Optimize/landing - #635
Closed
DavidK4leido5 wants to merge 105 commits into
Closed
Optimize/landing#635DavidK4leido5 wants to merge 105 commits into
DavidK4leido5 wants to merge 105 commits into
Conversation
Release: dev to master
merged dev
dev to master
The Be a Member page filtered developers on role_id 10, but the accept flow sets role_id 4 (Intern). Accepted developers therefore never appeared there, and whoever still held the legacy role_id 10 showed up instead — the "iba ung devs rendered" report. application_status is the reliable signal: the accept flow always sets it to "passed", while role_id tracks the Intern → Codev → Mentor progression and changes as a developer advances. Filtering on it matches what /profiles already does, so both public pages now read from the same source of truth. The existing post-filter is unchanged and still removes admins, failed and applying codevs, then narrows to active — so no one who should be hidden becomes visible. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
fix(codevs): show accepted developers on Be a Member (cherry-pick to prod)
Enrollment generated 10 recovery codes with Math.random(), showed them in a modal telling users to store them safely, and never persisted them. The 2FA challenge screen's backup-code branch was a placeholder that rejected every code unconditionally, so no code could ever work. Users who enabled 2FA and lost their authenticator were left holding codes that did nothing. Removes the backup-code path from the challenge form and replaces the recovery-code modal with an honest confirmation that says recovery codes are not available yet and losing the device means asking an admin to reset 2FA. The TOTP flow is unchanged — the middleware assurance-level check and challengeAndVerify were correct and are left alone. Real recovery codes are specced separately in docs/tasks/Jury-FS/2FA-Recovery-Codes-Are-Non-Functional.md. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…overy-codes fix(2fa): remove non-functional backup recovery code UI (cherry-pick to prod)
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
…ffecting the whole page
…edMetrics component
…ted html and data during build time
AccountSettingsUsername: the mount fetch is a query and the debounced availability typeahead is a query keyed on the debounced value, replacing a timer-managed effect. Settings profile, hire and admin-controls: mount fetches become queries; JobListingsTable mutations write through the cache; TicketManagementView derives the selected ticket instead of syncing it in an effect. Also types ProfilePointsResponse.points as the row it actually is, so consumers stop narrowing it inline. dataFetchingEffects 14 -> 11
Certificate preview derived the certificate name/type in an effect and fetched users and points in two more; all three become queries keyed on the viewed user, with the editable name held as a draft. MeetingBasedAttendance mirrors the attendance grid: the month travels with the editable draft so a month change re-derives during render. MyTeamPage drops the client-instance-in-state effect; 2FA factors are a query the enrolment path invalidates. dataFetchingEffects 11 -> 7
AddMembersModal's roster load, ChecklistManageModal's template load and the test route's projects fetch become queries. The add-members reset-on-close effect is deleted: the parent mounts that modal only while open, so it was unreachable. The four remaining data-fetching effects are all Rule-2 synchronisation and are intentionally kept: the dashboard codev_points realtime channel, notification polling plus realtime, the test route's live clock, and the survey widget's one-shot imperative modal open.
The plan flags mutating actions with no revalidatePath as more dangerous now that staleTimes holds the client Router Cache longer, and asks for them to be fixed on any route touched. /home/hire's table and the my-team attendance/member-points queries both read data these actions write, so a delete or status change would otherwise leave the table showing the old rows. Five hire mutations and both attendance sync paths now revalidate.
staleTimes holds the client Router Cache for a long window, so an action that writes without invalidating leaves the page showing pre-mutation rows. Nine action files were missing it: news banners, survey questions, services, promote-modal, ticket-support, dashboard timers and schedule, attendance warnings, reapply, and signup. A script now reports any mutating action file with no invalidation; it reads 35/35 covered, and it recognises the kanban helper so that file is not a false positive.
Four of five useEditor calls omitted the flag, so Tiptap logged an SSR warning per render and dropped the subtree to a client render, which shows as a hydration mismatch. TaskEditModal already had it; the other four call sites now match. Verified by console capture on /home/overflow with the composer open: 18 warnings before, 0 after.
A project with no team_leader row is a normal state, but .single() reports it as PGRST116 'The result contains 0 rows', which the code then logged as 'Error fetching team lead' on every /home/my-team load. maybeSingle returns null, which the existing !data?.codev branch already handles. Verified by console capture: the error fires before, and is absent after.
Keeps the probes that established each fix rather than the throwaway edit scripts: the landing pagination flow, the Tiptap console capture, the project-member error capture, and the revalidation audit.
/home/announcements is a component directory, not a route, so probing it only ever hit the 404 page. The modal opens from the navbar button on any /home page, which is what this exercises: it loads its tabs from the database and reports no errors.
Filters on /home/in-house and /home/interns never reached the server. A filter change built a new query key, TanStack handed it initialData, and with no initialDataUpdatedAt it counted as fresh under staleTime: 60s, so queryFn never ran and the table rendered the unfiltered seed forever. - usePaginatedQuery takes initialDataKey and seeds only on an exact hashKey match - both pages render pageSize-matching pages, so the server seed and the client's first request describe the same rows - add hooks/ui/use-debounced-value and debounce in-house search before it enters the key; 7 chars now issue 1 request Measured /home/in-house: search Shaina 10->1 rows, 1 request; dropdown GRADUATED 10->9 rows, 1 request
The route awaited every project and sliced the page in the client, so the first paint shipped the whole table. - getProjectsPage filters the many-to-many category relation in the database and returns Page<Project>; the card's own columns only (main_image, not a nonexistent image_url), members hydrated by the existing two-query shape - ProjectCardContainer drops usePagination and the tabPages map for usePaginatedQuery keyed on category + page Measured: 12 cards rendered, tab Web Application -> 1 request, cards changed.
The page awaited every client and sliced page 1 in the browser. - getClientsPage pages and counts in the database, ordering active-first there instead of sorting after the fetch - ClientsCard drops usePagination for usePaginatedQuery keyed on page, with a skeleton gated on isPending rather than data length Measured: 7 clients (whole set) render, 0 page errors.
The route awaited every task and sliced 9 per page in the browser, and derived totalPages from the array in an effect. - getTasksPage pages and counts in the database, filtering by codev_id directly (the old .single() codev lookup failed for users without a codev row) - TasksContainer drops the slice and the effect for usePaginatedQuery keyed on codevId + page Measured: 9 cards, 0 page errors, dataFetchingEffects unchanged at 3.
The route fetched every non-passed codev with select('*') plus the full applicant
join, then filtered the array in a useEffect on every keystroke.
- getApplicantsPage pages and counts per pipeline status in the database, with
an explicit column list matching what newApplicantsSchema requires and the
table renders; getApplicantStatusCounts gives the tab badges real totals
- ApplicantLists pages per tab with usePaginatedQuery, debounced search, and no
effect; the header is now just the search field
- delete applicantFilters, applicantFiltersBadge and applicantSorters: they
re-filtered the whole table in the browser and nothing else imported them
Measured: tabs show Applying 2 / Onboarding 4 / Waitlist 7 / Denied 1044;
search DAVID -> 4 rows to 2 on 1 request for 5 chars.
news-banners and surveys both fetched every column of every row to render cards
that use ten fields each. Column lists now match their row interfaces.
The five select('*') calls in settings/profile feed one user's own record into
six edit forms; enumerating columns there couples the page to every field a form
touches, so they stay (plan 5.1).
Measured: surveys and news-banners both render their real rows, 0 page errors;
p9-role-check allMatch true.
MemberChecklist fetched every column of every checklist row in the project to render eleven fields. The column list now matches its ChecklistItem interface. Left alone: MemberRating and ChecklistManageModal both use .limit(1).single(), and AddMembersModal already debounces its typeahead at 300ms (plan 5.1, 4.4).
…w search
certificate-preview was a client page.tsx that read the viewer's skill and
attendance points from the browser after hydration.
- new read-only getCodevPointsSummary (cache()); deliberately not
computeProfilePoints, which scores profile completeness and writes on read
- page.tsx is now a server shell that resolves the viewer and seeds the client
island; admins switching users still fetch that user (F3)
- browser REST calls on first paint: 3 -> 1 (the remainder is the admin picker)
overflow search filtered only the five loaded rows and issued no request.
- fetchQuestions takes a search term and filters server-side, so a term matches
every post; explicit column list replaces select('*')
- OverflowView pages and searches through a keyed useQuery seeded by the server
render, with optimistic like/edit/delete writes into the query cache
- skeleton no longer gates on isFetching, which flashed over live rows
Measured /home/overflow: search 'zzzz' 15 -> 4 cards on 1 request (was 0).
columns.ts imported revalidateKanbanBoardLists but never called it, so creating, renaming, moving or deleting a column left every cached board stale. updateTask was the one mutating function in tasks.ts with no invalidation, and saveDraft and deleteDraft had none either. The p9 audit reported 35/35 covered because its pattern matched the import line. Coverage is now per mutation: columns 4/4, drafts 5/5, tasks 11/11. No kanban refactor: a missing revalidate is cache correctness (plan 3, 5.1).
PersonalInfo fetched the positions reference table on mount even though the
profile page already renders on the server. Positions now arrive as a prop from
the cached getPositions in lib/server/reference-data.ts, and the duplicate
select('*') copy in actions/settings/profile.ts is gone.
getEducation and getJobStatuses take explicit column lists matching what their
callers read.
Left alone: codev-queries.ts select('*') is a default parameter callers override,
and the two in feeds/post.ts are head:true count queries that return no rows.
TaskCard read a status variable that was never destructured, so the server render threw ReferenceError and React fell back to client rendering for the whole route. The block was dead: Task has no status field. Found by looking at page text rather than the row count; the card grid still rendered nine rows through the client fallback.
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.