@azure-id/orc 1.9.2 → 2.1.0

package/bin/webui/js/01-i18n.js

@@ -1,154 +1,155 @@
-"use strict";
-/* 01-i18n.js — orc ui client
-   LANGS, loadLang, setLang, t, applyStaticText, cycleLang.
-   
-   Keys are written out IN FULL, never assembled from a fragment. The v0.48.1
-   file split is by key PREFIX only; `docs.ship.confirm` is still spelled
-   `docs.ship.confirm` at every call site.
-
-   Loaded by app.html in the order its numeric prefix names. Classic
-   script, no import/export: an ES module import carries no query string,
-   and every static request here needs the per-launch session token. */
-
-/* ------------------------------------------------------------------- i18n -- */
-/*
-   THE SCOPE RULE, and it is deliberately narrow: only the panel's OWN prose is
-   translated. Everything that arrives from `bin/cli.js --json` stays exactly as
-   the CLI wrote it — config keys and their descriptions, values, agent names,
-   model ids, file paths, commands, `orc doctor` messages, command output. Those
-   are identifiers, not sentences: a translated config key is a key that does not
-   exist, and a translated command is a command you cannot type. So `t()` is
-   never applied to a value read out of a payload, and neither string table
-   contains a single ORC key name.
-
-   The tables are plain JSON served as static assets. Adding a language is a
-   FOLDER in i18n/ plus one row in LANGS — no build step, no library, no
-   dependency, and no server change (serve.js walks i18n/ at boot).
-*/
-const LANGS = [
-  { code: "en", label: "English" },
-  { code: "id", label: "Bahasa Indonesia" },
-];
-const LANG_KEY = "orc-ui-lang";
-
-// v0.48.1 — one file per namespace instead of one 53 KB table, so changing the
-// Docs panel's wording means opening i18n/en/docs.json rather than paging
-// through every string in the panel.
-//
-// A plain array, and deliberately NOT an index file to fetch: an index would be
-// one more round trip and one more thing to forget to update. A namespace owns
-// one or more key PREFIXES (`stats` owns `stats.*` and `cost.*`), which the
-// test suite asserts — a key whose prefix belongs to no namespace would load
-// nowhere and render as a raw dotted string.
-const NAMESPACES = [
-  "common",
-  "nav",
-  "banner",
-  "overview",
-  "settings",
-  "lanes",
-  "rules",
-  "runs",
-  "knowledge",
-  "stats",
-  "flow",
-  "crosslink",
-  "learn",
-  "mockrun",
-  "maintenance",
-  "pact",
-  "boundary",
-  "wait",
-  "hookui",
-  "handoff",
-  "challenge",
-  "docs",
-  "test",
-  "extra",
-  "experiment",
-  "tour",
-];
-
-let lang = "en";
-// English is loaded first and kept as the FALLBACK table: a key the other
-// language has not translated yet renders in English rather than as a raw
-// dotted key. A half-finished translation degrades to mixed prose, never to
-// debug output on somebody's screen.
-let DICT = {};
-let DICT_EN = {};
-
-// The namespaces are fetched in parallel and merged into ONE flat table, so
-// `t()` is unchanged and no call site knows the split happened.
-//
-// A namespace that fails to load falls back to ENGLISH for its keys rather than
-// rejecting the whole language: a network hiccup on one file should cost that
-// file's prose, not the entire panel. English itself has no fallback below it,
-// so a failure there rejects and boot() lands on raw keys — ugly, never blank.
-function loadLang(code) {
-  return Promise.all(
-    NAMESPACES.map((ns) =>
-      fetch(`/i18n/${code}/${ns}.json`, { headers: { "x-orc-token": TOKEN }, cache: "no-store" })
-        .then((r) => {
-          if (!r.ok) throw new Error(`i18n ${code}/${ns} ${r.status}`);
-          return r.json();
-        })
-        .catch((e) => {
-          if (code === "en") throw e;
-          return null;
-        })
-    )
-  ).then((parts) => Object.assign({}, ...parts.filter(Boolean)));
-}
-
-// A missing or broken table must never blank the UI. Every failure path here
-// lands on English, so the worst case is untranslated text — never empty text.
-async function setLang(code, opts) {
-  const rerender = !opts || opts.rerender !== false;
-  const want = LANGS.some((l) => l.code === code) ? code : "en";
-  try {
-    DICT = want === "en" ? DICT_EN : await loadLang(want);
-    lang = want;
-  } catch (_) {
-    DICT = DICT_EN;
-    lang = "en";
-  }
-  try {
-    localStorage.setItem(LANG_KEY, lang);
-  } catch (_) {}
-  document.documentElement.setAttribute("lang", lang);
-  applyStaticText();
-  if (rerender) route();
-}
-
-function t(key, vars) {
-  let s = DICT[key];
-  if (s === undefined) s = DICT_EN[key];
-  if (s === undefined) s = key;
-  if (vars) for (const k of Object.keys(vars)) s = s.split("{" + k + "}").join(String(vars[k]));
-  return s;
-}
-
-// English has one plural form and Indonesian has none, so count-aware lookup is
-// a second key rather than a plural-rules engine: `n === 1` picks the singular.
-const tn = (n, key, vars) => t(n === 1 ? key : key + "Plural", Object.assign({ n }, vars || {}));
-
-// The parts of the shell that are markup rather than a render function. Called
-// on every language change; panels re-render themselves through `route()`.
-function applyStaticText() {
-  for (const node of document.querySelectorAll("[data-i18n]")) node.textContent = t(node.dataset.i18n);
-  const tt = $("#theme-toggle");
-  if (tt) tt.textContent = document.documentElement.getAttribute("data-theme") === "light" ? t("rail.dark") : t("rail.light");
-  const hint = $("#shortcut-hint");
-  if (hint) hint.title = t("rail.shortcuts");
-  const label = $("#lang-label");
-  if (label) label.textContent = (LANGS.find((l) => l.code === lang) || LANGS[0]).label;
-  const btn = $("#lang-toggle");
-  if (btn) btn.title = t("lang.title");
-}
-
-// Two languages, so the control is a TOGGLE rather than a menu: one click, no
-// dropdown to open. A third language would make this a <select>; two do not.
-function cycleLang() {
-  const i = LANGS.findIndex((l) => l.code === lang);
-  setLang(LANGS[(i + 1) % LANGS.length].code);
-}
+"use strict";
+/* 01-i18n.js — orc ui client
+   LANGS, loadLang, setLang, t, applyStaticText, cycleLang.
+   
+   Keys are written out IN FULL, never assembled from a fragment. The v0.48.1
+   file split is by key PREFIX only; `docs.ship.confirm` is still spelled
+   `docs.ship.confirm` at every call site.
+
+   Loaded by app.html in the order its numeric prefix names. Classic
+   script, no import/export: an ES module import carries no query string,
+   and every static request here needs the per-launch session token. */
+
+/* ------------------------------------------------------------------- i18n -- */
+/*
+   THE SCOPE RULE, and it is deliberately narrow: only the panel's OWN prose is
+   translated. Everything that arrives from `bin/cli.js --json` stays exactly as
+   the CLI wrote it — config keys and their descriptions, values, agent names,
+   model ids, file paths, commands, `orc doctor` messages, command output. Those
+   are identifiers, not sentences: a translated config key is a key that does not
+   exist, and a translated command is a command you cannot type. So `t()` is
+   never applied to a value read out of a payload, and neither string table
+   contains a single ORC key name.
+
+   The tables are plain JSON served as static assets. Adding a language is a
+   FOLDER in i18n/ plus one row in LANGS — no build step, no library, no
+   dependency, and no server change (serve.js walks i18n/ at boot).
+*/
+const LANGS = [
+  { code: "en", label: "English" },
+  { code: "id", label: "Bahasa Indonesia" },
+];
+const LANG_KEY = "orc-ui-lang";
+
+// v0.48.1 — one file per namespace instead of one 53 KB table, so changing the
+// Docs panel's wording means opening i18n/en/docs.json rather than paging
+// through every string in the panel.
+//
+// A plain array, and deliberately NOT an index file to fetch: an index would be
+// one more round trip and one more thing to forget to update. A namespace owns
+// one or more key PREFIXES (`stats` owns `stats.*` and `cost.*`), which the
+// test suite asserts — a key whose prefix belongs to no namespace would load
+// nowhere and render as a raw dotted string.
+const NAMESPACES = [
+  "common",
+  "nav",
+  "banner",
+  "overview",
+  "settings",
+  "lanes",
+  "rules",
+  "runs",
+  "knowledge",
+  "stats",
+  "behaviour",
+  "flow",
+  "crosslink",
+  "learn",
+  "mockrun",
+  "maintenance",
+  "pact",
+  "boundary",
+  "wait",
+  "hookui",
+  "handoff",
+  "challenge",
+  "docs",
+  "test",
+  "extra",
+  "experiment",
+  "tour",
+];
+
+let lang = "en";
+// English is loaded first and kept as the FALLBACK table: a key the other
+// language has not translated yet renders in English rather than as a raw
+// dotted key. A half-finished translation degrades to mixed prose, never to
+// debug output on somebody's screen.
+let DICT = {};
+let DICT_EN = {};
+
+// The namespaces are fetched in parallel and merged into ONE flat table, so
+// `t()` is unchanged and no call site knows the split happened.
+//
+// A namespace that fails to load falls back to ENGLISH for its keys rather than
+// rejecting the whole language: a network hiccup on one file should cost that
+// file's prose, not the entire panel. English itself has no fallback below it,
+// so a failure there rejects and boot() lands on raw keys — ugly, never blank.
+function loadLang(code) {
+  return Promise.all(
+    NAMESPACES.map((ns) =>
+      fetch(`/i18n/${code}/${ns}.json`, { headers: { "x-orc-token": TOKEN }, cache: "no-store" })
+        .then((r) => {
+          if (!r.ok) throw new Error(`i18n ${code}/${ns} ${r.status}`);
+          return r.json();
+        })
+        .catch((e) => {
+          if (code === "en") throw e;
+          return null;
+        })
+    )
+  ).then((parts) => Object.assign({}, ...parts.filter(Boolean)));
+}
+
+// A missing or broken table must never blank the UI. Every failure path here
+// lands on English, so the worst case is untranslated text — never empty text.
+async function setLang(code, opts) {
+  const rerender = !opts || opts.rerender !== false;
+  const want = LANGS.some((l) => l.code === code) ? code : "en";
+  try {
+    DICT = want === "en" ? DICT_EN : await loadLang(want);
+    lang = want;
+  } catch (_) {
+    DICT = DICT_EN;
+    lang = "en";
+  }
+  try {
+    localStorage.setItem(LANG_KEY, lang);
+  } catch (_) {}
+  document.documentElement.setAttribute("lang", lang);
+  applyStaticText();
+  if (rerender) route();
+}
+
+function t(key, vars) {
+  let s = DICT[key];
+  if (s === undefined) s = DICT_EN[key];
+  if (s === undefined) s = key;
+  if (vars) for (const k of Object.keys(vars)) s = s.split("{" + k + "}").join(String(vars[k]));
+  return s;
+}
+
+// English has one plural form and Indonesian has none, so count-aware lookup is
+// a second key rather than a plural-rules engine: `n === 1` picks the singular.
+const tn = (n, key, vars) => t(n === 1 ? key : key + "Plural", Object.assign({ n }, vars || {}));
+
+// The parts of the shell that are markup rather than a render function. Called
+// on every language change; panels re-render themselves through `route()`.
+function applyStaticText() {
+  for (const node of document.querySelectorAll("[data-i18n]")) node.textContent = t(node.dataset.i18n);
+  const tt = $("#theme-toggle");
+  if (tt) tt.textContent = document.documentElement.getAttribute("data-theme") === "light" ? t("rail.dark") : t("rail.light");
+  const hint = $("#shortcut-hint");
+  if (hint) hint.title = t("rail.shortcuts");
+  const label = $("#lang-label");
+  if (label) label.textContent = (LANGS.find((l) => l.code === lang) || LANGS[0]).label;
+  const btn = $("#lang-toggle");
+  if (btn) btn.title = t("lang.title");
+}
+
+// Two languages, so the control is a TOGGLE rather than a menu: one click, no
+// dropdown to open. A third language would make this a <select>; two do not.
+function cycleLang() {
+  const i = LANGS.findIndex((l) => l.code === lang);
+  setLang(LANGS[(i + 1) % LANGS.length].code);
+}