fix(client): name the input limit in the autocomplete error

.changeset/name-autocomplete-input-limit.md

@@ -0,0 +1,5 @@
+---
+'@gatepost/client': patch
+---
+
+Name the 64 code point limit in the error for autocomplete text over it.

packages/client/src/requests.ts

@@ -129,7 +129,11 @@ export function autocompleteRequest(q: string): {
   readonly typed: string;
   readonly request: GatewayRequest;
 } {
-  const typed = isOverInputLimit(q) ? '' : normalize(q);
+  if (isOverInputLimit(q)) {
+    const rule = `The text has more than ${String(INPUT_LIMIT)} code points.`;
+    throw clientError('invalid_input', `${rule} Send no call for it.`);
+  }
+  const typed = normalize(q);
   if (!QUERY.test(typed)) {
     const rule = 'The text must hold 1 to 11 letters and digits after normalize.';
     throw clientError('invalid_input', `${rule} Send no call for empty text.`);

packages/client/test/requests.test.ts

@@ -146,6 +146,14 @@ describe('autocompleteRequest', () => {
     expect(failure(() => autocompleteRequest(overLimit)).code).toBe('invalid_input');
   });
 
+  it('names the input limit when the text is over it', () => {
+    // Five letters after normalize, so only the length is wrong.
+    const overLimit = 'FC 10'.padEnd(maxInputCodePoints + 1);
+    const { message } = failure(() => autocompleteRequest(overLimit));
+    expect(message).toContain(`${String(maxInputCodePoints)} code points`);
+    expect(message).not.toContain('1 to 11');
+  });
+
   it('counts code points, not UTF-16 units, for the input limit', () => {
     // Each mathematical bold capital is one code point in two UTF-16 units. NFKC makes it a
     // plain letter, so text at the limit still works.