[fjakobs]

Why

• The type generator types numeric columns as number and BOOLEAN columns as boolean in analytics.d.ts.
• The /query route returns them as strings. The warehouse's JSON_ARRAY format sends every cell as a string, and the connector passed them through unchanged.
• So useAnalyticsQuery rows had total_trips: "21932" while the type said number. Apps (and agents writing apps) kept adding their own Number(...) workarounds. The dev-playground did the same in its KPI cards.

Reproduced on e2-dogfood (Serverless Starter Warehouse). This is a query run through the real type generator and the connector's JSON_ARRAY path:

column	generated type	value before
BIGINT	number	"42"
DECIMAL(10,2)	number	"1.50"
INT / SMALLINT / TINYINT	number	"7" / "3" / "1"
DOUBLE / FLOAT	number	"2.5" / "1.25"
BOOLEAN	boolean	"true"

Change

• The SQL warehouse connector attaches the manifest column types (columnTypes) next to the columnNames it already carries. It does this on JSON rows and on inline Arrow results. Cell values are unchanged at this layer.
• deliverJsonResult takes a coerceScalars option. With it set, coerceScalarCells converts cells by manifest typeName:
  • BYTE, SHORT, INT, LONG, FLOAT, DOUBLE and DECIMAL become numbers.
  • BOOLEAN becomes a boolean.
  • This happens on both the native JSON_ARRAY path and the ARROW_STREAM fallback, so the two paths still match.
• Only the /query route sets coerceScalars. The /metric route keeps its documented string | null wire values, so metric-view types and docs don't change.
• NULL stays null. DATE and TIMESTAMP stay ISO strings. STRUCT, ARRAY and MAP values stay as delivered.
• DOUBLE/FLOAT NaN, Infinity and -Infinity stay the strings "NaN", "Infinity" and "-Infinity". JSON has no number for them, and converting them would put null on the wire, the same as SQL NULL.
• Dev-playground: remove toNum and the other string-coercion guards in kpi-cards.tsx.
• Docs: add a "Result value types" note to type-generation.md.

Trade-offs

• A BIGINT beyond Number.MAX_SAFE_INTEGER and a DECIMAL beyond about 15 significant digits lose precision as JS numbers. DECIMAL also loses trailing zeros ("1.50" becomes 1.5). The docs say to CAST(... AS STRING) for exact values. The generated type then becomes string, so it stays truthful.
• This changes runtime values on /query: they become what the generated types already claimed. Code that relied on getting strings at runtime will see numbers.
• Not in scope: nested types (STRUCT, ARRAY, MAP) still arrive as JSON strings although they are typed as unknown[] or Record<string, unknown>.

Testing

• Unit tests:
  • connector: columnNames and columnTypes are attached.
  • deliverJsonResult: native path, Reyden fallback using the captured fixture, and no conversion without the option.
  • /query route: converts values.
  • /metric route: keeps strings.
• pnpm build, pnpm docs:build, pnpm check:fix, pnpm -r typecheck and pnpm test all pass (5934 tests).
• End to end, dev-playground against e2-dogfood:
  • POST /api/analytics/query/dashboard_kpis now returns {"total_trips":21932,"avg_fare":12.35,…}.
  • Sparkline rows have numeric values and trip_date stays "2015-12-31".
  • POST /api/analytics/metric/revenue still returns {"arr":"1.604847408E9"}.
  • The smart-dashboard KPI cards, sparklines and trends render correctly without toNum.
• After review: a query with every scalar type, typed NULLs and DOUBLE NaN/±Infinity through /query, on both a standard serverless warehouse and a Reyden (REALTIME) warehouse. Every value matches its generated type except the three non-finite strings, and NULLs are null. Also ran the ARROW_STREAM fallback with live Reyden Arrow bytes (the JSON_ARRAY rejection was simulated, since the warehouse now serves JSON_ARRAY): the same values as native JSON_ARRAY.

This pull request and its description were written by Isaac.

[github-actions]

📦 Bundle size report

Compared against bundle-size-baseline.json (main).

@databricks/appkit

npm tarball (packed): 1.3 MB (+2.3 KB) — gzipped download (dist + bin; excludes release-only docs/NOTICE).

dist	raw	gzip
JS (runtime)	1.3 MB (+2.6 KB)	479 KB (+938 B)
Type declarations	470 KB	169 KB
Source maps	2.6 MB (+4.4 KB)	889 KB (+1.4 KB)
Other	11 KB	3.7 KB
Total	4.4 MB (+6.9 KB)	1.5 MB (+2.3 KB)

Per-entry composition (own code — deps external (as shipped))

Entry	Initial (gz)	Lazy (gz)	Total (gz)	node_modules (min)	Own code (min)
.	108 KB (+257 B)	2.5 KB	110 KB (+257 B)	external	348 KB (+645 B)
./beta	113 KB (+32 B)	485 B	114 KB (+32 B)	external	339 KB (+154 B)
./testing	46 KB	33 KB	79 KB	external	229 KB
./tsdown	520 B	0 B	520 B	external	813 B
./type-generator	26 KB	0 B	26 KB	external	74 KB

Chunks:

Entry	Chunk	Load	Size (gz)
.	index.js	initial	103 KB
.	utils.js	initial	4.6 KB
.	remote-tunnel-manager.js	lazy	2.5 KB
./beta	beta.js	initial	93 KB
./beta	modular.js	initial	6.0 KB
./beta	stream-manager.js	initial	5.9 KB
./beta	service-context.js	initial	4.3 KB
./beta	databricks.js	initial	3.3 KB
./beta	client.js	initial	649 B
./beta	index.js	initial	20 B
./beta	supervisor-api.js	lazy	193 B
./beta	databricks.js	lazy	178 B
./beta	index.js	lazy	114 B
./testing	manifest.js	initial	33 KB
./testing	index.js	initial	10 KB
./testing	wide-event-emitter.js	initial	2.9 KB
./testing	index.js	lazy	29 KB
./testing	remote-tunnel-manager.js	lazy	2.5 KB
./testing	utils.js	lazy	1.8 KB
./tsdown	index.js	initial	520 B
./type-generator	index.js	initial	26 KB

@databricks/appkit-ui

npm tarball (packed): 406 KB (-5 B) — gzipped download (dist + bin; excludes release-only docs/NOTICE).

dist	raw	gzip
JS (runtime)	436 KB	150 KB
Type declarations	271 KB	100 KB (-1 B)
Source maps	862 KB	291 KB
CSS	16 KB	3.2 KB
Total	1.5 MB	544 KB (-1 B)

Per-entry composition (consumer bundle — deps bundled, peerDeps external)

Entry	Initial (gz)	Lazy (gz)	Total (gz)	node_modules (min)	Own code (min)
./js	5.3 KB	49 KB	55 KB	208 KB	14 KB
./js/beta	2.1 KB	0 B	2.1 KB	0 B	5.0 KB
./react	440 KB	49 KB	489 KB	1.4 MB	178 KB
./react/beta	14 KB	0 B	14 KB	25 KB	16 KB

Chunks:

Entry	Chunk	Load	Size (gz)
./js	index.js	initial	5.2 KB
./js	chunk	initial	120 B
./js	apache-arrow	lazy	49 KB
./js/beta	beta.js	initial	2.1 KB
./react	index.js	initial	438 KB
./react	tslib	initial	2.1 KB
./react	apache-arrow	lazy	49 KB
./react/beta	beta.js	initial	14 KB

[github-actions]

🤖 AppKit PR bot

🔬 Run evals

Start an eval for this PR from the evals-monitor app: Go to Evals Monitor →

📦 Try this PR's app template

Scaffolds a new app from this PR's SDK build. Run it in any folder (requires the GitHub CLI — gh auth login — and the Databricks CLI):

gh run download 43204330191 -R databricks/appkit -n appkit-template-0.94.0-pr.b55196f-fjakobs-la-2230-numeric-types-709 -D appkit-pr-709 \
  && unzip -o "appkit-pr-709/appkit-template-0.94.0-pr.b55196f-fjakobs-la-2230-numeric-types-709.zip" -d "appkit-pr-709" \
  && databricks apps init --template "appkit-pr-709"

The template pins @databricks/appkit and @databricks/appkit-ui to tarballs built from this branch, so the scaffolded app runs against this PR's code.

[fjakobs]

@pffigueiredo thanks for the review. I replied in your thread, which is resolved and so collapsed: #709 (comment)

Short version: I kept plain Number() / === "true" instead of zod. .parse throws on the "NaN" / "Infinity" / "-Infinity" strings the warehouse really returns for DOUBLE/FLOAT, and that would fail the whole query. Your .finite() did point at a real bug, though: those values were reaching the client as null, the same as SQL NULL. 4f6d390 keeps them as the warehouse strings and adds tests and a docs note. Could you take another look?

[fjakobs]

@pffigueiredo friendly ping: I replied to your zod suggestion in the (resolved) thread and pushed a follow-up commit for the NaN/Infinity case it uncovered. Could you take another look and update your review?

[fjakobs]

@MarioCadenas could you review this when you get a chance? It needs a code-owner approval; pffigueiredo has approved, CI is green and it's up to date with main.