mirror of
https://github.com/multica-ai/multica.git
synced 2026-08-05 09:30:05 +02:00
tokens.css defined colours, radii and font families but not a single --text-* step, so font sizes had no baseline to align to and grew wherever they were needed: 51 distinct sizes across web + desktop, 370 written as arbitrary values, six at half a pixel (10.5 / 11.5 / 12.5 / 13.5 / 14.5 / 15.5px). text-xs and text-sm carried nearly all UI text while the range between them — 11, 13, 15px — could only be reached with arbitrary values. Hierarchy does not come from having more sizes; past a handful, each extra size makes the hierarchy blurrier. Add ten role-named steps, each with its own line-height so leading cannot fragment the way size did, and move every product-UI call site onto them. Steps are named for what the text is for, not for a t-shirt size, because that is what keeps the scale from drifting again. Six steps deliberately keep the exact size/line-height pairs of the Tailwind defaults they replace, so the ~1,900-call-site rename moves nothing on screen. The visible changes are confined to former arbitrary values snapping to a step: 8/9/10px -> micro (11px) on badges and overlines; 17 -> 18; 22 -> 24; 30 (text-3xl) -> 36 on headings and stat numbers; 12.8px -> label (13px) on small buttons and toggles. Half-pixel sizes are gone. This supersedes #6108, which was reverted by #6116 because the sidebar group labels rendered at the inherited 16px. The cause was not the scale but cn(): `text-<x>` is ambiguous in Tailwind, and tailwind-merge resolves it against a table listing only the default sizes, so it filed every role step under text-colour and dropped whichever of `text-caption` / `text-sidebar-foreground/70` came first. Registering the steps as a font-size class group restores the real conflict groups — size beats size, colour beats colour, the two coexist — and a test pins the list against the scale, since the failure is silent in source. Hand-written CSS is covered too. The transcript kept a 12.5px body long after every Tailwind call site was on the scale, so the "no half-pixel sizes" claim was true of the classes and false of the product; the editor's prose, code and mermaid ramps had the same blind spot, and seven of their eight values already equalled a step exactly. All now reference var(--text-*). The guard test reads raw `font-size:` declarations as well as class names, exempting only the 16px iOS input-zoom workaround in base.css and the landing pages' marketing ramp. apps/mobile (own NativeWind config) and apps/docs (fumadocs' own type system) keep Tailwind's default scale and are untouched. Landing display type (rem/clamp, 2.2-6.4rem) stays on its separate ramp, as do four decorative emoji / serif-hero sizes. Verified on a running local stack: pinned sidebar rows and group labels measure 12px/16px, nav items 14px/20px — identical to pre-migration. An audit of every rendered font size across the product surfaces finds nothing off the scale; the only exceptions are avatar initials and emoji, which actor-avatar.tsx sizes proportionally to the avatar diameter by design. Co-authored-by: Lambda <lambda@multica.ai> Co-authored-by: multica-agent <github@multica.ai>
293 lines
13 KiB
TypeScript
293 lines
13 KiB
TypeScript
"use client";
|
|
|
|
import { useRef, useState, useCallback, useEffect } from "react";
|
|
import { cn } from "@multica/ui/lib/utils";
|
|
import { ContentEditor, type ContentEditorRef, useFileDropZone, FileDropOverlay, useLazyEditor, useUploadGate, useComposerSubmit } from "../../editor";
|
|
import { FileUploadButton } from "@multica/ui/components/common/file-upload-button";
|
|
import { SubmitButton } from "@multica/ui/components/common/submit-button";
|
|
import { contentReferencesAttachment } from "@multica/core/types";
|
|
import { formatShortcut, useShortcut } from "@multica/core/shortcuts";
|
|
import { useCommentComposerStore, useCommentDraftStore } from "@multica/core/issues/stores";
|
|
import { useT } from "../../i18n";
|
|
import { CommentTriggerChips } from "./comment-trigger-chips";
|
|
import { useCommentTriggerPreview } from "../hooks/use-comment-trigger-preview";
|
|
import { useCommentUploads } from "./use-comment-uploads";
|
|
|
|
interface CommentInputProps {
|
|
issueId: string;
|
|
/** Resolves true on success, false on failure. The composer keeps the text
|
|
* (editor locked + button spinning) until this settles, then clears only on
|
|
* success — a failed send must not silently discard the user's draft. */
|
|
onSubmit: (content: string, attachmentIds?: string[], suppressAgentIds?: string[]) => Promise<boolean>;
|
|
}
|
|
|
|
function CommentInput({ issueId, onSubmit }: CommentInputProps) {
|
|
const { t } = useT("issues");
|
|
const { t: tEditor } = useT("editor");
|
|
const sendShortcut = useShortcut("send");
|
|
const editorRef = useRef<ContentEditorRef>(null);
|
|
// Sending mid-upload would strip the pending image's blob URL out of the
|
|
// markdown and bind no attachment id — the comment posts without the file.
|
|
const uploadGate = useUploadGate(editorRef);
|
|
// Read the persisted draft once on mount. ContentEditor only honors
|
|
// `defaultValue` at mount time, so this snapshot drives both the editor's
|
|
// initial content and the submit-button enable state — without this the
|
|
// button would be disabled even though the editor visibly contains text.
|
|
const draftKey = `new:${issueId}` as const;
|
|
const [initialDraft] = useState(() =>
|
|
useCommentDraftStore.getState().getDraft(draftKey),
|
|
);
|
|
const [content, setContent] = useState(initialDraft ?? "");
|
|
const [isEmpty, setIsEmpty] = useState(() => !initialDraft?.trim());
|
|
const [suppressedAgentIds, setSuppressedAgentIds] = useState<Set<string>>(() => new Set());
|
|
const triggerPreview = useCommentTriggerPreview({ issueId, content });
|
|
// Uploads for this composer session (MUL-5181). Owned by the module-level
|
|
// coordinator and persisted in the draft store, so closing/scrolling the
|
|
// composer away no longer drops an in-flight upload — its result lands in the
|
|
// draft. `attachments` (completed rows) drives both the submit `attachment_ids`
|
|
// payload and the editor's AttachmentDownloadProvider; `uploads` drives the
|
|
// status chips (uploading / failed / interrupted).
|
|
// `gate` widens the editor gate with coordinator-owned placeholders, so a
|
|
// composer reopened over a still-in-flight upload cannot send past it.
|
|
const { attachments: pendingAttachments, handleUpload, gate } =
|
|
useCommentUploads(draftKey, { issueId }, uploadGate, editorRef);
|
|
|
|
// Readonly-first: the composer renders as a same-looking static shell until
|
|
// the user shows intent (click / keyboard / file drop). An unsent draft is
|
|
// standing intent — mount the real editor immediately so the draft is
|
|
// visible and editable, exactly like the pre-lazy behavior.
|
|
const lazy = useLazyEditor({
|
|
initialActive:
|
|
!!initialDraft?.trim() ||
|
|
useCommentDraftStore.getState().getUploads(draftKey).length > 0,
|
|
editorRef,
|
|
});
|
|
const { isDragOver, dropZoneProps } = useFileDropZone({
|
|
onDrop: lazy.uploadOrQueue,
|
|
});
|
|
// Sticky preference (Settings → Preferences): issue-detail pins this
|
|
// composer to the bottom of the scroll viewport when enabled.
|
|
const sticky = useCommentComposerStore((s) => s.sticky);
|
|
|
|
// Draft persistence. Hydrate from store on mount via `defaultValue` above
|
|
// (ContentEditorRef has no setContent, so this is the only injection point).
|
|
// Flush on every onUpdate (debounced upstream) + visibilitychange/pagehide
|
|
// so tab close / mobile background doesn't lose work. Cleared on submit.
|
|
const setDraft = useCommentDraftStore((s) => s.setDraft);
|
|
useEffect(() => {
|
|
const flush = () => {
|
|
const md = editorRef.current?.getMarkdown();
|
|
if (md && md.trim().length > 0) setDraft(draftKey, md);
|
|
};
|
|
const onVis = () => { if (document.visibilityState === "hidden") flush(); };
|
|
document.addEventListener("visibilitychange", onVis);
|
|
window.addEventListener("pagehide", flush);
|
|
return () => {
|
|
document.removeEventListener("visibilitychange", onVis);
|
|
window.removeEventListener("pagehide", flush);
|
|
};
|
|
}, [draftKey, setDraft]);
|
|
|
|
useEffect(() => {
|
|
setSuppressedAgentIds(new Set());
|
|
}, [issueId]);
|
|
|
|
useEffect(() => {
|
|
const visible = new Set(triggerPreview.agents.map((agent) => agent.id));
|
|
setSuppressedAgentIds((prev) => {
|
|
const next = new Set([...prev].filter((id) => visible.has(id)));
|
|
return next.size === prev.size ? prev : next;
|
|
});
|
|
}, [triggerPreview.agents]);
|
|
|
|
const toggleSuppressedAgent = useCallback((agentId: string) => {
|
|
setSuppressedAgentIds((prev) => {
|
|
const next = new Set(prev);
|
|
if (next.has(agentId)) next.delete(agentId);
|
|
else next.add(agentId);
|
|
return next;
|
|
});
|
|
}, []);
|
|
|
|
// Await-then-render send (MUL-5181): the shared hook reads the markdown,
|
|
// guards empty/in-flight, re-checks the upload gate, locks + spins via
|
|
// `submitting`, and clears only once the server accepts — a failed send keeps
|
|
// the draft instead of silently dropping it.
|
|
// Stale-submit guard (MUL-5181 P0): if this composer unmounts mid-submit
|
|
// (issue detail closed) and the user reopens and types a new draft under the
|
|
// same key, the late success may only clear the draft it submitted.
|
|
const mountedRef = useRef(true);
|
|
useEffect(() => {
|
|
mountedRef.current = true;
|
|
return () => {
|
|
mountedRef.current = false;
|
|
};
|
|
}, []);
|
|
const submittedEntryRef = useRef<unknown>(null);
|
|
// Set by `onAccepted` only on the branch that actually wiped the editor; read
|
|
// back by `afterAccepted`. Bound this way because the stale-submit guard can
|
|
// leave a mid-flight draft in place: dropping the caret then would yank it
|
|
// out of the sentence the user is still typing.
|
|
const editorScrubbedRef = useRef(false);
|
|
|
|
const { submitting, submit } = useComposerSubmit({
|
|
editorRef,
|
|
uploadGate: gate,
|
|
// A top-level comment ends a turn: the caret is dropped rather than kept,
|
|
// so the composer stops reading as "still writing" once the comment is
|
|
// posted above it. Thread replies are the opposite — see ReplyInput.
|
|
afterAccepted: () => (editorScrubbedRef.current ? "blur" : "none"),
|
|
onSubmit: (content) => {
|
|
editorScrubbedRef.current = false;
|
|
// Flush the editor's pending debounce before snapshotting — a late flush
|
|
// of pre-submit typing must not read as an edit made during the request.
|
|
const pending = editorRef.current?.flushPendingUpdate?.();
|
|
if (pending != null) setDraft(draftKey, pending);
|
|
submittedEntryRef.current = useCommentDraftStore.getState().drafts[draftKey];
|
|
// Bind only uploads the BODY still references (MUL-5181): deleting an
|
|
// inline image really unbinds it. Uploads that finished after a close
|
|
// are written back into the body by the settle handler, so surviving
|
|
// files are referenced too — never silently attached.
|
|
const activeIds = pendingAttachments
|
|
.filter((a) => contentReferencesAttachment(content, a))
|
|
.map((a) => a.id);
|
|
const suppressAgentIds = triggerPreview.agents
|
|
.filter((agent) => suppressedAgentIds.has(agent.id))
|
|
.map((agent) => agent.id);
|
|
return onSubmit(
|
|
content,
|
|
activeIds.length > 0 ? activeIds : undefined,
|
|
suppressAgentIds.length > 0 ? suppressAgentIds : undefined,
|
|
);
|
|
},
|
|
onAccepted: () => {
|
|
// Success may only consume the entry it submitted (MUL-5181 P0): edits
|
|
// made while the request was in flight — or by a reopened composer after
|
|
// this one unmounted — survive both in the store and in the editor.
|
|
// Flush the pending debounce first so typing still inside the window is
|
|
// judged correctly (a no-op flush preserves entry identity).
|
|
const lateMd = editorRef.current?.flushPendingUpdate?.();
|
|
if (lateMd != null) setDraft(draftKey, lateMd);
|
|
const store = useCommentDraftStore.getState();
|
|
const live = store.drafts[draftKey];
|
|
const untouched = live === undefined || live === submittedEntryRef.current;
|
|
if (untouched) store.clearDraft(draftKey);
|
|
if (!mountedRef.current || !untouched) return;
|
|
editorRef.current?.clearContent();
|
|
setContent("");
|
|
setIsEmpty(true);
|
|
setSuppressedAgentIds(new Set());
|
|
editorScrubbedRef.current = true;
|
|
},
|
|
});
|
|
|
|
return (
|
|
<div
|
|
{...dropZoneProps}
|
|
className="relative flex flex-col rounded-lg bg-card pb-8 ring-1 ring-border"
|
|
>
|
|
{/* Lock the editor while the send is in flight. ContentEditor can't
|
|
toggle Tiptap's `editable` post-mount (see its docstring), so the
|
|
documented way to make it non-interactive is a pointer-events-none +
|
|
dimmed wrapper. */}
|
|
{lazy.active && (
|
|
<div
|
|
className={cn(
|
|
"flex-1 min-h-0 overflow-y-auto px-3 py-2",
|
|
// Pinned to the viewport bottom the composer grows upward; cap it
|
|
// so a long draft can't swallow the whole timeline (the editor
|
|
// area scrolls internally instead).
|
|
sticky && "max-h-[40vh]",
|
|
submitting && "pointer-events-none opacity-60",
|
|
!lazy.ready && "hidden",
|
|
)}
|
|
aria-busy={submitting || undefined}
|
|
>
|
|
<ContentEditor
|
|
ref={editorRef}
|
|
defaultValue={initialDraft}
|
|
onReady={lazy.onReady}
|
|
placeholder={t(($) => $.comment.leave_comment_placeholder)}
|
|
onUpdate={(md) => {
|
|
setContent(md);
|
|
setIsEmpty(!md.trim());
|
|
// Debounced upstream (debounceMs=100). Persist on every tick so a
|
|
// reload or scroll-out-of-viewport restores work to the keystroke.
|
|
// setDraft keeps any pending attachments and drops the entry only
|
|
// when text AND attachments are both empty.
|
|
setDraft(draftKey, md);
|
|
}}
|
|
onSubmit={submit}
|
|
onUploadFile={handleUpload}
|
|
onUploadingChange={uploadGate.onUploadingChange}
|
|
debounceMs={100}
|
|
currentIssueId={issueId}
|
|
attachments={pendingAttachments}
|
|
enableSlashCommands
|
|
slashCommandMode="command"
|
|
/>
|
|
</div>
|
|
)}
|
|
{/* Static shell — visually clones the empty single-line composer.
|
|
Real editor mounts (hidden) on first intent; shell stays visible
|
|
until it's ready so the card never blanks or shifts. */}
|
|
{!lazy.ready && (
|
|
<div
|
|
data-testid="comment-composer-shell"
|
|
role="button"
|
|
tabIndex={0}
|
|
aria-label={t(($) => $.comment.leave_comment_placeholder)}
|
|
className="flex-1 min-h-0 cursor-text px-3 py-2"
|
|
onClick={() => lazy.activate()}
|
|
onKeyDown={(e) => {
|
|
if (e.key === "Enter" || e.key === " ") {
|
|
e.preventDefault();
|
|
lazy.activate();
|
|
}
|
|
}}
|
|
>
|
|
{/* rich-text-editor + <p>: the shell line inherits the editor's
|
|
exact type metrics (line-height 1.625 from prose.css), so the
|
|
shell→editor swap doesn't shift layout. */}
|
|
<div className="rich-text-editor text-body">
|
|
<p className="text-muted-foreground">{t(($) => $.comment.leave_comment_placeholder)}</p>
|
|
</div>
|
|
</div>
|
|
)}
|
|
<div className="absolute bottom-1 left-2 right-28 min-w-0">
|
|
<CommentTriggerChips
|
|
agents={triggerPreview.agents}
|
|
blocked={triggerPreview.blocked}
|
|
draftContent={content}
|
|
suppressedAgentIds={suppressedAgentIds}
|
|
onToggle={toggleSuppressedAgent}
|
|
/>
|
|
</div>
|
|
<div className="absolute bottom-1 right-1.5 flex items-center gap-1">
|
|
<FileUploadButton
|
|
size="sm"
|
|
multiple
|
|
onSelect={(file) => lazy.uploadOrQueue([file])}
|
|
/>
|
|
<SubmitButton
|
|
onClick={submit}
|
|
disabled={isEmpty}
|
|
loading={submitting}
|
|
busy={gate.uploading}
|
|
tooltip={gate.uploading
|
|
? tEditor(($) => $.upload.in_progress)
|
|
: sendShortcut
|
|
? `${t(($) => $.comment.send_tooltip)} · ${formatShortcut(sendShortcut)}`
|
|
: t(($) => $.comment.send_tooltip)}
|
|
ariaLabel={gate.uploading
|
|
? tEditor(($) => $.upload.in_progress)
|
|
: t(($) => $.comment.send_tooltip)}
|
|
/>
|
|
</div>
|
|
{isDragOver && <FileDropOverlay />}
|
|
</div>
|
|
);
|
|
}
|
|
|
|
export { CommentInput };
|