Skip to content

Implement updates tracking OWID grapher#6

Open
xrendan wants to merge 1197 commits into
BuildCanada:masterfrom
owid:master
Open

Implement updates tracking OWID grapher#6
xrendan wants to merge 1197 commits into
BuildCanada:masterfrom
owid:master

Conversation

@xrendan
Copy link
Copy Markdown
Member

@xrendan xrendan commented Feb 26, 2026

Context

Links to issues, Figma, Slack, and a technical introduction to the work.

Screenshots / Videos / Diagrams

Add if relevant, i.e. might not be necessary when there are no UI changes.

Testing guidance

Step-by-step instructions on how to test this change

  • Does the change work in the archive?
  • Does the staging experience have sign-off from product stakeholders?

Reminder to annotate the PR diff with design notes, alternatives you considered, and any other helpful context.

Checklist

(delete all that do not apply)

Before merging

  • Google Analytics events were adapted to fit the changes in this PR
  • Changes to CSS/HTML were checked on Desktop and Mobile Safari at all three breakpoints
  • Changes to HTML were checked for accessibility concerns

If DB migrations exists:

  • If columns have been added/deleted, all necessary views were recreated
  • The DB type definitions have been updated
  • The DB types in the ETL have been updated
  • If tables/views were added/removed, the Datasette export has been updated to take this into account
  • Update the documentation in db/docs

After merging

  • If a table was touched that is synced to R2, the sync script to update R2 has been run

sophiamersmann and others added 29 commits April 17, 2026 18:24
🐛 (search) don't fallback to default entities as peers
When searching in the chart editor's "Origin url" field, typing a
space now matches a dash in option values, so "child labor" surfaces
"/child-labor". Also normalizes the highlighted substring so the
bolding stays in sync.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The onChange handler was running .trim() on every keystroke and writing
the trimmed value back to the controlled input, which made trailing
spaces disappear the instant they were typed — so you couldn't get from
"child" to "child labor". Store the raw value and only check .trim() to
decide whether the field is effectively empty.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Spaces are allowed while typing as a search convenience (so "child
labor" finds "/child-labor" in the dropdown), but the saved value must
never contain spaces. On blur, replace any lingering spaces with
dashes. Also hide the "↵ Enter to use custom URL" hint while the query
contains spaces — a URL with spaces isn't a valid custom URL to commit.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Keep all topic pages cards on the right when articles are present.

Fixes #5681
The file is broken and shouldn't be used by anything anyway. It's needed for migrations managed by Knex, but we use TypeORM instead for those.
Simpler model: instead of accepting spaces, filtering through them, and
normalizing on blur, just replace each space with "-" in onChange. The
input never holds a space, the stored value never encodes as %20, and
the filter/highlight/custom-URL-hint logic all collapse back to the
original form.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Deduplicate GDoc slugs before triggering Buildkite lightning builds and
before validating targeted bakes. Add a regression test for duplicate
lightning deploy slugs.
✨ reverse order of entities in the No data section
- Show fallback image only when JS is disabled
- Otherwise, render Grapher, which has its own loading indicator

This should improve the UX especially on data pages, where the image
previously flashed for a brief moment before the Grapher started
loading.
Keep inactive key insight slides measurable instead of hiding them with
`display: none`, preventing cached interactive charts from briefly
rendering titles at zero width when switching slides.
They run them too often, prolonging the feedback loop.
🔨 clarify split between interactive/static captioned chart
🐛 Admin origin URL autocomplete: treat spaces as dashes
Nested DoDs should still work.

Fixes #5917
- Use react-query to fetch the multi-dim config
- Defer loading the config until the embed scrolls near the viewport,
  similar to `GrapherWithFallback`
- Show a fallback image when JS is disabled
- Log multi-dim analytics event only when the multi-dim becomes visible,
  similar to `Grapher`
- Introduce a single site-wide `QueryClient `and a
  `SiteQueryClientProvider`

Fixes #6353
MultiEmbedder is no longer used for multi-dims.
mlbrgl and others added 30 commits May 22, 2026 15:04
Match the convention used elsewhere in site search (SearchDataResults,
SearchWritingResults): on initial fetch render placeholder cards that
mirror the LatestHit layout instead of a centered "Loading…" string.
Reuses the existing .animate-pulse keyframes from search/skeletons.scss.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Collapses the (type, kicker) pair on PagesChronological records into
one flat latestType attribute with five values (article, data-insight,
data-update, website-upgrade, announcement). The raw OwidGdocType stays
on the record so the dynamic atom feed keeps filtering on type as before.

- Adds LatestType + LATEST_TYPE_VALUES in the shared types package.
- Derives latestType at index time; logs unrecognized announcement
  kickers via logErrorAndMaybeCaptureInSentry with a safe fallback to
  "announcement".
- Swaps kicker for latestType in attributesForFaceting.
- Simplifies queryLatestPages: filters is always the static base
  filter; user-selected type and topics each live in their own
  facetFilters group, dropped per-query to drive disabling counts.
- Drops LatestFilter tagged union, encodeFilter, the kicker: URL
  prefix, and the dual TYPE_META/KICKER_META lookup in favor of a
  single LATEST_TYPE_META map keyed by latestType.

URL changes from ?type=kicker:data-update to ?type=data-update.
Requires a PagesChronological reindex on deploy.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Header & mobile nav: rename "Insights" → "Latest", point at /latest
- Homepage hero: new "Articles" pill linking to /latest?type=article,
  backed by a server-side article count
- "See all Data Insights" homepage button: → /latest?type=data-insight
- Footer "Insights": → /latest?type=data-insight (kept alongside "Latest")
- Featured Data Insights, small-screen carousel, and DI permalink page
  "see all" + breadcrumb: → /latest?type=data-insight
- Homepage DI carousel cards: link to per-DI permalink instead of
  scroll-anchoring on the soon-to-be-deleted /data-insights index
- Centralize /latest URL construction in latestUrl() helper
  (renamed latestFilters.ts → latestUrl.ts)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The paginated /data-insights index is now redundant — every entry point
that used to land there now lands on /latest?type=data-insight (commit
237a063). Delete the page, its baking pipeline, sitemap entries, dev
preview branch, and orphaned constants/helpers. Add a permanent 301
redirect (bare + wildcard) so external links and search engines land in
the right place. Individual /data-insights/{slug} permalinks, the atom
feed, and the entire admin UI are unaffected. Existing slug-rename
redirects in the DB still take precedence over the new wildcard.

Also redirect /latest/page/N (retired SSR pagination, no SPA analog) and
collapse /blog/* onto /latest — the existing /blog/* /latest/:splat rule
was 404-ing for every input, since /latest has never had a /:slug route.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adopt the searchParamsToState / stateToSearchParams / urlNeedsSanitization
pattern from site/search/searchState.ts. Unknown params (e.g. legacy
?topic=Health from old /data-insights links), invalid topic names, and
unrecognized type values are now stripped from the URL on first paint
via a replaceState. Filter mutations build URLs from validated state
instead of copying URLSearchParams from the previous URL, so unknown
params can't sneak back in.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Introduce LatestFeedGdoc (subset of ChronologicalGdoc shown on /latest)
  derived structurally from LATEST_FEED_TYPE_VALUES, plus a
  checkIsLatestFeedGdoc predicate. Tighten ChronologicalGdoc so
  content.type is non-optional, eliminating one ! assertion in the indexer.
- Split deriveLatestType into a dispatcher that takes LatestFeedGdoc and
  a deriveAnnouncementLatestType(kicker) helper. The two announcement-only
  call sites (standalone preview, homepage card) call the helper directly,
  so the dispatcher is only used by the indexer (gated behind the new
  predicate). Dispatcher uses ts-pattern shape-match — no `as` casts.
- Subset enforcement: LATEST_FEED_TYPE_VALUES `satisfies` the chronological
  tuple at compile time. LATEST_BASE_FILTER (Algolia query) and
  LATEST_FEED_TYPES (Set) both derive from it, single source of truth.
- Narrow getHomepageAnnouncements return type to
  OwidGdocMinimalAnnouncementInterface (the SQL already filters
  type='announcement'); propagate through OwidGdocHomepageMetadata and
  HomepageIntro.
- Rename checkIsChronologicalFeedPost → checkIsChronologicalGdoc and
  checkIsLatestFeedPost → checkIsLatestFeedGdoc to match the type names
  and the codebase's checkIsX convention.
- Rename typeFacetFilter → latestTypeFacetFilter in queries.ts (it filters
  on the latestType attribute, not type — LATEST_BASE_FILTER is the type
  filter).
- Fix three misleading comments in pagesChronological.ts: the topic-page
  branches and isChronologicalGdocInstance gate were claimed to be
  unreachable / skipping, but topic pages are in fact chronologically
  indexed (for the atom feed) and reach those branches at runtime — the
  noop is the actual behavior.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds site/latest/README.md describing the architectural patterns of the
new /latest SPA: records carrying their own AttachmentsContext, the two
indexing paths (bulk + individual), URL as filter-state source of truth,
the batched facet-counting search, the latestType derived field, and the
preview-surface role of standalone announcement pages. Mirrors the shape
of site/search/README.md and notes the deliberate parallel between the
two surfaces.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mirror the /search analytics pattern on /latest. Two new GA4 events are
added to the registry: SiteLatest fires on filter state change, and
SiteLatestResultClick fires when a card link is clicked. A new
useLatestAnalytics hook debounces the filter event (skipping default
state and consecutive duplicates), and a LatestContext exposes the
SiteAnalytics instance to hit components.

Announcement cards intentionally don't track clicks — they have no
canonical link to their own slug, mirroring how links inside the rich
excerpt of an article hit are also untracked.

LatestState moves into @ourworldindata/types alongside SearchState.

GTM registration of owid.site_latest, owid.site_latest_result_click,
latestTopics, and latestType is required before GA4 will collect the
new params.
Drop the redundant "Site" prefix to match the dominant naming pattern
in SiteAnalytics (logSearch, logInstantSearchClick,
logSearchAutocompleteClick, and the new logLatestResultClick). The
EventCategory.SiteSearchResultClick GA4 event name is unchanged — the
"Site" prefix there is a deliberate namespace.
Point the SiteResources "Latest" link at the data-insight feed and
rename the SiteFooter "Insights" entry to "Data Insights" for consistency.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Add `parseLatestExcerpt` next to `parseSimpleText` in rawToEnriched. Per
block: `parseText` for raw text blocks; a synthetic EnrichedBlockText with
a parseError for any non-text block.

Surface those parseErrors in `GdocPost._validateSubclass`, mirroring the
existing `parsedFaqs.parseErrors` walk. Errors show up in the admin via
`gdoc.errors` → `getErrors()`, gating Publish on Error-severity entries.
Rename `getAndLoadPublishedDataInsightsPage` to
`getAndLoadLastPublishedDataInsights` and hardcode the limit. The atom
feed and `GdocDataInsight.getPublishedDataInsights` only ever needed the
most recent insights, never a page offset. Removes the now-unused
`LATEST_INDEX_PAGE_SIZE` constant and its export.
Drop a few comments that restate what the code already shows, and shorten
the `isChronologicalGdocInstance` doc to a single sentence. Add a brief
cross-reference from `checkIsChronologicalGdoc` to its class-narrowing
counterpart.
The /atom.xml ?type= param now accepts LatestType slugs (article,
data-insight, data-update, website-upgrade, announcement) and filters
Algolia on the indexed latestType facet — matching the /latest UI and
letting subscribers narrow to announcement subtypes.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Match the ~-separated set encoding used by /search and /latest so the
atom feed's ?topics= and ?type= params stay coherent across surfaces.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replace the hand-rolled content type dropdown with react-aria-components'
Select / ListBox / Popover. The trigger's active styling is now driven
by `data-pressed` (set while the popover is open) instead of by the
current selection.
Use react-aria ToggleButtonGroup for topic pills (keyboard nav, focus
ring) and Select for the type dropdown. Give the "All" pill and "All"
type option specific aria-labels ("All topics" / "All types") while
keeping the visible "All" text. Reset the negative right offset on the
scroll arrow under sm-only since the type dropdown isn't a sibling on
mobile.
LatestTopicFacets isn't memoized and the handlers aren't used in any
deps array, so useCallback was purely ceremony.
Bumps the newsletter email input to 16px at the sm breakpoint, matching the
existing sm-only convention used on the not-found page form.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…-with-algolia-spa-search

🎉🤖 Revamp Latest page with Algolia-powered SPA search
…est-page-with-algolia-spa-search"

This reverts commit f4fe2b0, reversing
changes made to 08d9c2e.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

10 participants