fix(service-analytics): 按聚合种类填充「查询从未报告过的分组」,并补上 compareTo 这一道接缝 (#4708) (#4822)

* fix(service-analytics): fill the empty-group value on every merge seam, by aggregate kind (#4708)

A dataset measure carrying its own `filter` runs as a separate grouped
sub-query merged back by dimension key. A `GROUP BY` over a filtered row set
emits no group at all for a dimension value the filter excludes entirely, so
the measure comes back ABSENT rather than 0, and `computeDerived` treats an
absent operand as unknowable — every ratio over it goes null too. The cell
renders blank: visually identical to "no data for this row", which is the
opposite of what the row means. The bias runs the worst possible way — the rows
that blank are the ones whose numerator matched nothing, i.e. the
worst-performing ones.

`fillEmptyGroups` now fills the identity element by aggregate kind:
count/count_distinct/sum → 0 (measured facts), avg/min/max → null (genuinely
undefined; flattening those would trade this lie for its mirror image). The
kind→identity mapping stays `emptyGroupValueFor` in @objectstack/spec/data, so
the runtime and the authoring-side coherence checks cannot drift. Only cells on
rows that already exist are filled — no group is invented.

The fill previously ran BEFORE the compareTo merge, and that merge appends a row
for every bucket the previous window had and this one does not; every base
measure on those rows — including unfiltered ones — stayed absent, so a lead
source that sold last month and nothing this month rendered as "no data"
instead of 0. It now runs after every merge and covers all base measures plus
their `<measure>__compare` columns.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015Br2xsJsczFsTR9bvbh2Ny

* chore(changeset): 标 minor —— fillEmptyGroups 是新增公共导出 (#4708)

`fillEmptyGroups` 从包根 index.ts 导出,属于新增公共 API,按 #4815
(objectql / AuditAbortSignal)与 #4791(service-automation /
sealNodeTypeVocabulary)确立的同一条 bar,changeset 从 patch 改为 minor,
并在正文点名这个新导出。

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015Br2xsJsczFsTR9bvbh2Ny

---------

Co-authored-by: Claude <noreply@anthropic.com>

Author: os-zhuang

.changeset/analytics-empty-group-fill-compare-seam.md

@@ -0,0 +1,50 @@
+---
+"@objectstack/service-analytics": minor
+---
+
+fix(service-analytics): a measure a query never reported reads 0 for a count/sum on every merge seam (#4708)
+
+A dataset measure carrying its own `filter` runs as a separate grouped
+sub-query and is merged back onto the selected dimensions. A `GROUP BY` over a
+filtered row set emits **no group at all** for a dimension value the filter
+excludes entirely, so the measure comes back **absent**, not `0` — and
+`computeDerived` treats an absent operand as unknowable, so every ratio over it
+goes null too. The cell then renders blank, which is visually identical to "no
+data for this row" and means the opposite.
+
+The bias runs the worst possible way: the rows that blank are the ones whose
+numerator matched nothing — the **worst-performing rows**. A `lead_source` that
+won nothing rendered as "no data" while one that won everything rendered fine.
+
+The empty-group value is now filled **by aggregate kind** into every measure
+column the assembled grid lists but no query reported:
+
+| aggregate | over an excluded group | why |
+|:---|:---|:---|
+| `count`, `count_distinct` | `0` | "how many rows matched" has an exact answer when the answer is none |
+| `sum` | `0` | the identity element of the empty set |
+| `avg`, `min`, `max` | stays `null` | genuinely undefined — there is nothing to average |
+
+Filling all five with `0` would trade this lie for its mirror image, reporting a
+measurement nobody made, so the kinds are judged separately (via
+`emptyGroupValueFor`, shared with the authoring-side coherence checks).
+
+**Only cells are filled, never rows.** A dimension value no query reported at
+all has genuinely no data and stays out of the grid.
+
+**What changes beyond the measure-scoped seam.** The fill previously ran before
+the `compareTo` merge, and that merge *appends* a row for every bucket the
+PREVIOUS window had and this one does not. Every base measure on those rows —
+including unfiltered ones — was absent, so a lead source that sold last month
+and nothing this month rendered as "no data" instead of `0`: the same worst-row
+bias, one merge later. The fill now runs after every merge and covers all base
+measures plus their `<measure>__compare` columns.
+
+Widgets that worked around this with `?? 0` in the consumer or a `coalesce` in
+the measure can drop it; the coercion belongs in the executor, which is the only
+layer that knows which aggregate produced the gap.
+
+**New export.** `fillEmptyGroups(rows, columnAggregates)` is exported from the
+package root beside `mergeByDimensions`, so a host assembling a grid outside
+`DatasetExecutor` can apply the same aggregate-kind rule rather than
+reimplementing it — which is what makes this a `minor` rather than a `patch`.

packages/services/service-analytics/src/__tests__/dataset-empty-group-fill.test.ts

@@ -0,0 +1,273 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+/**
+ * Empty-group fill (#4708, objectui#3136).
+ *
+ * A measure carrying its own `filter` runs as a separate grouped sub-query and
+ * is merged back by dimension key. A `GROUP BY` over a filtered row set emits
+ * NO group for a dimension value the filter excludes entirely, so the measure
+ * comes back ABSENT rather than `0` — and a derived ratio over an absent
+ * operand goes null, so the cell renders blank: visually identical to "no data
+ * for this row", which is the opposite of what the row means.
+ *
+ * The bias is what makes it worth a dedicated suite: the rows that blank are
+ * exactly the WORST-performing ones (nothing matched the numerator's filter),
+ * while a row that matched everything renders fine. A dashboard that hides its
+ * worst rows and shows its best is the least acceptable direction for the
+ * error to run, so `renders the worst row …` below asserts the asymmetry
+ * directly rather than only the individual cells.
+ *
+ * The fill is strictly by aggregate kind — over-filling would trade this lie
+ * for its mirror image (an `avg` of nothing reported as 0 is a measurement
+ * nobody made), so `does NOT fill avg/min/max` pins the other side.
+ */
+
+import { describe, it, expect, vi } from 'vitest';
+import type { IAnalyticsService, AnalyticsQuery, AnalyticsResult } from '@objectstack/spec/contracts';
+import { DatasetSchema } from '@objectstack/spec/ui';
+import { compileDataset } from '../dataset-compiler.js';
+import { DatasetExecutor, fillEmptyGroups } from '../dataset-executor.js';
+
+function fakeService(handler: (q: AnalyticsQuery) => AnalyticsResult): IAnalyticsService {
+  return { query: vi.fn(async (q: AnalyticsQuery) => handler(q)), getMeta: async () => [] };
+}
+
+// ── the issue's reproduction, verbatim (hotcrm#593 / hotcrm#656) ─────────────
+
+const winRate = DatasetSchema.parse({
+  name: 'pipeline', label: 'Pipeline', object: 'opportunity',
+  dimensions: [{ name: 'lead_source', field: 'lead_source', type: 'string' }],
+  measures: [
+    { name: 'won_count', aggregate: 'count', filter: { stage: 'closed_won' } },
+    { name: 'lost_count', aggregate: 'count', filter: { stage: 'closed_lost' } },
+    { name: 'decided_count', aggregate: 'count', filter: { stage: { $in: ['closed_won', 'closed_lost'] } } },
+    { name: 'win_rate', derived: { op: 'ratio', of: ['won_count', 'decided_count'] } },
+  ],
+});
+
+/**
+ * The four lead sources of the issue's table. Each sub-query returns only the
+ * groups its filter left non-empty — which is what a real `GROUP BY` does:
+ *   - `partner` won everything      → absent from the `lost_count` result
+ *   - `cold_call` won nothing       → absent from the `won_count` result
+ */
+const winRateService = fakeService((q) => {
+  switch (q.measures[0]) {
+    case 'won_count':
+      return { rows: [
+        { lead_source: 'content', won_count: 2 },
+        { lead_source: 'referral', won_count: 1 },
+        { lead_source: 'partner', won_count: 1 },
+      ], fields: [] };
+    case 'lost_count':
+      return { rows: [
+        { lead_source: 'content', lost_count: 1 },
+        { lead_source: 'referral', lost_count: 1 },
+        { lead_source: 'cold_call', lost_count: 1 },
+      ], fields: [] };
+    case 'decided_count':
+      return { rows: [
+        { lead_source: 'content', decided_count: 3 },
+        { lead_source: 'referral', decided_count: 2 },
+        { lead_source: 'partner', decided_count: 1 },
+        { lead_source: 'cold_call', decided_count: 1 },
+      ], fields: [] };
+    default:
+      return { rows: [], fields: [] };
+  }
+});
+
+const runWinRate = () =>
+  new DatasetExecutor(winRateService).execute(compileDataset(winRate), {
+    dimensions: ['lead_source'],
+    measures: ['won_count', 'lost_count', 'decided_count', 'win_rate'],
+  });
+
+describe('empty-group fill — a filtered count that matched nothing (#4708)', () => {
+  it('reports 0 (not blank) for the group its filter excluded, so the ratio computes', async () => {
+    const rows = (await runWinRate()).rows;
+    const cold = rows.find((r) => r.lead_source === 'cold_call')!;
+    // cold_call won nothing and lost one. The correct answer is 0%, not blank.
+    expect(cold.won_count).toBe(0);
+    expect(cold.decided_count).toBe(1);
+    expect(cold.win_rate).toBe(0);
+  });
+
+  it('fills the mirror gap too — a source that never lost reads 0 losses', async () => {
+    const rows = (await runWinRate()).rows;
+    expect(rows.find((r) => r.lead_source === 'partner')).toMatchObject({
+      won_count: 1, lost_count: 0, decided_count: 1, win_rate: 1,
+    });
+  });
+
+  it('renders the worst row exactly as legibly as the best one (the asymmetry)', async () => {
+    const rows = (await runWinRate()).rows;
+    // The defect's signature: `partner` (won everything) rendered fine while
+    // `cold_call` (won nothing) rendered blank — the dashboard hid its worst
+    // row. No row may come back with an unmeasured cell.
+    for (const row of rows) {
+      expect(row.win_rate, `win_rate blank for ${row.lead_source}`).not.toBeNull();
+      expect(row.win_rate).toEqual(expect.any(Number));
+      for (const m of ['won_count', 'lost_count', 'decided_count']) {
+        expect(row[m], `${m} blank for ${row.lead_source}`).toEqual(expect.any(Number));
+      }
+    }
+    // …and the worst row is the one the reader must be able to act on.
+    const byRate = [...rows].sort((a, b) => Number(a.win_rate) - Number(b.win_rate));
+    expect(byRate[0]).toMatchObject({ lead_source: 'cold_call', win_rate: 0 });
+    expect(byRate[byRate.length - 1]).toMatchObject({ lead_source: 'partner', win_rate: 1 });
+  });
+
+  it('invents no group — a dimension value no query reported stays out of the grid', async () => {
+    const rows = (await runWinRate()).rows;
+    // Only the four sources some sub-query actually returned. `webinar` has no
+    // opportunities at all, so it is absent from every result and must not be
+    // materialised as a row of zeroes.
+    expect(rows.map((r) => r.lead_source).sort()).toEqual(['cold_call', 'content', 'partner', 'referral']);
+  });
+});
+
+// ── the other side: aggregates with no answer over an empty set ──────────────
+
+const amounts = DatasetSchema.parse({
+  name: 'deals', label: 'Deals', object: 'opportunity',
+  dimensions: [{ name: 'lead_source', field: 'lead_source', type: 'string' }],
+  measures: [
+    { name: 'won_count', aggregate: 'count', filter: { stage: 'closed_won' } },
+    { name: 'won_total', aggregate: 'sum', field: 'amount', filter: { stage: 'closed_won' } },
+    { name: 'won_avg', aggregate: 'avg', field: 'amount', filter: { stage: 'closed_won' } },
+    { name: 'won_min', aggregate: 'min', field: 'amount', filter: { stage: 'closed_won' } },
+    { name: 'won_max', aggregate: 'max', field: 'amount', filter: { stage: 'closed_won' } },
+    { name: 'all_count', aggregate: 'count' },
+  ],
+});
+
+describe('empty-group fill — strictly by aggregate kind (#4708)', () => {
+  it('does NOT fill avg/min/max: nothing to average over an empty group', async () => {
+    const svc = fakeService((q) => {
+      if (q.measures.includes('all_count')) {
+        return { rows: [
+          { lead_source: 'content', all_count: 4 },
+          { lead_source: 'cold_call', all_count: 1 },
+        ], fields: [] };
+      }
+      // every won_* sub-query: cold_call won nothing, so no group for it
+      const m = q.measures[0];
+      return { rows: [{ lead_source: 'content', [m]: 7 }], fields: [] };
+    });
+    const res = await new DatasetExecutor(svc).execute(compileDataset(amounts), {
+      dimensions: ['lead_source'],
+      measures: ['all_count', 'won_count', 'won_total', 'won_avg', 'won_min', 'won_max'],
+    });
+    const cold = res.rows.find((r) => r.lead_source === 'cold_call')!;
+    // Measured facts: no rows matched, so the count is 0 and the sum is 0.
+    expect(cold.won_count).toBe(0);
+    expect(cold.won_total).toBe(0);
+    // Genuinely unknown — filling these would report a measurement nobody made
+    // ("average deal size 0" is a different claim from "no deals").
+    for (const m of ['won_avg', 'won_min', 'won_max']) {
+      expect(cold[m] ?? null, `${m} must stay null`).toBeNull();
+      expect(cold[m], `${m} must not be flattened to 0`).not.toBe(0);
+    }
+    // The group that did have data is untouched.
+    expect(res.rows.find((r) => r.lead_source === 'content')).toMatchObject({
+      won_count: 7, won_total: 7, won_avg: 7, won_min: 7, won_max: 7,
+    });
+  });
+});
+
+// ── the compareTo seam: rows APPENDED by the comparison merge ────────────────
+
+const compareDs = DatasetSchema.parse({
+  name: 'trend', label: 'Trend', object: 'opportunity',
+  dimensions: [
+    { name: 'lead_source', field: 'lead_source', type: 'string' },
+    { name: 'close_date', field: 'close_date', type: 'date' },
+  ],
+  measures: [
+    { name: 'revenue', aggregate: 'sum', field: 'amount' },
+    { name: 'avg_deal', aggregate: 'avg', field: 'amount' },
+    { name: 'won_count', aggregate: 'count', filter: { stage: 'closed_won' } },
+  ],
+});
+
+describe('empty-group fill — buckets the comparison window added (#4708)', () => {
+  it('fills every base measure on a row only the previous period produced', async () => {
+    const svc = fakeService((q) => {
+      const shifted = JSON.stringify(q.timeDimensions ?? []).includes('2025-12');
+      if (shifted) {
+        return { rows: [
+          { lead_source: 'content', revenue: 80, avg_deal: 40, won_count: 2 },
+          { lead_source: 'cold_call', revenue: 10, avg_deal: 10, won_count: 1 },
+        ], fields: [] };
+      }
+      if (q.measures.includes('won_count')) {
+        return { rows: [{ lead_source: 'content', won_count: 3 }], fields: [] };
+      }
+      // this period, `cold_call` has no opportunities at all
+      return { rows: [{ lead_source: 'content', revenue: 100, avg_deal: 50 }], fields: [] };
+    });
+    const res = await new DatasetExecutor(svc).execute(compileDataset(compareDs), {
+      dimensions: ['lead_source'],
+      measures: ['revenue', 'avg_deal', 'won_count'],
+      timeDimensions: [{ dimension: 'close_date', dateRange: ['2026-01-01', '2026-01-31'] }],
+      compareTo: { kind: 'previousPeriod', dimension: 'close_date' },
+    });
+    const cold = res.rows.find((r) => r.lead_source === 'cold_call')!;
+    // The row exists because the comparison window had data for it; "how much
+    // did cold_call sell this period" therefore has an exact answer — none.
+    // Before #4708 the fill ran BEFORE this merge, so all three read blank.
+    expect(cold.revenue).toBe(0);
+    expect(cold.won_count).toBe(0);
+    expect(cold.avg_deal ?? null).toBeNull();
+    // the comparison columns are untouched — they were reported
+    expect(cold).toMatchObject({ revenue__compare: 10, won_count__compare: 1 });
+  });
+
+  it('fills a comparison column the previous window never reported', async () => {
+    const svc = fakeService((q) => {
+      const shifted = JSON.stringify(q.timeDimensions ?? []).includes('2025-12');
+      // the previous period knew nothing of `content`
+      if (shifted) return { rows: [], fields: [] };
+      if (q.measures.includes('won_count')) {
+        return { rows: [{ lead_source: 'content', won_count: 3 }], fields: [] };
+      }
+      return { rows: [{ lead_source: 'content', revenue: 100, avg_deal: 50 }], fields: [] };
+    });
+    const res = await new DatasetExecutor(svc).execute(compileDataset(compareDs), {
+      dimensions: ['lead_source'],
+      measures: ['revenue', 'avg_deal', 'won_count'],
+      timeDimensions: [{ dimension: 'close_date', dateRange: ['2026-01-01', '2026-01-31'] }],
+      compareTo: { kind: 'previousPeriod', dimension: 'close_date' },
+    });
+    expect(res.rows[0]).toMatchObject({
+      revenue: 100, won_count: 3,
+      // "sold nothing last month" — a fact, not a gap
+      revenue__compare: 0, won_count__compare: 0,
+    });
+    expect(res.rows[0].avg_deal__compare ?? null).toBeNull();
+  });
+});
+
+// ── the helper in isolation ─────────────────────────────────────────────────
+
+describe('fillEmptyGroups', () => {
+  it('fills count/count_distinct/sum, leaves avg/min/max and reported values alone', () => {
+    const rows = [
+      { g: 'a', c: 2, d: 1, s: 5, avg: 2.5, min: 1, max: 4 },
+      { g: 'b' },
+    ];
+    fillEmptyGroups(rows, {
+      c: 'count', d: 'count_distinct', s: 'sum', avg: 'avg', min: 'min', max: 'max',
+      unknown: undefined,
+    });
+    expect(rows[0]).toEqual({ g: 'a', c: 2, d: 1, s: 5, avg: 2.5, min: 1, max: 4 });
+    expect(rows[1]).toEqual({ g: 'b', c: 0, d: 0, s: 0 });
+  });
+
+  it('touches only the columns it is given, and never adds rows', () => {
+    const rows = [{ g: 'a' }];
+    expect(fillEmptyGroups(rows, { c: 'count' })).toHaveLength(1);
+    expect(rows[0]).toEqual({ g: 'a', c: 0 });
+  });
+});