fix(driver-memory): render the analytics echo as the query it describes (#7117)

.changeset/memory-analytics-echo-like-wildcards.md

@@ -0,0 +1,60 @@
+---
+"@objectstack/driver-memory": patch
+---
+
+fix(driver-memory): the analytics echo renders the query it describes (#7117)
+
+`MemoryAnalyticsService` has two exits for one normalized filter tree, and they
+disagreed about what the LIKE family MEANS. `query()` builds a real containment
+pattern; `generateSql()` emitted the comparand as a bare literal with no
+wildcard anywhere, so `{name: {$contains: 'acme'}}` echoed
+
+```sql
+WHERE name LIKE 'acme'
+```
+
+— an **equality** — beside a chart drawn from every row *containing* `acme`.
+The echo's only job is reproducing execution, so an author who ran it to debug
+the chart got a **narrower** row set and read the filter as broken.
+`$notContains` mirrored it through `NOT LIKE`.
+
+**What the echo emits now.** The `$contains` family renders `GLOB '*v*'` /
+`NOT GLOB`, and `$icontains` renders `lower(col) GLOB lower('*v*')`. `GLOB`
+rather than `LIKE` because this exit emits SQLite-shaped SQL and SQLite's `LIKE`
+folds ASCII case unconditionally: #4706 Q2 = A rules the `$contains` family
+case-**sensitive**, and #7723 put this package's execution faces on that answer,
+so a `LIKE` echo would have contradicted execution on a second axis the moment
+the missing wildcards were added. The two halves are one fix because `GLOB`
+speaks a different pattern language from `LIKE` — choosing the construct and
+rendering the wildcards are the same decision. The translation is the spec's
+shared `likePatternToGlobPattern`, and the comparand is escaped first, so an
+author's own `%` / `_` / `*` / `?` / `[` stay literal instead of becoming the
+match-every-row bypass (#5567).
+
+**The `|| '='` fallback is gone with it.** `operatorToSql` was a
+name→name map, which cannot hold a wildcard, a list, or a null-safe negation;
+it is now a builder table keyed by `CubeOperator`, the shape #5374 gave the
+mingo exit, so a widened vocabulary fails to compile until its SQL spelling
+exists. Three operators were reaching that fallback and are fixed with it —
+measured against `query()` on a six-row fixture:
+
+| `where` | `query()` | echoed, before | echoed, now |
+|---|---|---|---|
+| `{name: {$in: [a, b]}}` | both rows | `name = a` — one row | `name IN (a, b)` |
+| `{name: {$nin: [a]}}` | the other five | `name = a` — the **complement** | `(name IS NULL OR name NOT IN (a))` |
+| `{name: {$exists: true}}` | five rows | `name = 1` — **no** rows | `name IS NOT NULL` |
+
+Three smaller divergences on the same builder went with them: negations are
+null-safe (`$ne` / `$nin` / `$notContains` kept only rows whose column was not
+NULL, where the pipeline returns them — the #5146 / #5297 rule the rest of the
+repo already follows); an empty `$in` / `$nin` list now renders a predicate
+instead of no `WHERE` at all (an empty `$in` echoed the whole table while the
+pipeline returns nothing); and a bare-day `$lte` bound renders half-open, as the
+pipeline has read it since #4042.
+
+`$startsWith` and `$endsWith` never reached the fallback and are unchanged: this
+face does not lower them, so both exits refuse them with `INVALID_FILTER` / 400
+(#5345).
+
+Only the *displayed* SQL changes — this exit produces the statement shown for
+transparency, never the query that runs, and `query()`'s rows are untouched.

packages/drivers/driver-memory/package.json

@@ -25,6 +25,8 @@
   },
   "devDependencies": {
     "@types/node": "^26.1.2",
+    "@types/sql.js": "^1.4.11",
+    "sql.js": "^1.14.1",
     "typescript": "^6.0.3",
     "vitest": "^4.1.10"
   },