Final Ofsted/education-expert review of the detail page (Batch F) found one
clarity gap: on an all-through school the Admissions block shows a single
entry point's figures (Hessle's are the Year 7 secondary intake —
admissions.school_phase = "Secondary") but is titled just "Admissions", so a
parent interested in the reception/primary intake could read the Year 7
numbers as the whole school.
Add a phase subtitle for all-through schools ("These figures are for secondary
entry (Year 7)." / "...primary entry (Reception)."). Pure-primary and
pure-secondary pages are unambiguous and unchanged.
Rest of the review verified correct against source payloads: all three Ofsted
regimes (report card / ungraded post-Sept-2024 / carried-forward), England-
average anchoring, gap-honest trend charts (KS2 + KS4), all-through dual-phase
rendering, and the mobile layout.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Batch E of applying the compare-screen learnings to the school detail page —
all-through handling (point 11).
An all-through school carries both KS2 and KS4 figures in the same yearly
rows, but SchoolDetailView flipped it to isSecondary and rendered GCSE-only,
hiding the entire primary phase (SATs, phonics, KS2 trend). The Results
snapshot already gated its KS2/KS4 blocks purely on data availability, so both
already appeared there — but the section title, the trend chart, phonics, the
nav label and the history table all still assumed a single phase.
- Add an explicit `isAllThrough` flag (+ `showPrimaryContent = isPrimary ||
isAllThrough`); pure-secondary behaviour is unchanged.
- Hero: an "All-through (primary & secondary)" meta chip for all-ages framing.
- Results section: title "SATs & GCSE Results", a combined subtitle, and
"Primary — KS2 SATs (Year 6)" / "Secondary — GCSEs (Year 11)" sub-headings
separating the two blocks.
- Results Over Time: render two stacked PerformanceCharts (KS2 SATs + GCSEs)
rather than crowding both stages' series — on different scales with
different gap stories — onto one axis. Each gets its correct England overlay.
- Phonics section + nav item now show for all-through (primary-stage metric).
- History table: an all-through column set covering both phases (RWM, Exceeding,
Attainment 8, Progress 8, Eng & Maths 4+).
- Nav "Results" label for all-through instead of "GCSEs".
e2e: a new journey asserts an all-through school (Hessle, 137306) shows both
the KS2 and KS4 results and the All-through label, with a data-driven
precondition skip if staging data drifts.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Applies the compare screen's chart-truthfulness rules (spec §8.1) to the
school detail page's time-series charts.
PerformanceChart (Results Over Time, both phases):
- Fill every academic year between the first and last data point via the
shared fillAcademicYears helper, so cancelled/unpublished years (2019/20,
2020/21, and — for KS2 — 2021/22) render as real gaps instead of
compressed time. Each series and the England overlay map onto this filled
axis with null for missing years; spanGaps:false so school lines never
bridge a gap.
- Replace the primary-only COVID note with a distinct, honest gap caption:
KS2 names the cancelled tests plus the unpublished 2021/22 school-level
year; KS4 names the unpublished 2019/20–2020/21 GCSE grading years.
- Mobile x-axis switches to autoSkip so the longer (gap-honest) axis stays
readable; the broken line still marks a missing year even when its tick
label is skipped.
AdmissionsTrendChart:
- Same gap-honest axis + spanGaps:false so a missing admissions year is a
real gap, not compressed time.
Point 12 (definite canvas heights): desktop is already a definite 280px;
fix the secondary detail's mobile .chartContainer, which fixed the outer
box at 220px and double-constrained PerformanceChart's own 220px canvas +
chip strip (clipping the chips onto the plot) — now height:auto to match
the primary view.
SatsChart is out of scope (single-year per-subject CSS bars — no year axis,
no canvas height to constrain).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Batch C (points 7-10):
- Point 7: the admissions section shows the first-choice offer rate but no
sense of how severe a low rate is. Reuse the compare screen's banded
summariseAdmissions ('More than half of first choices missed out' etc.)
as an interpretive line under the tiles. (Selective/faith framing was
already present via admissionsTag.)
- Point 8: replace the jargon P8-suspended banner ('not used for
accountability purposes ... treat with caution') with the compare
screen's plain-English 'no KS2 baseline (COVID)' explanation.
- Points 9 (disadvantaged cohort sizes) and 10 (FSM benchmark) don't apply
to the detail page: it shows disadvantaged composition, not disadvantaged
attainment, and no FSM-vs-benchmark comparison — so there's no small-cohort
caveat or cross-definition fallback to fix here.
- Also fixed a stray lowercase 'national avg' provenance string Batch B
missed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Batch B (points 5-6): the detail page called every national figure a
'National avg'. All of them come from the official-DfE national-averages
marts (KS2/KS4 headlines), so they are England averages — relabel to
match the compare screen's provenance convention ('England average' for
official figures; the detail page has no computed benchmarks, so no
'state-school average (computed)' label is needed). Point 6 (anchoring
every number) is already satisfied on the detail page via DeltaChip.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Batch A of applying compare-screen learnings to the detail page (points 2-3):
- Point 2: the legacy OEIF sub-judgement grid rendered any non-null value
via OFSTED_LABELS[value], so Ofsted's sentinel code 9 ('not applicable')
leaked as an undefined label, and the sixth-form judgement was dropped
entirely. New shared ofstedLegacyAreas() helper returns only real grades
(1-4) and includes Sixth Form provision. Both detail views + the
all-same-grade check use it.
- Point 3: the OEIF disclaimer always claimed 'Ofsted no longer makes an
overall judgement' even when showing a carried-forward grade. It's now
grade_source-aware and labels carried-forward grades as such, matching
the compare screen.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Review fixes on the list/map report-card PR:
- e2e precondition now hard-asserts ofsted_rc_date instead of test.skip, so
the backend dropping the field fails loudly (that's the regression under
test), not silently skips.
- Use page_size=5 (the real backend param); per_page was ignored and fell
back to the default page size.
- Update buildOfstedListBadge docstring to describe the ofsted_rc_date-based,
report-card-wins-first detection instead of the removed framework check.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The search-result cards and map pins keyed report-card detection off
ofsted_framework === 'ReportCard', but the API sets ofsted_framework to the
raw event grouping ('Schools - S5'); worse, ofsted_grade (the carried-forward
legacy grade) was checked first and won. So report-card schools were badged
by their old grade — Barclay's pin/card read 'Outstanding · 2021' instead of
'Report Card · 2026'. Same root cause as the detail-page fix, different
surface.
Expose ofsted_rc_date on the list serialization (the report-card inspection
date, non-null only for report cards) and make both badge builders treat a
present rc-date as winning over any grade, using its year. Removes the dead
framework === 'ReportCard' branches.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The detail views detected report cards via ofsted.framework === 'ReportCard',
but the API sets framework to the raw event grouping ('Schools - S5') even
for report-card schools — so the check never matched: report-card schools
rendered as legacy 'Ofsted Rating' badges dated to a pre-Nov-2025
inspection (e.g. Barclay shown as 'Outstanding, Inspected 7 October 2021'
instead of its Feb 2026 report card).
Detect report cards by the presence of the report_card object (matching
the compare screen), and date them with rc_inspection_date, never the
legacy inspection_date. Applies to both primary and secondary detail views.
New e2e assertion against a real report-card school gates it.
Note: the list/map surface has the same latent issue via ofsted_framework
(LeafletMapInner) — flagged as a separate follow-up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Promotion's 'Verify the staging E2E gate' step called the Gitea
commit-status API with REGISTRY_TOKEN, which has container-registry scope
but no repository scope — so the API returned 403 and promotion failed.
Switch to the built-in GITHUB_TOKEN (repo read scope), matching how
pr-checks.yml already authenticates to the Gitea API. REGISTRY_TOKEN is
still used for the docker registry login, which is its correct scope.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The link repeated the full school name ('Watford Grammar School for Boys's
Ofsted page') — long on mobile, and the colour-coded school name already
leads the row. Show just 'Ofsted page →'; keep the full name in aria-label
so screen readers can still tell the per-school links apart.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The colour-coded school name that leads each row in a mobile measure card
was 0.8rem — a touch small to scan which school a value belongs to. Bump
to 0.9rem (with a matching 10px dot) so it's easier to read. Mobile only;
desktop uses the column header and hides this label.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
SF1 (expert sign-off): the secondary academics 'Grade 5+ in English &
maths' and 'EBacc entry' rows rendered as bare numbers — the one place
the 'every number anchored against the England average' promise broke.
Both now show the official England anchor (already in the payload) + an
Above/Close/Below verdict chip, matching Attainment 8 (which also gains a
chip for consistency).
Button affordance: globals.css has a duplicate .btn block whose
'border: none' overrode the base '1px solid transparent', so
.btn-secondary/.btn-active's 'border-color' had no width — every outline
button (the modal's '+ Compare', search-result '+ Compare', 'Comparing')
rendered as borderless teal text. Give the outline variants the full
'border' shorthand so the outline renders regardless of the clobbered base.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The bottom-sheet overlay was position:fixed inset:0, sized to the LAYOUT
viewport, so when the mobile keyboard opened it kept full height: the
sheet (anchored to the overlay's bottom) and the lower half of the dim
backdrop sat behind the keyboard, hiding the search input and results.
Track window.visualViewport (height + offsetTop) while the modal is open
and pin the overlay to the visible region, so backdrop and sheet stay
above the keyboard. Size the sheet against the overlay (max-height:100%,
min-height:min(55vh,100%)) instead of vh, and drop the results list's own
inner scroll on mobile so the sheet is a single scroll area.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The 'add school' modal passed no title to Modal, so its header held only
the close button — shoved to the far left by justify-content:space-between
in an otherwise-empty bar — while the real title was rendered separately
inside the content. Pass the title to Modal so the header reads title
(left) + close (right), and drop the duplicate in-content heading.
Also: the mobile full-width result button targeted a dead .addButton
selector (the button never had that class), so it rendered inconsistently
— give the button a real module class and full width on mobile, tighten
the stacked cards, and give the bottom sheet a stable min-height so its
empty state isn't a tiny stub.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The chart note (and the 2-row mobile chip legend) lived inside the fixed
360px/420px chartBox and competed with the canvas for it, so a longer
caption — e.g. the new KS4 gap note — shrank the mobile canvas to ~178px
and tripped the e2e >220px guard. The canvas now owns a definite height
(280px mobile / 380px desktop) with chips above and the note flowing
below at natural size, so neither can eat the plot. Verified on staging:
patched canvas renders 280px vs the squashed 178px.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Review findings: (1) the 'no KS2 baseline (COVID)' explanation was derived
from the basket-wide max year, so a school with an unrelated data gap
borrowed it from a neighbour with 2024/25 data — now judged per school on
its own latest year. (2) The '~N disadvantaged pupils' cohort multiplied
eligible_pupils and disadvantaged_pct resolved independently across years
— now both come from the same yearly row that supplies the displayed
percentage.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
S1: first-choice chip banded (More than half / About 1 in 3 / Over 1 in 4
missed out) so a 44%-offered grammar isn't understated by half.
S2: Progress 8 explains its absence for 2024/25+ cohorts (no KS2 baseline,
COVID) instead of a bare 'No data'.
S3: KS4 trend charts get their own honest gap caption (2019/20-2020/21
unpublished; later years not in our dataset yet); y-axis 'Value'→'Score';
buildCompareChart exposes englandOnlyYears.
S4: all-through schools labelled in chips, rail caption says 'N schools ·
<phase> view' for mixed baskets, whole-school roll no longer judged
against the single-phase median, community section carries an all-ages
caveat.
S6 (spec §8.5): disadvantaged attainment shows the cohort behind it
('of ~50 disadvantaged pupils').
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The DAG's dbt --select list predates the official-KS4-nationals stream,
so the extract loaded raw.ees_ks4_national but the staging model and
fact_ks4_national_averages were never rebuilt — staging kept serving the
old computed means after the DAG run.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
M1: admissions rounds are now selected by the active phase tab
(admissionsForPhase) — an all-through school's Year 7 round no longer
masquerades as Reception odds beside pure primaries; honest per-cell and
section fallbacks name the round (Reception / Year 7).
M2: Ofsted sentinel codes (9 = not applicable) never render as judgement
chips, and the sixth-form judgement — previously dropped — now renders
for schools that have one.
M3: 'What this means' is phase- and type-aware: selective schools get
entrance-test framing, secondary faith schools a faith-criteria note, and
the primaries' distance template never appears on the secondary tab
(admissions_policy now exposed in compare school_info).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Review finding: merging left a re-added school's old entry in the map
forever, so the covered check served stale data and the map grew
unboundedly. The cancelled flag already discards superseded responses,
which is the actual race fix; replacing keeps the map bounded to the
current selection and guarantees fresh refetches.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Opening a compare link while localStorage held a different basket raced a
fetch for the OLD school set against the URL's SSR data; the stale
response replaced comparisonData, so no active school had data and every
section (including the trends chart) vanished until a hard refresh.
Responses from superseded effect runs are now dropped, successful ones
merge instead of replace, and the URL-seed effect re-runs when a
client-side navigation changes ?urns=.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
'COMPARING / 3 primary schools' — 0.72rem uppercase eyebrow over a
0.95rem semibold count, sized to sit alongside the 0.92rem chip names
without dominating. Desktop only; mobile pills unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The bar was a plain flex row while sections use a 200px-label + per-school
grid, so chips drifted off the columns they label. The bar now shares the
sections' grid template (via --school-count) at the same 761px breakpoint;
verified pixel-exact against live section columns (236/615/994). Mobile
pills unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Swept in accidentally by a broad 'git add pipeline'. They embed local
absolute paths and a personal usage-tracking UUID, and a stale committed
manifest causes partial-parse/version-mismatch noise for others.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The final review claimed DfE published school-level 2021/22 KS2 in Dec
2022. Re-verified: the GOV.UK announcement 'Primary school performance
tables: 2022' is CANCELLED ('will not be published in key stage 2
performance tables in academic year 2021/22'), and the EES 2021/22
release carries the same statement. No code change.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
New ees_ks4_national stream ingests the EES 'National characteristics
summary data' series (England, state-funded, all pupils). The old mart's
unweighted school means were 7-15 points off every headline measure and
produced an impossible national Progress 8 (-0.27). The API's computed
fallback is gone too: the footnote calls these figures official, so an
unbuilt mart now yields an empty series, never a stand-in.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The FSM chip anchored against disadvantaged_pct (a different measure,
FSM6+CLA) whenever fsm_pct was null — which it always was, since the
performance df has no fsm_pct. New fact_census_benchmarks mart supplies
pupil-weighted FSM/EAL means per phase; the KS2-column medians that
produced a bogus 50% 'secondary disadvantaged' anchor are gone.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The MI file's report-card grade columns belong to the latest FULL
inspection (col 'Inspection start date'), but inspection_date maps to the
legacy OEIF graded/ungraded dates — so report cards were being dated with
pre-Nov-2025 inspections. Also discover_csv_url() returned matches[0],
the oldest (2017) link on the GOV.UK page.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The Navigation header is position:sticky top:0 (z-index 1000). The school
bar was also sticky top:0 (z-index 10), so when scrolled it pinned at the
same top:0 *behind* the header — on mobile 57 of its 72px were covered,
leaving only a sliver, so you couldn't see which schools were selected.
Offset the bar's sticky top to the header height (65px desktop, 57px
mobile — the Navigation breakpoint is also 640px) so it pins just below.
e2e guard asserts the bar's sticky offset is at least the header height
(verified it fails against the pre-fix build).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The grid sections (At a glance, Ofsted, Getting a place, Who goes there)
collapsed generically on mobile — grey label pills, full names wrapping
to 3 lines, no dots — making the page ~2x the mockup's height and
'significantly different' from the mobile design.
Each measure is now wrapped in a <Measure> that is display:contents on
desktop (so the label + cells still flow into the shared aligned grid,
unchanged) and a white card on mobile with compact [dot][short name]
[value] rows — matching the mobile mockup. The sticky school bar becomes
scrollable short-name pills on mobile too. Adds a shortName() util.
Desktop layout is unchanged (display:contents dissolves the wrapper).
Validated the card mechanism and real content shapes (report-card cell,
badges, %+chip rows) via static previews at both widths.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The mockup's Explore trends section is the measure picker + chart only —
no data table. Removes the table (and the now-unused progressBand / band
chip / formatMetricValue plumbing that only fed it).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
ComparisonChart runs Chart.js with maintainAspectRatio:false, so it sizes
to its container's height — which must be definite. TrendsExplorer gave
.chartBox a min-height, which doesn't resolve the chart wrapper's
height:100%, so Chart.js fell back to its ~150px default: a squashed
8.6:1 sliver that didn't match the mockups. Set a definite height (420px
desktop, 360px mobile where the chips row sits above the canvas).
Verified on staging by patching the live height: canvas went from
1287x150 to 1287x392 (desktop) / 284 (mobile) — proper ~3:1 proportions
matching the mockup, with the England dashed line, COVID/2021-22 gap and
table all reading correctly.
An e2e guard asserts the trends canvas is taller than 220px so the
squash can't regress.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The compare page's phase tabs put all-through schools (which carry KS4
data) on the secondary tab, so comparing an all-through school with a
pure primary splits them across tabs and only the active tab renders its
link. The test picked the first two 'primary' search hits without
guaranteeing same phase, so it flaked whenever a search returned an
all-through school first (e.g. URN 137306). Now selects two pure-Primary
URNs via the API — deterministic and data-invariant.
Verified against staging: was a 15.6s timeout, now passes in ~1.8s.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The job does ~3s of real work (typecheck 1.4s + jest 1.3s) but installs
452 MB / 460 packages every run. Two changes:
- Cache nextjs-app/node_modules keyed on the lockfile hash (OS + node
major pinned) and skip npm ci entirely on a hit — deps change rarely,
so most PR pushes now do zero install. On miss, npm ci runs with
--prefer-offline --no-audit --no-fund.
- Workflow-level concurrency with cancel-in-progress: a new commit (or an
empty re-trigger) aborts the previous run instead of stacking a second
full matrix.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
get_supplementary_data ran ~5 sequential DB round-trips per URN, so
/api/compare scaled at ~37ms/school (measured on staging: 1 school 155ms,
3 schools 220ms, 6 schools 340ms). get_supplementary_data_batch fetches
each table once with WHERE urn IN (...) and groups in Python, collapsing
5*N round-trips to a constant 5. get_supplementary_data is now a thin
wrapper so the detail endpoint is unchanged; the compare endpoint makes
one batched call. Each table degrades independently on failure.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Refresh bug: on mount the basket is empty for a beat before it hydrates
from the URL. The fetch effect nulled comparisonData on that transient
empty urnKey, then the one-shot 'SSR covers it' skip suppressed the
refetch — leaving the page blank on reload. The effect is now gated on
isInitialized, never blanks on empty (the render already shows the empty
state when nothing is selected), and decides fetch-vs-skip by whether it
already holds each requested school's data (SSR or a prior fetch).
Perf: useComparison ran a useSWR('/api/compare') whose result nothing
consumed — dead weight that fired on every page (Navigation + Toast are
global) whenever the basket was non-empty, and duplicated ComparisonView's
own fetch on the compare page. Removed; the hook now exposes basket state
only.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
- Metric changes no longer refire /api/compare (the data is already
client-side; the picker is presentational) — the fetch effect depends
only on the URN set, with URL sync split into its own effect.
- The initial client fetch is skipped when the SSR payload already covers
the selected schools; national averages + benchmarks now arrive via SSR
props so nothing is lost by skipping.
- page.tsx fetches comparison and metrics in parallel.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
fact_ks4_national_averages is computed once at dbt build time (covered by
the EES DAG's stg_ees_ks4+ selector). _national_averages_payload now reads
both national-averages marts instead of scanning the performance dataframe
per year on every /api/compare request (~250ms saved per call). Fallback
for the deploy-before-DAG window computes the latest year only.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Addresses the AI review findings on PR #33:
- severe: the workflow_dispatch sha input was interpolated directly into
the run script (shell injection with REGISTRY_TOKEN + prod webhook in
scope). It now reaches the shell only via env, is rejected if it
starts with '-', and is resolved locally with git rev-parse.
- minor: the resolved sha must be a 40-hex ancestor of origin/main —
non-main refs are refused explicitly instead of implicitly.
- minor: a prod-promotion concurrency group serialises promotions
(cancel-in-progress: false).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
The AI review gate caught that stg_ees_ks2's 7 new columns broke the
positional UNION ALL with stg_legacy_ks2 in int_ks2_with_lineage, and
that the lineage CTEs never emitted them (same class of bug fixed for
KS4 in 34a5de2). Legacy gets typed null placeholders at matching
positions; both lineage CTEs pass the columns through. 45/45 columns
verified name-identical in order across both union branches.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Evidence trail for the rc_* mapping in the prior commit: real value_counts()
over the 7 MI report-card columns, confirming the 5-value grade vocabulary
and that 'Achievement'/'Safeguarding standards' match by exact string only
(no legacy OEIF column accidentally consumed).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Wires the tap TODO in stg_ofsted_inspections.sql: maps the 7 confirmed
report-card MI columns (Safeguarding standards, Inclusion, Curriculum
and teaching, Achievement, Attendance and behaviour, Personal
development and wellbeing, Leadership and governance) into rc_*
fields, parsed via the new parse_report_card_grade macro against
real sampled grade values (Exceptional/Strong standard/Expected
standard/Needs attention/Urgent improvement). rc_safeguarding_met
becomes boolean from Met/Not met. rc_early_years/rc_sixth_form have
no MI column yet and are intentionally omitted from COLUMN_PRIORITY,
staying NULL.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
DfE never published school-level KS2 2021/22 data publicly (confirmed via
EES release notes and by walking the Compare School Performance download
wizard, which has no ks2 checkbox for 2021-2022, same as the COVID-cancelled
2020-2021 year). No archive exists to verify column headers against or
upload to the filebrowser; Task 6 is blocked at the source-data level.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
Widen the year filter in stg_ees_ks2_national.sql from >= 201617 to
>= 201516 so the England national-averages line no longer starts a
year late; the catalogue CSV has a real, comparable 201516 row (2015/16
was the first year of the current expected-standard tests, so it's the
correct floor).
GPS/science/scaled-score national columns confirmed present at source
with correct mapping; prod NULLs are stale raw data, backfilled by the
next extract run. No _KS2_NATIONAL_COL_MAP change accompanies this fix.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
int_ks4_with_lineage.sql unions stg_ees_ks4 and stg_legacy_ks4 via
`select *`, which PostgreSQL aligns positionally. stg_legacy_ks4 listed
sen_support_pct before sen_ehcp_pct while stg_ees_ks4 lists sen_ehcp_pct
before sen_support_pct, swapping the two values for legacy-sourced rows
in marts.fact_ks4_performance. Reordered stg_legacy_ks4's final select
to match.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0146VHeLAWjDVE2B5uU67jCB
ReligiousCharacter 99 (~4k schools) and AdmissionsPolicy 9 (~5.6k) carry a
code with a blank name in the GIAS CSV; the generator skipped them so they
hit the Unknown(<code>) path — wrongly triggering the Faith-priority tag
and polluting filters. Blank-only codes now map to "" (byte-identical to
the old name pipeline); accepted_values lists extended to match the seed.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Addresses AI-review findings: the annual IDACI DAG also rebuilds a mart
(fact_deprivation) and needs the reload; curl gets connect/max timeouts
so an unreachable backend fails fast instead of hanging the task.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The daily/monthly/annual DAG docstring promised an Invalidate Cache step
that never existed — after a marts rebuild the backend kept serving its
startup-cached (possibly empty) DataFrame until a container restart.
Add a POST /api/admin/reload task at the end of each pipeline DAG,
mirroring the sitemap DAG's admin-call pattern.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
str(ProgrammingError) embeds the full SQL, which contains every column
name — the substring check matched any error and could take the wrong
retry branch. Parse the missing column from exc.orig instead.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Closes the deploy window flagged by CI review — the backend now works
against both the old (name) and new (code) mart schemas.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
