🎨 Improved child display in thread sidebar#26250
Conversation
📝 WalkthroughWalkthroughAdds a thread-focused comment fetch hook and centralizes thread link construction; refactors frontend to drive thread-view navigation via generated links (removing callback-based navigation), renames backend relation from Changes
Sequence DiagramsequenceDiagram
participant User
participant UI as Comments UI
participant Router as URL/Navigation
participant Sidebar as CommentThreadSidebar
participant API as useThreadComments
participant List as CommentThreadList
User->>UI: Click replies metric
UI->>UI: buildThreadLink(searchParams, commentId)
UI->>Router: Navigate with thread param
Router->>Sidebar: Mount with thread param
Sidebar->>API: Fetch threadComments(commentId)
API-->>Sidebar: Return threadReplies
Sidebar->>List: render(selectedComment, selectedCommentId, threadReplies)
List-->>User: Show selected comment + replies
Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 1 | ❌ 2❌ Failed checks (1 warning, 1 inconclusive)
✅ Passed checks (1 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
E2E Tests FailedTo view the Playwright test report locally, run: REPORT_DIR=$(mktemp -d) && gh run download 21691597768 -n playwright-report -D "$REPORT_DIR" && npx playwright show-report "$REPORT_DIR" |
kevinansfield
force-pushed
the
issue/BER-3276-improve-child-display-in-thread-sidebar
branch
from
d4307af to
bd9a166
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
apps/posts/src/views/comments/components/comment-thread-sidebar.tsx (1)
55-75:⚠️ Potential issue | 🔴 CriticalFix undefined
parentCommentreference (runtime error).
parentCommentis no longer defined after the selected-comment refactor, so this block will throw. UseselectedCommentinstead.🛠️ Proposed fix
- {parentComment?.post && ( + {selectedComment?.post && ( <> <div className="flex items-center gap-4"> <div className="min-w-0 flex-1"> <h3 className="line-clamp-1 text-xl font-semibold text-foreground"> - {parentComment.post.title} + {selectedComment.post.title} </h3> - {parentComment.post.excerpt && ( + {selectedComment.post.excerpt && ( <p className="mt-1 line-clamp-2 text-sm text-muted-foreground"> - {parentComment.post.excerpt} + {selectedComment.post.excerpt} </p> )} </div> - {parentComment.post.feature_image && ( + {selectedComment.post.feature_image && ( <img - alt={parentComment.post.title || 'Post feature image'} + alt={selectedComment.post.title || 'Post feature image'} className="hidden aspect-video h-18 shrink-0 rounded object-cover lg:block" - src={parentComment.post.feature_image} + src={selectedComment.post.feature_image} /> )} </div> <Separator className="-mx-6 my-6 w-auto" /> </> )}
🤖 Fix all issues with AI agents
In `@apps/posts/src/views/comments/components/comment-thread-list.tsx`:
- Around line 37-40: The UI reads comment.count.direct_replies (used by
hasReplies and metrics) but the thread query only requests count.replies,
leaving direct_replies undefined and inflating nested counts; update the query
used by useThreadComments to request count.direct_replies (in addition to
count.replies) so comment.count.direct_replies is populated, then verify
components referencing hasReplies, comment.count.direct_replies and
comment.count.replies still behave correctly; locate the query definition in
useThreadComments and the API query helper in
apps/admin-x-framework/src/api/comments.ts to add the direct_replies field.
In `@ghost/core/core/server/models/comment.js`:
- Around line 45-47: Update the unit test fixtures that mock Comment model JSON
to use snake_case in_reply_to instead of camelCase inReplyTo: locate the
serializer/mapper tests that supply mocked Comment objects to the mapper (the
tests referencing the mapper) and change any inReplyTo properties to in_reply_to
so they match the Bookshelf toJSON output and the mapper's expected in_reply_to
field (the relation method in_reply_to() is already renamed correctly).
🧹 Nitpick comments (2)
e2e/helpers/pages/admin/comments/comments-page.ts (2)
19-42: Potential selector conflict betweenthreadSidebaranddisableCommentsModal.Both
threadSidebar(line 40) anddisableCommentsModal(line 34) usepage.getByRole('dialog'). If both dialogs could theoretically be present or if tests interact with both, this could cause flaky behavior.Consider differentiating using a more specific selector or
data-testid:- this.threadSidebar = page.getByRole('dialog'); + this.threadSidebar = page.getByRole('dialog', {name: 'Thread'});Alternatively, if the Ghost codebase doesn't have distinguishing attributes on these dialogs, consider suggesting a
data-testid="thread-sidebar"be added to the component.
90-96: Parent traversal in locators may be brittle.The
getRepliedToLinkandgetMainListRepliedToLinkmethods use.locator('..')for parent traversal, which couples tests to DOM structure. This could break if the markup changes.If this becomes problematic, consider requesting a
data-testidon the replied-to link container in the Ghost codebase for more stable selection.
apps/posts/src/views/comments/components/comment-thread-list.tsx
Outdated
Show resolved
Hide resolved
jonatansberg
left a comment
jonatansberg
left a comment
There was a problem hiding this comment.
Looks good! Added a few comments that we could either adress now or treat as followups if we want to get this merged pronto.
| searchParams: { | ||
| filter: `(parent_id:${commentId}+in_reply_to_id:null),in_reply_to_id:${commentId}`, | ||
| order: 'created_at asc', | ||
| include: 'member,post,count.replies,count.likes,count.reports,parent,in_reply_to', |
There was a problem hiding this comment.
Shouldn't this say count.direct_replies?
There was a problem hiding this comment.
Yes, that would explain an inconsistency I saw too!
| icon: React.ReactNode; | ||
| count: number; | ||
| label: string; | ||
| link?: string; |
There was a problem hiding this comment.
Nit: I'd probably name this to to align with the React Router Link conventions.
| // Use direct_replies for checking if a reply has further replies (reply-to-replies) | ||
| // Use replies or total_replies for root comments, direct_replies for reply comments | ||
| const hasReplies = (comment.replies?.length ?? 0) > 0 || (comment.count?.direct_replies ?? comment.count?.replies ?? 0) > 0; |
There was a problem hiding this comment.
Comment is outdated (there aren't any total_replies)
| /> | ||
|
|
||
| {comment.in_reply_to_snippet && !isReply && ( | ||
| {comment.in_reply_to_snippet && (isSelectedComment || (comment.in_reply_to_id && comment.in_reply_to_id !== selectedCommentId)) && ( |
There was a problem hiding this comment.
Can we name this (combination of) criteria something so that the intent here is clearer?
There was a problem hiding this comment.
This could be simplified further I think because we're only showing direct replies in the sidebar now. It's only the selected comment that needs the replied-to line
| await expect(commentsPage.getThreadCommentByText('Reply to first level reply')).toBeVisible(); | ||
| }); | ||
|
|
||
| test('url structure uses only thread param without reply_to', async ({page}) => { |
There was a problem hiding this comment.
This test name makes no sense other than in the context of an agent conversation. I think we can scrap the full test case?
| await expect(selectedComment.getByText('Replied to:')).toBeVisible(); | ||
| }); | ||
|
|
||
| test('shows replied to line for non-direct children in thread', async ({page}) => { |
There was a problem hiding this comment.
I read that as "shows that line on the left" not that it shows the "reply to context"
| await expect(nonDirectReply.getByText('Replied to:')).toBeVisible(); | ||
| }); | ||
|
|
||
| test('hides replied to line for direct children in thread', async ({page}) => { |
There was a problem hiding this comment.
Same here, maybe we can merge this with the test above?
| const replyToComment = jsonModel.inReplyTo || (!isPublicRequest && jsonModel.parent_id && jsonModel.parent); | ||
| // Use in_reply_to if available, otherwise fall back to parent for first-level replies only | ||
| // (first-level replies have parent_id but no in_reply_to_id) | ||
| const replyToComment = jsonModel.in_reply_to || (!isPublicRequest && jsonModel.parent_id && !jsonModel.in_reply_to_id && jsonModel.parent); |
There was a problem hiding this comment.
Do we need this extra condition? Seems like we would already hit the in_reply_to on the left hand side
There was a problem hiding this comment.
Claude thinks yes but there could still be something to tidy up here 🤔
The in_reply_to relation is based on in_reply_to_id (see comment.js:45-46):
in_reply_to() {
return this.belongsTo('Comment', 'in_reply_to_id');
},
For first-level replies (direct replies to a top-level comment), in_reply_to_id is NULL. They only have a parent_id. So in_reply_to will also be null for these.
The fallback handles this case for admin requests:
- !isPublicRequest - admin only
- jsonModel.parent_id - it's a reply
- !jsonModel.in_reply_to_id - it's a first-level reply (not a nested reply)
- jsonModel.parent - use the parent comment to generate the snippet
Without this fallback, first-level replies in admin would have no in_reply_to_snippet, even though they're replies to something. The admin UI wants to show "replying to..." snippets for ALL replies, not just nested ones.
The
&& !jsonModel.in_reply_to_idcondition isn't needed though? We've already checked that on the left side of the OR
Good question, but the check is still needed. Here's the distinction:
in_reply_to being falsy doesn't mean in_reply_to_id is null. Consider:
- in_reply_to_id is null → in_reply_to is null (first-level reply)
- in_reply_to_id exists but referenced comment was deleted → in_reply_to could be null
- in_reply_to_id exists but relation wasn't loaded → in_reply_to is undefined/null
The !jsonModel.in_reply_to_id check protects against cases 2 and 3. If someone replied to a nested comment that was later deleted, we shouldn't fall back to showing the parent (root) comment's snippet - that would be misleading since they didn't reply to the root.
The check ensures we only use the parent fallback for true first-level replies where in_reply_to_id was never set in the first place.
ref https://linear.app/tryghost/issue/BER-3276 Follows the convention used by other model relations (snake_case)
kevinansfield
force-pushed
the
issue/BER-3276-improve-child-display-in-thread-sidebar
branch
2 times, most recently
from
bdf119b to
f40c0fb
Compare
There was a problem hiding this comment.
Actionable comments posted: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
apps/posts/src/views/comments/components/comment-thread-sidebar.tsx (1)
55-78:⚠️ Potential issue | 🔴 CriticalCritical bug:
parentCommentis undefined.The code references
parentComment(lines 55, 60, 62, 63, 64, 68, 70, 72) but this variable is never defined. Based on the refactoring that renamedparentCommenttoselectedComment(line 44), all these references need to be updated.This will cause a
ReferenceErrorat runtime, crashing the component.🐛 Proposed fix
- {parentComment?.post && ( + {selectedComment?.post && ( <> <div className="flex items-center gap-4"> <div className="min-w-0 flex-1"> <h3 className="line-clamp-1 text-xl font-semibold text-foreground"> - {parentComment.post.title} + {selectedComment.post.title} </h3> - {parentComment.post.excerpt && ( + {selectedComment.post.excerpt && ( <p className="mt-1 line-clamp-2 text-sm text-muted-foreground"> - {parentComment.post.excerpt} + {selectedComment.post.excerpt} </p> )} </div> - {parentComment.post.feature_image && ( + {selectedComment.post.feature_image && ( <img - alt={parentComment.post.title || 'Post feature image'} + alt={selectedComment.post.title || 'Post feature image'} className="hidden aspect-video h-18 shrink-0 rounded object-cover lg:block" - src={parentComment.post.feature_image} + src={selectedComment.post.feature_image} /> )} </div> <Separator className="-mx-6 my-6 w-auto" /> </> )}
🧹 Nitpick comments (2)
e2e/helpers/pages/admin/comments/comments-page.ts (2)
34-34: Potential selector collision betweenthreadSidebaranddisableCommentsModal.Both
disableCommentsModal(line 34) andthreadSidebar(line 40) usepage.getByRole('dialog'), which will match any dialog on the page. If both dialogs could be open simultaneously or if tests check visibility, this could cause flaky behavior.Consider using more specific selectors, such as adding a
data-testidto each dialog in the Ghost codebase or scoping by dialog title:- this.threadSidebar = page.getByRole('dialog'); + this.threadSidebar = page.getByRole('dialog', {name: 'Thread'});Also applies to: 40-40
90-92: DOM traversal with.locator('..')is fragile.Using
.locator('..')to traverse to the parent element is brittle and can break with minor markup changes. Consider adding adata-testidto the "Replied to" link container in the Ghost codebase for more resilient test selectors.
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Fix all issues with AI agents
In `@e2e/helpers/pages/admin/comments/comments-page.ts`:
- Around line 19-21: The thread sidebar locator is colliding with the disable
comments dialog because both use page.getByRole('dialog') unscoped; update the
threadSidebar locator (and any related selectors used by waitForThreadSidebar)
to scope the dialog to a unique ancestor or aria/name/test-id (e.g., use
page.locator('[data-testid="thread-sidebar"]').getByRole('dialog') or
page.getByRole('dialog', { name: 'Thread sidebar' }) so it only matches the
sidebar), and similarly verify and scope disableCommentsModal and cancelButton
selectors so they target the intended dialog to avoid cross-matching.
- Around line 90-92: The getRepliedToLink function uses both XPath parent
traversal (locator('..')) and a CSS selector ('a, button'); replace these with
semantic Playwright locators: in getRepliedToLink(row: Locator) use
row.getByRole('link', { name: /Replied to:/ }) or row.getByRole('button', {
name: /Replied to:/ }) to target the actionable element by its accessible name
(or if the accessible name isn't available, add a data-testid on the
parent/action in the app and use row.getByTestId('replied-to-link')); update the
implementation in the getRepliedToLink function (and any callers) to rely solely
on getByRole or getByTestId rather than locator('..') or locator('a, button').
| getRepliedToLink(row: Locator): Locator { | ||
| return row.getByText('Replied to:').locator('..').locator('a, button'); | ||
| } |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# First, find and examine the actual file
fd "comments-page.ts" --path "*/e2e/helpers/pages/*"Repository: TryGhost/Ghost
Length of output: 273
🏁 Script executed:
# Read the file around lines 90-92
head -100 e2e/helpers/pages/admin/comments/comments-page.ts | tail -20Repository: TryGhost/Ghost
Length of output: 689
🏁 Script executed:
# Check if there are more usages to understand context better
head -110 e2e/helpers/pages/admin/comments/comments-page.ts | tail -30Repository: TryGhost/Ghost
Length of output: 963
🏁 Script executed:
# Search for how this method is used in tests to understand the intended behavior
rg "getRepliedToLink" e2e/Repository: TryGhost/Ghost
Length of output: 281
🏁 Script executed:
# Check the actual HTML structure or tests to understand what element needs to be selected
cat -n e2e/tests/admin/comments/thread-sidebar.test.ts | grep -A 5 -B 5 "getRepliedToLink"Repository: TryGhost/Ghost
Length of output: 703
🏁 Script executed:
# Check if there's a data-testid or accessible role we could use instead
# Search for the HTML structure or context around "Replied to:" text
rg "Replied to:" e2e/ -A 3 -B 3Repository: TryGhost/Ghost
Length of output: 2911
Remove both CSS and XPath selectors; use only semantic locators.
The code uses two selector types that violate the semantic locator guideline:
locator('..')— an XPath selector for parent traversallocator('a, button')— a CSS selector
The suggested fix addresses the CSS selector but leaves the XPath selector. Use semantic locators throughout:
♻️ Improved semantic locator
getRepliedToLink(row: Locator): Locator {
- return row.getByText('Replied to:').locator('..').locator('a, button');
+ const repliedTo = row.getByText('Replied to:').locator('..');
+ return repliedTo.getByRole('link').or(repliedTo.getByRole('button'));
}If locator('..') cannot be replaced with a semantic alternative, consider adding data-testid to the parent element in the Ghost codebase.
As per coding guidelines, never use CSS or XPath selectors in page objects—only semantic locators or data-testid.
🤖 Prompt for AI Agents
In `@e2e/helpers/pages/admin/comments/comments-page.ts` around lines 90 - 92, The
getRepliedToLink function uses both XPath parent traversal (locator('..')) and a
CSS selector ('a, button'); replace these with semantic Playwright locators: in
getRepliedToLink(row: Locator) use row.getByRole('link', { name: /Replied to:/
}) or row.getByRole('button', { name: /Replied to:/ }) to target the actionable
element by its accessible name (or if the accessible name isn't available, add a
data-testid on the parent/action in the app and use
row.getByTestId('replied-to-link')); update the implementation in the
getRepliedToLink function (and any callers) to rely solely on getByRole or
getByTestId rather than locator('..') or locator('a, button').
E2E Tests FailedTo view the Playwright test report locally, run: REPORT_DIR=$(mktemp -d) && gh run download 21713927816 -n playwright-report -D "$REPORT_DIR" && npx playwright show-report "$REPORT_DIR" |
kevinansfield
force-pushed
the
issue/BER-3276-improve-child-display-in-thread-sidebar
branch
from
f40c0fb to
21354f6
Compare
ref https://linear.app/tryghost/issue/BER-3276 - Added useThreadComments hook with NQL filter to fetch only direct replies - Thread sidebar can now open for any comment (not just top-level) - "Replied to" links use URL navigation to open threads for referenced comments - Uses count.direct_replies for accurate reply counts in thread view
kevinansfield
force-pushed
the
issue/BER-3276-improve-child-display-in-thread-sidebar
branch
from
21354f6 to
abb9c30
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Fix all issues with AI agents
In `@e2e/helpers/pages/admin/comments/comments-page.ts`:
- Around line 91-93: The getThreadCommentById method currently uses a CSS
attribute selector which violates the E2E guideline to use semantic/test-id
locators; update the UI to add a combined data-testid for each thread row (e.g.,
data-testid="comment-thread-row-<id>") and change getThreadCommentById to use
the Playwright semantic locator (page.getByTestId with
`comment-thread-row-${commentId}`) instead of the CSS attribute selector so
tests rely only on data-testid locators.
🧹 Nitpick comments (2)
e2e/tests/admin/comments/thread-sidebar.test.ts (1)
30-228: Make Arrange/Act/Assert sections explicit for readability.Consider minimal section markers or helper blocks to clearly separate setup, action, and assertion in each test while keeping comments light.
As per coding guidelines, use the AAA Pattern (Arrange, Act, Assert) in E2E tests with clear sections.apps/posts/src/views/comments/components/comment-thread-list.tsx (1)
68-81: Consider guarding against an empty link target.If neither
comment.in_reply_to_idnorcomment.parent_idis set (data inconsistency),buildThreadLinkreturnsundefinedand the fallback|| ''results into="", which could cause unexpected navigation behavior.Consider conditionally rendering the Link only when a valid target exists:
♻️ Suggested guard
- {comment.in_reply_to_snippet && isSelectedComment && ( + {comment.in_reply_to_snippet && isSelectedComment && (comment.in_reply_to_id || comment.parent_id) && ( <div className={`mb-1 line-clamp-1 text-sm ${comment.status === 'hidden' && 'opacity-50'}`}> <span className="text-muted-foreground">Replied to:</span> <Link className="text-sm font-normal text-muted-foreground hover:text-foreground" - to={buildThreadLink(searchParams, comment.in_reply_to_id || comment.parent_id) || ''} + to={buildThreadLink(searchParams, comment.in_reply_to_id || comment.parent_id)!} onClick={(e: React.MouseEvent) => { e.stopPropagation(); }}
| getThreadCommentById(commentId: string): Locator { | ||
| return this.page.locator(`[data-testid="comment-thread-row"][data-comment-id="${commentId}"]`); | ||
| } |
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
# First, let's examine the full context of the comments-page.ts file
wc -l e2e/helpers/pages/admin/comments/comments-page.tsRepository: TryGhost/Ghost
Length of output: 111
🏁 Script executed:
# Read the relevant section with more context
sed -n '80,110p' e2e/helpers/pages/admin/comments/comments-page.tsRepository: TryGhost/Ghost
Length of output: 805
🏁 Script executed:
# Check class declaration and other locator methods for consistency
head -50 e2e/helpers/pages/admin/comments/comments-page.tsRepository: TryGhost/Ghost
Length of output: 2611
🏁 Script executed:
# Search for usage of getThreadCommentById in tests
rg "getThreadCommentById" e2eRepository: TryGhost/Ghost
Length of output: 571
🏁 Script executed:
# Look for the comment component or related UI files
find . -path ./node_modules -prune -o -type f \( -name "*comment*" -o -name "*thread*" \) \( -name "*.ts" -o -name "*.tsx" -o -name "*.hbs" \) -print | head -20Repository: TryGhost/Ghost
Length of output: 1307
🏁 Script executed:
# Check the comment thread sidebar component
cat -n apps/posts/src/views/comments/components/comment-thread-sidebar.tsx | head -150Repository: TryGhost/Ghost
Length of output: 5679
🏁 Script executed:
# Search for where comment-thread-row data-testid is used/rendered
rg "comment-thread-row" --type tsx --type ts -B 2 -A 2Repository: TryGhost/Ghost
Length of output: 85
🏁 Script executed:
# Check the test file to understand the usage pattern better
sed -n '1,50p' e2e/tests/admin/comments/thread-sidebar.test.tsRepository: TryGhost/Ghost
Length of output: 1948
🏁 Script executed:
# Find and check the comment-thread-list component
find . -name "comment-thread-list.*" -type f | head -5Repository: TryGhost/Ghost
Length of output: 124
🏁 Script executed:
# Check how comment rows are rendered in the thread list
cat -n ./apps/posts/src/views/comments/components/comment-thread-list.tsx | head -100Repository: TryGhost/Ghost
Length of output: 5651
Avoid CSS selectors in page objects; prefer semantic/test-id locators.
The current implementation uses a chained CSS attribute selector to target both data-testid and data-comment-id. Consider adding a combined data-testid (e.g., data-testid="comment-thread-row-{commentId}") in the UI and updating this method to use getByTestId() instead. As per coding guidelines, never use CSS or XPath selectors—only semantic locators or data-testid in E2E tests.
🔧 Suggested update (requires adding a matching data-testid in the UI)
- getThreadCommentById(commentId: string): Locator {
- return this.page.locator(`[data-testid="comment-thread-row"][data-comment-id="${commentId}"]`);
- }
+ getThreadCommentById(commentId: string): Locator {
+ return this.page.getByTestId(`comment-thread-row-${commentId}`);
+ }🤖 Prompt for AI Agents
In `@e2e/helpers/pages/admin/comments/comments-page.ts` around lines 91 - 93, The
getThreadCommentById method currently uses a CSS attribute selector which
violates the E2E guideline to use semantic/test-id locators; update the UI to
add a combined data-testid for each thread row (e.g.,
data-testid="comment-thread-row-<id>") and change getThreadCommentById to use
the Playwright semantic locator (page.getByTestId with
`comment-thread-row-${commentId}`) instead of the CSS attribute selector so
tests rely only on data-testid locators.
There was a problem hiding this comment.
I have a fix for this but I'll create a separate PR once this is merged so we don't have to wait for CI all over again just for a test improvement.
There was a problem hiding this comment.
@kevinansfield, sounds good! Thank you for acknowledging this and planning to address it in a follow-up PR. That's a practical approach to avoid the CI overhead. Looking forward to seeing the fix! 👍
🧠 Learnings used
Learnt from: rob-ghost
Repo: TryGhost/Ghost PR: 26219
File: ghost/core/test/e2e-api/admin/comments.test.js:630-676
Timestamp: 2026-02-04T15:58:05.895Z
Learning: In the Ghost comments API (ghost/core/test/e2e-api/admin/comments.test.js and related files), count.replies is a backward-compatible alias of count.total_replies (all descendants), not direct replies. The new count.direct_replies field provides direct reply counts. This aliasing is intentional to maintain API compatibility with existing consumers.
Learnt from: CR
Repo: TryGhost/Ghost PR: 0
File: e2e/AGENTS.md:0-0
Timestamp: 2026-02-02T08:26:12.442Z
Learning: Applies to e2e/**/*.{test,spec}.{ts,tsx} : Use test IDs (data-testid) only when semantic locators are unavailable, and suggest adding `data-testid` to Ghost codebase when needed
Learnt from: CR
Repo: TryGhost/Ghost PR: 0
File: e2e/AGENTS.md:0-0
Timestamp: 2026-02-02T08:26:12.442Z
Learning: Applies to e2e/**/*.{test,spec}.{ts,tsx} : Never use CSS or XPath selectors - only use semantic locators or data-testid in E2E tests
Learnt from: CR
Repo: TryGhost/Ghost PR: 0
File: e2e/AGENTS.md:0-0
Timestamp: 2026-02-02T08:26:12.442Z
Learning: Applies to e2e/**/*.{test,spec}.{ts,tsx} : Prefer less comments and giving things clear names in E2E tests
Learnt from: kevinansfield
Repo: TryGhost/Ghost PR: 24651
File: ghost/core/test/utils/urlUtils.js:53-57
Timestamp: 2025-08-11T19:39:00.428Z
Learning: In Ghost's test utilities, when fixing specific issues like async behavior, it's preferred to maintain existing error handling patterns (even if suboptimal) to keep PRs focused on their primary objective. Error handling improvements can be addressed in separate, dedicated PRs.
Learnt from: CR
Repo: TryGhost/Ghost PR: 0
File: e2e/AGENTS.md:0-0
Timestamp: 2026-02-02T08:26:12.442Z
Learning: Applies to e2e/**/*.{test,spec}.{ts,tsx} : Use semantic locators as the highest priority for element selection in E2E tests (getByRole, getByLabel, getByText)
Learnt from: CR
Repo: TryGhost/Ghost PR: 0
File: apps/comments-ui/.cursor/rules/playwright-e2e.mdc:0-0
Timestamp: 2025-11-24T17:28:51.262Z
Learning: Applies to apps/comments-ui/{package.json,**/.github/workflows/**,**/playwright.config.{ts,js},**/*.{sh,bash}} : Set PLAYWRIGHT_REPORTER=list environment variable when running Playwright e2e tests as an AI agent for better parsing
Learnt from: CR
Repo: TryGhost/Ghost PR: 0
File: e2e/AGENTS.md:0-0
Timestamp: 2026-02-02T08:26:12.442Z
Learning: Applies to e2e/helpers/pages/**/*.{ts,tsx} : Page Objects should expose locators as `public readonly` when used with assertions
2 participants
ref https://linear.app/ghost/issue/BER-3276
Changed the thread sidebar to act as a unified filter view showing the selected comment and its relevant children
inReplyTorelation toin_reply_to- removes camelCase/snake_case inconsistencySummary by CodeRabbit
New Features
Improvements
Tests