fix(vscode): address chat search review findings from testing round

High: visible tool content (e.g. a todowrite checklist) could be highlighted in the DOM while the counter still reported "No results" and navigation was disabled. toolText() previously only indexed state.title for non-bash tools, independently of what transcript-search-highlight.ts actually scans in the DOM. Rewrote it to recursively collect every string leaf from state.input/state.metadata plus state.output for any tool, so matching and highlighting draw from one canonical, much more comprehensive notion of "the tool's text" instead of two divergent ones. read/glob/grep/list remain excluded -- kilo-ui's context-tool-results.tsx confirmed they never render raw input/output text, even expanded, so including it there would reintroduce the same class of mismatch for those tools. High: search only covered the initially-loaded message page (80 messages), silently missing older history in long sessions. MessageList now auto-requests older pages (session.loadOlderMessages()) while a search is active with a non-empty query, looping via reactivity on hasOlderMessages()/loadingOlderMessages() until history is exhausted. A new searchingHistory state surfaces this to the widget, which shows "Searching earlier messages..." and withholds a final "No results" until the whole session has actually been searched (a live match count still updates progressively as pages load). Medium: whole-word matching used plain \b, which only treats ASCII letters/digits/underscore as word characters and silently breaks for Cyrillic, Arabic, CJK, and similar text. Replaced with Unicode-aware boundary lookarounds using \p{L}/\p{M}/\p{N} property escapes, with the `u` flag applied to every compiled pattern. New chat.search.searchingHistory i18n key added across all 20 locales. (cherry picked from commit 47d4ebd73b)
2026-08-28 19:11:03 +08:00 · 2026-07-10 11:08:41 +02:00
parent 3d28bfc50e
commit f96771d302
23 changed files with 108 additions and 12 deletions
@@ -61,7 +61,7 @@ import {
  parseProviderAuthError,
  unwrapError,
 } from "../../utils/errorUtils"
-import type { Part, QuestionRequest, SuggestionRequest } from "../../types/messages"
+import type { Part, QuestionRequest, SuggestionRequest, ToolState } from "../../types/messages"

 interface MessageListProps {
  onSelectSession?: (id: string) => void
@@ -210,19 +210,37 @@ export const MessageList: Component<MessageListProps> = (props) => {

  // Extracts only the text kilo-ui's tool renderers actually put on screen —
  // matched field-by-field rather than reading `state.title` generically.
-  // The bash/shell renderer never shows `state.title` (its header is a
-  // static "Shell" label); the visible command/description come from
-  // `state.input` instead, and its output is shown as plain scrollable
-  // text. Other tools' `state.output` is typically internal data (file
-  // contents, JSON) that isn't rendered inline, so including it — or bash's
-  // unused `title` — produces search matches with no corresponding
-  // highlight, which is what made navigation appear to skip past matches.
+  // Uses one canonical extraction for both counting/navigation (this
+  // function) and highlighting (transcript-search-highlight.ts scans the
+  // rendered DOM) rather than two independently maintained notions of "the
+  // tool's text" — a hand-picked field list here previously missed content
+  // that's genuinely always on screen (e.g. a todowrite checklist's item
+  // text), so a visible match could be highlighted in the DOM while the
+  // counter still reported "No results" and navigation was disabled.
+  //
+  // read/glob/grep/list are the one confirmed exception: kilo-ui always
+  // collapses them into a context-group summary (context-tool-results.tsx)
+  // that never renders raw input/output text, even expanded — including
+  // that text here would count matches with no corresponding highlight,
+  // the same class of bug this rewrite fixes for every other tool.
+  const CONTEXT_GROUP_TOOLS = new Set(["read", "glob", "grep", "list"])
+
  function toolText(part: Part & { type: "tool" }): string[] {
    const state = part.state
    if (state.status === "running") return state.title ? [state.title] : []
    if (state.status === "error") return state.error ? [state.error] : []
    if (state.status !== "completed") return []
-    if (part.tool !== "bash") return state.title ? [state.title] : []
+    if (CONTEXT_GROUP_TOOLS.has(part.tool)) return state.title ? [state.title] : []
+    if (part.tool === "bash") return bashText(state)
+    const chunks: string[] = []
+    if (state.title) chunks.push(state.title)
+    collectStrings(state.input, chunks)
+    collectStrings(state.metadata, chunks)
+    if (typeof state.output === "string" && state.output) chunks.push(state.output)
+    return chunks
+  }
+
+  function bashText(state: Extract<ToolState, { status: "completed" }>) {
    const input = state.input as { command?: string; description?: string } | undefined
    const metadata = state.metadata as { command?: string; description?: string } | undefined
    const command = input?.command ?? metadata?.command
@@ -238,6 +256,27 @@ export const MessageList: Component<MessageListProps> = (props) => {
    return chunks
  }

+  // Recursively collects every string leaf value from a tool's `input`/
+  // `metadata` (JSON-like objects/arrays of unknown shape), so nested
+  // rendered text — a todo item's `content`, a question's `question` text,
+  // a skill's `name` — is included without hand-modeling each tool's shape.
+  function collectStrings(value: unknown, out: string[], depth = 0): void {
+    if (depth > 4 || value === undefined || value === null) return
+    if (typeof value === "string") {
+      if (value) out.push(value)
+      return
+    }
+    if (Array.isArray(value)) {
+      for (const item of value) collectStrings(item, out, depth + 1)
+      return
+    }
+    if (typeof value === "object") {
+      for (const key of Object.keys(value as Record<string, unknown>)) {
+        collectStrings((value as Record<string, unknown>)[key], out, depth + 1)
+      }
+    }
+  }
+
  function buildPattern(query: string, matchCase: boolean, wholeWord: boolean, regex: boolean): RegExp | undefined {
    if (!query) return undefined
    try {
@@ -246,9 +285,14 @@ export const MessageList: Component<MessageListProps> = (props) => {
        pattern = query.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")
      }
      if (wholeWord) {
-        pattern = `\\b(?:${pattern})\\b`
+        // Unicode-aware boundary: plain `\b` only treats ASCII letters/
+        // digits/underscore as "word" characters, so it silently breaks
+        // whole-word matching for Cyrillic, Arabic, CJK, and other non-ASCII
+        // text. `\p{L}`/`\p{M}`/`\p{N}` (letters/marks/numbers) require the
+        // `u` flag, applied below for every pattern, not just this one.
+        pattern = `(?<![\\p{L}\\p{M}\\p{N}_])(?:${pattern})(?![\\p{L}\\p{M}\\p{N}_])`
      }
-      return new RegExp(pattern, matchCase ? "g" : "gi")
+      return new RegExp(pattern, matchCase ? "gu" : "giu")
    } catch {
      return undefined
    }
@@ -268,6 +312,24 @@ export const MessageList: Component<MessageListProps> = (props) => {
    search.setInvalid(search.active() && !!q && search.regex() && !pattern())
  })

+  // Sessions only load the most recent page (session.tsx's MESSAGE_PAGE_LIMIT)
+  // up front; matches() only ever sees currently-loaded rows(). Without this,
+  // an active search would silently miss everything in older, not-yet-loaded
+  // history — undermining the main "find something in a long session" use
+  // case. While a query is active, keep requesting older pages until there
+  // either aren't any more or the search is no longer active; each
+  // completed load feeds back into hasOlderMessages()/loadingOlderMessages(),
+  // both tracked here, so this effect naturally re-fires and continues the
+  // chain without an explicit loop. searchingHistory (surfaced to the
+  // widget) stays true for that whole stretch, so "No results"/a final
+  // count aren't shown until the whole session has actually been searched.
+  createEffect(() => {
+    const searching = search.active() && !!search.query() && session.hasOlderMessages()
+    search.setSearchingHistory(searching)
+    if (!searching || session.loadingOlderMessages()) return
+    session.loadOlderMessages()
+  })
+
  const matches = createMemo(() => {
    const p = pattern()
    if (!p) return []
@@ -116,7 +116,12 @@ export const TranscriptSearch: Component = () => {
        <Show when={search.invalid()}>
          <span data-slot="transcript-search-error">{language.t("chat.search.invalidRegex")}</span>
        </Show>
-        <Show when={!search.invalid() && search.query().length > 0 && search.count() === 0}>
+        <Show when={!search.invalid() && search.searchingHistory() && search.count() === 0}>
+          <span data-slot="transcript-search-empty">{language.t("chat.search.searchingHistory")}</span>
+        </Show>
+        <Show
+          when={!search.invalid() && !search.searchingHistory() && search.query().length > 0 && search.count() === 0}
+        >
          <span data-slot="transcript-search-empty">{language.t("chat.search.noResults")}</span>
        </Show>
        <Show when={!search.invalid() && search.count() > 0}>
@@ -32,6 +32,12 @@ interface TranscriptSearchContextValue {
   * indistinguishable from a plain "no matches". */
  invalid: Accessor<boolean>
  setInvalid: (value: boolean) => void
+  /** True while MessageList is auto-loading older message pages to search
+   * them too — the session only loads the most recent page by default, so
+   * without this the widget would report "No results"/a final count while
+   * older history hadn't been searched yet. */
+  searchingHistory: Accessor<boolean>
+  setSearchingHistory: (value: boolean) => void
 }

 const TranscriptSearchContext = createContext<TranscriptSearchContextValue>()
@@ -46,6 +52,7 @@ export const TranscriptSearchProvider: ParentComponent = (props) => {
  const [count, setCount] = createSignal(0)
  const [jump, setJump] = createSignal(0)
  const [invalid, setInvalid] = createSignal(false)
+  const [searchingHistory, setSearchingHistory] = createSignal(false)

  return (
    <TranscriptSearchContext.Provider
@@ -68,6 +75,8 @@ export const TranscriptSearchProvider: ParentComponent = (props) => {
        requestJump: () => setJump((n) => n + 1),
        invalid,
        setInvalid,
+        searchingHistory,
+        setSearchingHistory,
      }}
    >
      {props.children}
@@ -1845,4 +1845,5 @@ export const dict = {
  "chat.search.close": "إغلاق البحث",
  "chat.search.invalidRegex": "تعبير عادي غير صالح",
  "chat.search.noResults": "لا توجد نتائج",
+  "chat.search.searchingHistory": "جارٍ البحث في الرسائل السابقة…",
 }
@@ -1894,4 +1894,5 @@ export const dict = {
  "chat.search.close": "Fechar pesquisa",
  "chat.search.invalidRegex": "Expressão regular inválida",
  "chat.search.noResults": "Nenhum resultado",
+  "chat.search.searchingHistory": "Pesquisando mensagens anteriores…",
 }
@@ -1886,4 +1886,5 @@ export const dict = {
  "chat.search.close": "Zatvori pretragu",
  "chat.search.invalidRegex": "Nevažeći regularni izraz",
  "chat.search.noResults": "Nema rezultata",
+  "chat.search.searchingHistory": "Pretraživanje ranijih poruka…",
 }
@@ -1878,4 +1878,5 @@ export const dict = {
  "chat.search.close": "Luk søgning",
  "chat.search.invalidRegex": "Ugyldigt regulært udtryk",
  "chat.search.noResults": "Ingen resultater",
+  "chat.search.searchingHistory": "Søger i tidligere beskeder…",
 }
@@ -1916,4 +1916,5 @@ export const dict = {
  "chat.search.close": "Suche schließen",
  "chat.search.invalidRegex": "Ungültiger regulärer Ausdruck",
  "chat.search.noResults": "Keine Ergebnisse",
+  "chat.search.searchingHistory": "Frühere Nachrichten werden durchsucht…",
 } satisfies Partial<Record<Keys, string>>
@@ -1873,4 +1873,5 @@ export const dict = {
  "chat.search.close": "Close search",
  "chat.search.invalidRegex": "Invalid regular expression",
  "chat.search.noResults": "No results",
+  "chat.search.searchingHistory": "Searching earlier messages…",
 }
@@ -1902,4 +1902,5 @@ export const dict = {
  "chat.search.close": "Cerrar búsqueda",
  "chat.search.invalidRegex": "Expresión regular no válida",
  "chat.search.noResults": "Sin resultados",
+  "chat.search.searchingHistory": "Buscando en mensajes anteriores…",
 }
@@ -1926,4 +1926,5 @@ export const dict = {
  "chat.search.close": "Fermer la recherche",
  "chat.search.invalidRegex": "Expression régulière non valide",
  "chat.search.noResults": "Aucun résultat",
+  "chat.search.searchingHistory": "Recherche dans les messages précédents…",
 }
@@ -1811,4 +1811,5 @@ export const dict = {
  "chat.search.close": "Chiudi ricerca",
  "chat.search.invalidRegex": "Espressione regolare non valida",
  "chat.search.noResults": "Nessun risultato",
+  "chat.search.searchingHistory": "Ricerca nei messaggi precedenti…",
 } as const
@@ -1871,4 +1871,5 @@ export const dict = {
  "chat.search.close": "検索を閉じる",
  "chat.search.invalidRegex": "正規表現が無効です",
  "chat.search.noResults": "見つかりませんでした",
+  "chat.search.searchingHistory": "以前のメッセージを検索しています…",
 }
@@ -1853,4 +1853,5 @@ export const dict = {
  "chat.search.close": "검색 닫기",
  "chat.search.invalidRegex": "정규식이 잘못되었습니다",
  "chat.search.noResults": "검색 결과 없음",
+  "chat.search.searchingHistory": "이전 메시지를 검색하는 중…",
 }
@@ -1913,4 +1913,5 @@ export const dict = {
  "chat.search.close": "Zoeken sluiten",
  "chat.search.invalidRegex": "Ongeldige reguliere expressie",
  "chat.search.noResults": "Geen resultaten",
+  "chat.search.searchingHistory": "Eerdere berichten doorzoeken…",
 }
@@ -1871,4 +1871,5 @@ export const dict = {
  "chat.search.close": "Lukk søk",
  "chat.search.invalidRegex": "Ugyldig regulært uttrykk",
  "chat.search.noResults": "Ingen resultater",
+  "chat.search.searchingHistory": "Søker i tidligere meldinger…",
 } satisfies Partial<Record<Keys, string>>
@@ -1884,4 +1884,5 @@ export const dict = {
  "chat.search.close": "Zamknij wyszukiwanie",
  "chat.search.invalidRegex": "Nieprawidłowe wyrażenie regularne",
  "chat.search.noResults": "Brak wyników",
+  "chat.search.searchingHistory": "Wyszukiwanie we wcześniejszych wiadomościach…",
 }
@@ -1882,4 +1882,5 @@ export const dict = {
  "chat.search.close": "Закрыть поиск",
  "chat.search.invalidRegex": "Недопустимое регулярное выражение",
  "chat.search.noResults": "Нет результатов",
+  "chat.search.searchingHistory": "Поиск в более ранних сообщениях…",
 }
@@ -1851,4 +1851,5 @@ export const dict = {
  "chat.search.close": "ปิดการค้นหา",
  "chat.search.invalidRegex": "นิพจน์ทั่วไปไม่ถูกต้อง",
  "chat.search.noResults": "ไม่มีผลลัพธ์",
+  "chat.search.searchingHistory": "กำลังค้นหาข้อความก่อนหน้า…",
 }
@@ -1900,4 +1900,5 @@ export const dict = {
  "chat.search.close": "Aramayı kapat",
  "chat.search.invalidRegex": "Geçersiz normal ifade",
  "chat.search.noResults": "Sonuç yok",
+  "chat.search.searchingHistory": "Önceki mesajlarda aranıyor…",
 }
@@ -1897,4 +1897,5 @@ export const dict = {
  "chat.search.close": "Закрити пошук",
  "chat.search.invalidRegex": "Недійсний регулярний вираз",
  "chat.search.noResults": "Немає результатів",
+  "chat.search.searchingHistory": "Пошук у попередніх повідомленнях…",
 }
@@ -1800,4 +1800,5 @@ export const dict = {
  "chat.search.close": "关闭搜索",
  "chat.search.invalidRegex": "正则表达式无效",
  "chat.search.noResults": "无结果",
+  "chat.search.searchingHistory": "正在搜索更早的消息…",
 } satisfies Partial<Record<Keys, string>>
@@ -1806,4 +1806,5 @@ export const dict = {
  "chat.search.close": "關閉搜尋",
  "chat.search.invalidRegex": "規則運算式無效",
  "chat.search.noResults": "無結果",
+  "chat.search.searchingHistory": "正在搜尋較早的訊息…",
 } satisfies Partial<Record<Keys, string>>