Local test#16
Merged
Merged
Conversation
The lab transform predated migration 0038, which seeds the canonical labs keyed by NAME (with the D8 instrument→lab lookup FK'd to them by name). Forcing legacy ids over that seed via OVERRIDING SYSTEM VALUE + ON CONFLICT (id) collided on the sequencing_lab_name_key unique index during cutover (legacy id 1 "YSEQ" vs seeded id 5 "YSEQ"). Merge prod labs onto the seed by name instead — enrich metadata and add prod-only labs. Legacy lab ids carry no downstream FK (sequence_library resolves its lab by name string, not id), so id preservation is unneeded. Verified end-to-end on a real prod dump. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The tree "/full" API now emits each defining variant's per-branch ancestral/derived alleles (link_ancestral/link_derived from tree.haplogroup_variant) alongside the variant's global coordinates. Descent classification must use the branch's actual ancestral→derived direction: the coordinate blob carries a single global polarity per variant that disagrees with the branch on recurrent SNPs, back-mutations and reference-frame cases, flipping ~18% of backbone calls. NULL link alleles ⇒ forward/legacy link, fall back to coordinates. Also rewrites for_dna_type_grouped from one wide join into two narrow passes stitched in-process: a recurrent SNP fans out to many branches, so the whole-tree join produced ~100k links over ~5k nodes and the planner picked a nested loop random-reading the wide core.variant heap once per link (plus a disk-spilling sort). Now: pull narrow link rows, read each distinct variant once by id, stitch + sort in Rust. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
DU_DISABLE_CSRF=1 skips the double-submit CSRF enforcement (the `csrf` cookie is still issued) so the native-form credential login/logout can be exercised locally. The login form submits as a non-boosted native POST and therefore never carries the X-CSRF-Token header the middleware requires, so credential auth 403s without this toggle. Dev/local-test only — never set DU_DISABLE_CSRF in production. The durable fix is to make login/logout carry the token (hidden field + form-field fallback, or HX-Redirect-boosted forms). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Variant Naming Authority fixes surfaced during the cutover smoke test: - Display the hs1 (T2T-CHM13) coordinate the de-novo catalog is built on, not GRCh38 (most de-novo variants had no GRCh38 coord → blank). Show the mutation state (ancestral→derived) the curator needs to name a variant. - Dedup by SITE (locus + ancestral + derived), not position alone: two distinct SNPs can share a position with different alleles (Z12236 A>C vs Y17125 A>G) and are not duplicates. - Reuse an established (non-DU) name via a new "adopt" action instead of minting — "named by definition" — guarded against the (name, branch) unique index so a taken name routes to merge review, not a 500. - Scope the "needs a name" queue to variants that DEFINE a branch: a canonical name is (name + branch), so branch-less catalog rows are reference data, not naming work. The (canonical_name, defining_haplogroup_id) model — same SNP name canonical on each branch it defines, replacing ISOGG's L270.1/.2 suffixes — was never populated (defining_haplogroup_id NULL everywhere), collapsing the unique index to global-name-uniqueness. scripts/backfill-defining-haplogroup.sql populates it from the tree links (row-per-(name,branch), recurrence split). Durable loader support (denovo::load setting it + clear_dna deleting the recurrence copies) is the tracked follow-up. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Naming queue is Y-DNA only: mtDNA variants are identified by their [anc]pos[der] coordinate string and never get a DU name, so exclude them from "needs a name" (they were sorting first and filling the queue). - Load bootstrap.bundle.min.js in <head> (deferred) instead of at end of <body>: hx-boost swaps the whole body on every nav, re-executing a body <script> each time and stacking duplicate Bootstrap event-delegation listeners (dropdowns/toggles double-fire; htmx content swaps misbehave until a full reload). In <head> it runs once and survives boosts. - hx-boost="false" on the nav dropdown-toggle anchors (href="#"): htmx was boosting their click and navigating to "#" instead of letting Bootstrap open the menu. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Prev/Next buttons put mode in their hx-get URL AND inherit hx-include="#nm-mode" from the parent #naming-table, so htmx appended mode a second time (?mode=X&page=N&mode=X). serde rejects the duplicate key → 400, so pagination silently failed. Drop mode from the button URLs and let the inherited hx-include supply it. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The de-novo loader stamps every branch-defining variant NAMED with a synthetic coordinate string as its canonical_name (e.g. "chrY:10158192 C->G") and also copies that string into common_names. The naming queue already surfaced these (canonical_name LIKE 'chr%:%'), but the detail panel keyed "already named" purely off naming_status = NAMED, so the curator saw them as named with no way to mint a DU identifier. Add du_db::naming::is_placeholder_name and thread it through: - assign_du_name / adopt_established_name accept a placeholder-named NAMED variant as still-nameable (real name blocks, placeholder doesn't) - assign_du_name purges chr placeholders from common_names on mint and never preserves a placeholder as an alias - established_name / common_names display filter skip placeholders so they aren't offered for adoption or shown as alias chips - list + detail render placeholder-named rows as UNNAMED (unnamed) Regression test covers the NAMED-placeholder mint path. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…e-set tools Same duplicate-query-param bug as the naming tool (24bab37): each of these fragment lists sits in a container with hx-include pointing at a filter control (search box or status select), and the Prev/Next buttons ALSO hard-coded that same param in their hx-get URL. htmx appends the included value on top of the URL's, producing a duplicate key (?query=&page=2&query=) which axum's Query extractor rejects with 400. Fix (matching the naming + haplogroups pattern): drop the hard-coded filter param from the pagination URL and keep only `page` — the inherited hx-include supplies the live filter value (also fixes stale-snapshot filters when paging). Tools fixed: regions (the reported one), variants, publications, proposals, change-sets. Audited all curator fragment pagers: haplogroups/naming already correct; instrument-proposals and denovo-conflicts have no hx-include so don't collide; inbox has no Prev/Next controls. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Two follow-ups to the pagination-400 sweep. 1. Defense-in-depth: crate::extract::Query, a drop-in replacement for axum::extract::Query that no longer 400s on a repeated query key. axum's Query (serde_urlencoded) and axum-extra's (serde_html_form) both reject a duplicated scalar key — exactly what HTMX produces when a pagination link hard-codes a param that the container also sends via hx-include. The new extractor splits the raw query into ordered pairs (a sequence tolerates repeats), collapses duplicates keeping the last value (form semantics), then deserializes. Swapped into all 11 curator fragment route modules. Unit tests cover last-wins, plain, and empty. 2. Inbox pagination: the team inbox rendered a "Page x / y" label with no controls, so curators couldn't get past page 1. Added Prev/Next buttons (its container has no hx-include, so status is carried in the URL — now doubly safe under the new extractor). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…tions The Sequencer-Lab tool was proposal-only — a consensus review queue that's empty until federation flows. The 36 preseeded (mig 0038) instrument→lab mappings were applied to sequencer_instrument.lab_id and served by the public lookup, but had no maintenance surface. Add a second tab that lists every instrument→lab association (search + pagination, unassigned first) and lets a curator reassign the lab, fix the model/manufacturer, and set the D2C flag directly — no proposal needed. Reuses the get-or-create-lab + audited-update path from accept_proposal, keyed by sequencer_instrument.id. du_db::sequencer: list_established, established_detail, list_labs, update_instrument_lab (audited REASSIGN_LAB). New routes under /curator/instrument-labs/* (distinct prefix — no :id route collision). Bootstrap nav-tabs on the page; each tab is its own two-panel pane. Pagination follows the ?page-only + hx-include pattern. i18n en/es/fr. DB test covers list/search/reassign/create-new-lab/audit. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…idation Reference samples used as de-novo tree building blocks showed "No haplogroup call" even though they're placed in the tree. Root cause: one individual is split across several core.biosample rows — multiple publication accessions (SAMN… NCBI + SAMEA… EBI) plus a de-novo tree tip keyed by the panel id (NA18530) — and the ETL gave each its own specimen_donor while leaving the tip donor-less. The tree placement (tree.haplogroup_sample) landed on the tip, so it couldn't surface on the catalog accessions. ~2,660 individuals affected. Fix uses the donor as the linking entity (the entity above biosamples): - du_db::donor::consolidate_denovo_donors — groups biosamples by shared panel id (tip.accession == ref.alias), points every member at the richest donor, fills its gaps + canonical identifier, prunes the emptied donors. New run-once job `consolidate-donors` (preview unless --apply). Applied to the cutover DB: 2,661 groups, 2,684 biosamples repointed, 23 donors pruned. - Sample report resolves the haplogroup at the DONOR level: a new TreePlacement call origin sits between FedConsensus and Original in the precedence, resolving tree.haplogroup_sample across the donor's biosamples. Shows on every accession with a "from de-novo tree placement" badge. - De-novo loader now attaches each tip to the shared donor at creation (reuse an existing catalog donor by alias/accession, else create one keyed by the panel id) so reloads don't recreate donor-less tips. - dedup merge plan: cover genomics.biosample_str_profile (mig 0053, was stale → assert_fk_coverage would reject any merge); KeepSurvivor now handles a table unique on the repoint column alone. Tests: donor_consolidation (split→consolidate→prune→report surfaces placement on both accessions); existing dedup_merge/merge_e2e still pass. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The landing page had a single Y-DNA tree button and no explanation of what Decoding Us is. Add: - an mtDNA tree CTA alongside the Y-DNA one (both now icon buttons), keeping the variant-search button; - a "What is Decoding Us?" section — three cards (open haplogroup trees, public variant catalog, private discovery & research); - a "How to participate" section — explore / contribute your DNA / own your data / collaborate, with a link to the About page. Renamed home.cta.tree → home.cta.ytree and added home.cta.mtree + the eco.*/participate.* strings across en/es/fr (25 home.* keys, at parity). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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
1 participant
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.