
+});
+
+describe('MEASURE_FIELD_TYPES', () => {
+  it('covers the numeric measure-backing field types', () => {
+    expect([...MEASURE_FIELD_TYPES].sort()).toEqual(['currency', 'number', 'percent']);
+  });
+});

packages/spec/src/data/aggregation-policy.ts

@@ -0,0 +1,34 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+/**
+ * Field → aggregation semantics. The single source of truth for "how should a
+ * numeric field be aggregated into a measure", shared by authoring (dataset
+ * derivation) and validation (build-time coherence checks) so the two cannot
+ * drift.
+ *
+ * The motivating defect: additive amounts (currency/number) SUM correctly, but
+ * a RATE — a `percent` field such as a win-probability or conversion rate — must
+ * AVG. Summing percentages is meaningless: the total routinely exceeds 100%.
+ */
+
+/** Numeric field types that can back a value measure. */
+export const MEASURE_FIELD_TYPES: ReadonlySet<string> = new Set(['number', 'currency', 'percent']);
+
+/**
+ * The aggregation a derived value measure should use for a field of this type:
+ * rates (`percent`) AVG, every other additive amount SUMs.
+ */
+export function defaultAggregateFor(fieldType: string | undefined): 'sum' | 'avg' {
+  return fieldType === 'percent' ? 'avg' : 'sum';
+}
+
+/**
+ * Is applying `aggregate` to a field of `fieldType` semantically incoherent?
+ * True only for cases that produce a meaningless number — today: SUM (or
+ * COUNT_DISTINCT) of a percentage/rate. Returns false when the field type is
+ * unknown (cannot judge) so callers never raise a false positive.
+ */
+export function isIncoherentAggregate(aggregate: string, fieldType: string | undefined): boolean {
+  if (fieldType === 'percent' && (aggregate === 'sum' || aggregate === 'count_distinct')) return true;
+  return false;
+}

packages/spec/src/data/index.ts

@@ -39,6 +39,10 @@ export * from './external-catalog.zod';
 // Analytics Protocol (Semantic Layer)
 export * from './analytics.zod';
 
+// Field → aggregation semantics (rates AVG, amounts SUM) — shared by authoring
+// and build-time coherence validation.
+export * from './aggregation-policy';
+
 // Feed & Activity Protocol
 export * from './feed.zod';