Copy all selected thoughts when Select All (multicursor) is active by Copilot · Pull Request #4438 · cybersemics/em

Copilot · 2026-06-23T15:04:56Z

On desktop, copying with Select All active only copied the focused thought even though the alert reported the full count ("Copied 3 thoughts"), so pasting yielded a single thought instead of the whole selection.

Root cause

copyCursor runs with permitDefault: true, so preventDefault() is never called. After the command writes the full multicursor selection to the clipboard, the browser's native copy event still fires on the focused editable, invoking useOnCopy, which exported only the single focused thought — overwriting the clipboard. This is desktop-only because headless Chrome does not emit a copy event for a collapsed contenteditable selection.

On Desktop Safari the same fix was insufficient: the structured clipboard formats (text/html and the text/em marker) were only written by the native copy event handler, but Safari (like headless Chrome) does not fire a copy event for a collapsed contenteditable selection. As a result only plain text reached the clipboard, and on paste Safari synthesized its own generic text/html that shadowed the clean plain text in importData — so the "Importing from clipboard..." alert flashed but nothing was pasted.

On Mobile Safari the rich-copy path could not run at all: it depends on a genuine, user-initiated Cmd+C copy event, but the copy is triggered by a Command Center tap (no keyboard), so no native copy event ever fires, and Safari ignores setData() during a programmatic copy. The rich-copy listener therefore waited, timed out, and wrote nothing — so nothing pasted.

Changes

src/selectors/getMulticursorThoughtIds.ts (new): returns the multicursor selection as document-ordered ThoughtId[], deduping descendants when an ancestor is also selected.
src/components/Editable/useOnCopy.ts: when hasMulticursor(state), export all selected thoughts (text/plain, text/html, text/em) and return early; single-thought path is unchanged.
src/commands/copyCursor.ts: reuse getMulticursorThoughtIds in execMulticursor in place of the inline dedup; pass the exported html to copy().
src/device/copy.ts: copy(text, { html }) now writes text/plain + text/html + text/em deterministically via a one-shot document copy listener (triggered through ClipboardJS's execCommand('copy')), independent of the native editable copy event. This makes structured paste work on browsers that do not fire a native copy event for a collapsed selection (Desktop Safari, headless Chrome). The native Capacitor and plain-only paths are unchanged, and selection is still saved/restored. On mobile Safari (isSafari() && isTouch), copy() skips the rich path and uses the plain-text ClipboardJS path, which runs synchronously within the tap gesture: the full multicursor selection is still copied as indented plain text that em reconstructs into the thought tree on paste (thought structure is preserved; inline formatting is not, since text/html cannot be written without a real copy event on iOS).
src/commands/__tests__/copyCursor.ts: updated assertions for the new copy() call signature.
src/e2e/puppeteer/__tests__/multiselect.ts: regression tests — one dispatching a synthetic copy event with multicursor active asserting the clipboard contains all selected thoughts, and one intercepting DataTransfer.setData to assert the copy command writes text/html + text/em for the full selection (verified to fail without the device-level fix).

// useOnCopy.ts
if (hasMulticursor(state)) {
  const ids = getMulticursorThoughtIds(state)
  e.clipboardData?.setData('text/plain', trimBullet(ids.map(id => strip(exportContext(state, id, 'text/plain'))).join('\n')))
  // ...text/html, text/em
  return
}

Co-authored-by: BayuAri <8419585+BayuAri@users.noreply.github.com>

BayuAri

Issue is partially fixed.
It works on Desktop Chrome but there is an issue on Desktop Safari.
Importing from clpboard... alert is flashed after Cmd + V but the copied thoughts are not pasted as it is on Chrome

@copilot
Please fix this issue on Desktop Safari

… copy (#3993) Co-authored-by: BayuAri <8419585+BayuAri@users.noreply.github.com>

Copilot · 2026-06-24T06:11:06Z

Issue is partially fixed.
It works on Desktop Chrome but there is an issue on Desktop Safari....

Fixed in 41ee407. The root cause was that the structured clipboard formats (text/html and the text/em marker) were only written by the native copy event handler — but Safari (like headless Chrome) does not fire a copy event for a collapsed contenteditable selection, so only plain text reached the clipboard. On paste, Safari then synthesized its own generic text/html from the plain text, which shadowed the clean text in the importer and broke structured paste.

The device-level copy() now writes text/plain + text/html + text/em deterministically, independent of the native copy event, so multicursor copy/paste now works on Safari as well as Chrome.

Safari only honors clipboard setData() in a genuine user-initiated copy event, and dispatches the Cmd+C copy event on a later task than setTimeout(0). With a multicursor no editable is focused, so the copy event targets <body> and useOnCopy never runs; the browser then serializes the empty collapsed selection and the previous async write clobbered it, leaving an empty clipboard. Register a capture-phase document copy listener that intercepts the late user-initiated copy event regardless of focus target, preventDefault()s the empty serialization, and writes text/plain + text/html + text/em via setData (honored by Safari). The listener self-removes on the copy event and is cleaned up after a generous window otherwise. Non-Safari browsers keep the proven execCommand+setData path used by the e2e suite. Add a selectNode() helper used by the non-Safari rich copy path.

BayuAri · 2026-06-24T14:57:48Z

Safari multicursor copy/paste fix (#3993)

Problem

The original fix resolved multicursor copy/paste on desktop Chrome, but paste was still broken on Safari — pasting produced nothing, and the clipboard contained only an empty <!DOCTYPE html> (15 chars) with no text/em marker.

Root cause (established empirically on real Safari)

Safari's clipboard behaves very differently from Chrome:

It ignores setData() during a programmatic execCommand('copy').
navigator.clipboard.write([ClipboardItem]) with text/html resolves but sanitizes the content to an empty document.
It only honors setData() in a genuine, user-initiated copy event (the real Cmd+C default action).
Crucially, Safari dispatches that Cmd+C copy event on a later task than setTimeout(0).

copyCursor runs on the Cmd+C keydown with permitDefault, so the browser fires its own copy event. But with a multicursor no editable is focused, so the event targets <body> and the Editable's useOnCopy handler never runs. The browser then serializes the empty collapsed selection, and the previous async copy() write resolved afterward and clobbered the clipboard — leaving it empty.

Solution

On Safari, copy() now registers a capture-phase document copy listener that:

Intercepts the late, user-initiated Cmd+C copy event regardless of which element is focused (works even when the event targets <body>).
Calls preventDefault() to block the empty serialization.
Writes text/plain + text/html + text/em via setData() — honored by Safari because the event is genuinely user-initiated.

The listener self-removes on the copy event and is cleaned up after a generous window otherwise (so it can't affect an unrelated later copy). Non-Safari browsers (Chrome + the Puppeteer e2e suite) keep the proven execCommand+setData rich-copy path.

Verification

Tested live on Safari across many attempts — copy/paste is now consistent: every paste arrives with isEmText: true, full text/plain + text/html, and the text/em marker. copyCursor unit tests (6/6) pass; lint, formatting, and type checks are clean.

Known separate issue (not addressed here)

When a copied selection contains an empty thought (<li></li> in the export), the HTML importer (htmlToJson.ts) throws TypeError: ... blockOrArray[0].scope because it reads the "first item" of an empty group. This is a distinct, pre-existing html-import bug, unrelated to the clipboard path, and will be addressed separately after merge.
Logged as #4448

BayuAri · 2026-06-25T00:45:17Z

Test Result

Desktop Chrome
Desktop Safari
Mobile Chrome
Mobile Safari

@copilot
Please fix the issue on Mobile Safari.
Nothing happens when pasting the copied select all thoughts

Step to Reproduce

Open em on Mobile Safari
Create some thought
Draw ←↓→ to select all the thoughts
Tap gear icon on the Command Center to Copy All
Create a new subthought
Paste the copied thought

Current behavior

Nothing happens

Expected behavior

Copied thoughts should be pasted

…3993) Co-authored-by: BayuAri <8419585+BayuAri@users.noreply.github.com>

Copilot · 2026-06-25T01:04:02Z

Test Result

Desktop Chrome
...

Fixed in 4b3d61b.

Root cause on Mobile Safari: the rich copy path (copyRichSafari) only writes the clipboard from inside a genuine Cmd+C copy event. On mobile the copy is triggered by a Command Center tap, so no native copy event ever fires, and Safari ignores setData() during a programmatic copy — so copyRichSafari waited, timed out after 1s, and wrote nothing to the clipboard (hence nothing pasted).

Mobile Safari now routes to the plain-text ClipboardJS execCommand('copy') path (isSafari() && isTouch), which runs synchronously within the tap gesture. The full multicursor selection is still copied — as indented plain text that em reconstructs into the thought tree on paste. This is the same reliable path mobile used before #3993.

Note: on Mobile Safari this preserves the full thought structure but not inline formatting (bold/italic/color), since that requires text/html, which can't be written without a real copy event on iOS. Desktop Safari (Cmd+C) and Chrome are unchanged and keep the rich, formatting-preserving path.

BayuAri

Test Result

Desktop Chrome
Desktop Safari
Mobile Chrome
Mobile Safari

Confirm all platforms are working as expected.

raineorshine

This touches a lot of the copy infrastructure which may affect other cases. Before I review, could you test the following:

Copy thoughts with subthoughts.
Copy thoughts and subthoughts with different colors.

These are the cases where a regression is most likely to occur.

Note that copying colored thoughts is already a bit broken on main, so if you encounter one of those cases you can just note it and I will make a judgment call about whether I think it is fully independent of this fix.

Thanks!

BayuAri · 2026-07-09T01:51:50Z

Copy thoughts with subthoughts.

Copy thoughts and subthoughts with different colors.

@raineorshine
Above use cases are working properly.
There is a known issue on Formatting and color are not working on Mobile and will be handed on #3960

I have a question about how the Copied alert modal when the Selected modal is on, can you help me with the expected result of this behavior, please?

Suggestion:

Multicursor selection state and alert are dismissed and Copied alert is prompted for 5 secs afterward.
Multicursor selection remain, Selected alert is dismissed, Copied alert is prompted for 1-2 secs and after that Selected alert reappear because the selection is

The expected result may relate to #3612

Currently, the Copied alert flashes in milli second

Flashing.Copied.Alert.when.Selected.Alert.is.onscreen.mov

raineorshine · 2026-07-09T02:01:58Z

Copy thoughts with subthoughts.

Copy thoughts and subthoughts with different colors.

@raineorshine Above use cases are working properly. There is a known issue on Formatting and color are not working on Mobile and will be handed on #3960

Thanks!

Suggestion:

Multicursor selection state and alert are dismissed and Copied alert is prompted for 5 secs afterward.

Multicursor selection remain, Selected alert is dismissed, Copied alert is prompted for 1-2 secs and after that Selected alert reappear because the selection is

Actually the new Liminal UI for multiselect won't have this conflict, because the multicursor selection state will no longer be shown in an alert. So we can leave the broken alert behavior for now.

raineorshine · 2026-07-09T18:36:10Z

+      // When a multicursor is active, copy all selected thoughts, not just the focused one.
+      // The Copy Cursor command also writes the full selection to the clipboard, but copyCursor uses
+      // permitDefault, so this native copy event fires afterward and would otherwise clobber it with
+      // only the focused thought (#3993).
+      if (hasMulticursor(state)) {
+        const ids = getMulticursorThoughtIds(state)
+        const plainText = trimBullet(ids.map(id => strip(exportContext(state, id, 'text/plain'))).join('\n'))
+        const htmlText = ids.map(id => exportContext(state, id, 'text/html')).join('\n')
+        clipboardData.setData('text/plain', plainText)
+        clipboardData.setData('text/html', htmlText)
+        clipboardData.setData('text/em', 'true')
+        return
+      }
+
      const currentText = selection.text()
      const currentHtml = selection.html()
-      const clipboardData = event.clipboardData
      clipboardData.setData('text/plain', currentText!)
      clipboardData.setData('text/html', currentHtml!)

      if (!currentText) {
        const thoughtHtml = exportContext(state, thoughtId, 'text/html')
        const thoughtText = exportContext(state, thoughtId, 'text/plain')
        clipboardData.setData('text/plain', trimBullet(thoughtText))
        clipboardData.setData('text/html', thoughtHtml)
      }
      clipboardData.setData('text/em', 'true')


There is a lot of duplication between these different blocks. Instead of doing this with separate blocks with multiple return statements, funnel all the logic into a single return statement at the end. Use ternary expressions to reduce conditional boilerplate.

Refactored: the handler now computes plainText and htmlText via ternaries (multicursor → all selected thoughts; else current selection, falling back to the focused thought when empty) and writes them in a single set of setData calls, removing the separate blocks and early return.

raineorshine · 2026-07-09T18:40:46Z

+  // onCopy self-removes on the first copy event (setting done); this timeout only removes the listener if no
+  // copy event ever arrives (e.g. the command was triggered from the command palette) so it cannot affect an
+  // unrelated later copy.
+  setTimeout(() => {


Does this handle the case where copyRichSafari is called multiple times within the 1s window? I don't see anything that cancels the timeout.

Good catch — it didn't. copyRichSafari now tracks the pending listener + timeout in a module-level pendingSafariCopy and calls clearPendingSafariCopy() at the start of each invocation, so a second call within the 1s window cancels the prior listener and its timeout before registering the new one. The timeout and the copy handler both route through clearPendingSafariCopy.

raineorshine · 2026-07-09T18:44:30Z

@copilot Please address the review feedback.

…afari copy timeout (#3993) Co-authored-by: raineorshine <750276+raineorshine@users.noreply.github.com>

github-actions · 2026-07-09T18:56:17Z

Vercel preview: https://em-6dojfejxb-cybersemics.vercel.app

Initial plan

316b87a

Copilot AI assigned Copilot and BayuAri Jun 23, 2026

Copilot started work on behalf of BayuAri June 23, 2026 15:05 View session

Copilot AI linked an issue Jun 23, 2026 that may be closed by this pull request

[Desktop] Copying when Select All is copying the active cursor #3993

Closed

vercel Bot deployed to Preview – em June 23, 2026 15:07 View deployment

Copilot AI and others added 2 commits June 23, 2026 15:22

test(e2e): add failing regression test for copy with Select All (#3993)

48db9d1

Co-authored-by: BayuAri <8419585+BayuAri@users.noreply.github.com>

test(e2e): add failing regression test for copy with Select All (#3993)

afabe81

Co-authored-by: BayuAri <8419585+BayuAri@users.noreply.github.com>

vercel Bot deployed to Preview – em June 23, 2026 15:27 View deployment

fix(copy): copy all selected thoughts when Select All is active (#3993)

9f04877

Co-authored-by: BayuAri <8419585+BayuAri@users.noreply.github.com>

vercel Bot deployed to Preview – em June 23, 2026 15:35 View deployment

Copilot AI changed the title ~~[WIP] Fix copying when Select All copies the active cursor~~ Copy all selected thoughts when Select All (multicursor) is active Jun 23, 2026

Copilot finished work on behalf of BayuAri June 23, 2026 15:44

Copilot AI requested a review from BayuAri June 23, 2026 15:44

BayuAri reviewed Jun 24, 2026

View reviewed changes

Copilot started work on behalf of BayuAri June 24, 2026 05:36 View session

fix(copy): write html + em marker to clipboard for Safari multicursor…

41ee407

… copy (#3993) Co-authored-by: BayuAri <8419585+BayuAri@users.noreply.github.com>

vercel Bot deployed to Preview – em June 24, 2026 06:11 View deployment

Copilot finished work on behalf of BayuAri June 24, 2026 06:12

Copilot AI requested a review from BayuAri June 24, 2026 06:12

vercel Bot deployed to Preview – em June 24, 2026 14:58 View deployment

Copilot started work on behalf of BayuAri June 25, 2026 00:45 View session

fix(copy): route mobile Safari multicursor copy to plain-text path (#…

4b3d61b

…3993) Co-authored-by: BayuAri <8419585+BayuAri@users.noreply.github.com>

vercel Bot deployed to Preview – em June 25, 2026 00:59 View deployment

Copilot finished work on behalf of BayuAri June 25, 2026 01:05

BayuAri reviewed Jun 25, 2026

View reviewed changes

BayuAri marked this pull request as ready for review June 25, 2026 01:56

BayuAri requested a review from raineorshine June 25, 2026 01:56

BayuAri mentioned this pull request Jun 25, 2026

Empty thought in a series of copied thoughts is not pasted on the destination #4448

Closed

Copilot AI mentioned this pull request Jun 25, 2026

fix(htmlToJson): preserve empty thoughts when pasting copied thoughts #4451

Merged

raineorshine reviewed Jul 7, 2026

View reviewed changes

BayuAri marked this pull request as draft July 9, 2026 01:54

raineorshine marked this pull request as ready for review July 9, 2026 02:02

BayuAri mentioned this pull request Jul 9, 2026

Prevent multiselect indicator alert from auto-dismissing on desktop #4546

Open

raineorshine requested changes Jul 9, 2026

View reviewed changes

Copilot started work on behalf of raineorshine July 9, 2026 18:45 View session

refactor(copy): funnel useOnCopy into single return; cancel pending S…

a186336

…afari copy timeout (#3993) Co-authored-by: raineorshine <750276+raineorshine@users.noreply.github.com>

vercel Bot deployed to Preview – em July 9, 2026 18:55 View deployment

Copilot finished work on behalf of raineorshine July 9, 2026 19:00

Copilot AI requested a review from raineorshine July 9, 2026 19:00

raineorshine approved these changes Jul 9, 2026

View reviewed changes

raineorshine merged commit a84cfca into main Jul 9, 2026
13 checks passed

raineorshine deleted the copilot/fix-copying-active-cursor-issue branch July 9, 2026 19:00

Uh oh!

Conversation

Copilot AI commented Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Root cause

Changes

Uh oh!

BayuAri left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI commented Jun 24, 2026

Uh oh!

BayuAri commented Jun 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Safari multicursor copy/paste fix (#3993)

Problem

Root cause (established empirically on real Safari)

Solution

Verification

Known separate issue (not addressed here)

Uh oh!

BayuAri commented Jun 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Step to Reproduce

Current behavior

Expected behavior

Uh oh!

Copilot AI commented Jun 25, 2026

Uh oh!

BayuAri left a comment

Choose a reason for hiding this comment

Uh oh!

raineorshine left a comment

Choose a reason for hiding this comment

Uh oh!

BayuAri commented Jul 9, 2026

Uh oh!

raineorshine commented Jul 9, 2026

Uh oh!

raineorshine Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

raineorshine Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jul 9, 2026

Choose a reason for hiding this comment

Uh oh!

raineorshine commented Jul 9, 2026

Uh oh!

github-actions Bot commented Jul 9, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Copilot AI commented Jun 23, 2026 •

edited

Loading

BayuAri commented Jun 24, 2026 •

edited

Loading

BayuAri commented Jun 25, 2026 •

edited

Loading