Spaces:

---

HuggingFaceBio

/

carbon-a-database-explorer

Like 8

Running

The Sankey reads as untranslated Latin above the familiar ranks: Eumetazoa and Porifera say nothing to most readers. Each group now carries a short English name under its scientific name, and the hover state no longer draws a pale slab across the ribbons or stacks two tooltips on top of each other.

English names

data/common_names.json, built by the new refresh_common_names.py, merges two sources:

• NCBI Taxonomy genbank common name and common name entries, about 9,200 taxa. This covers species, genera and familiar classes well (Porifera gets "sponges" straight from NCBI) but only 35 of the 200 largest groups in the atlas.
• A curated table of ~190 short glosses for the large unranked clades NCBI leaves unnamed, which are exactly the ones the atlas shows first: Opisthokonta "animals and fungi", Eumetazoa "animals with true tissues", Ecdysozoa "moulting animals", Endopterygota "insects with metamorphosis". Glosses are plain-language summaries of the living members of a clade, not formal synonyms.

The build verifies every curated taxid against the snapshot, so a renumbered clade fails the build instead of quietly mislabelling a branch, and it drops any gloss that only repeats the scientific name, so Fungi stays a single line.

Scientific names are now the primary label everywhere, replacing the ad-hoc DISPLAY_NAMES overrides. The gloss is the secondary line and also appears in the tooltip and the summary kicker. Rows grew from 36 to 46 points to seat the third line.

Hover fixes

• The hover highlight filled .tree-hit, a 210-point rounded rect that reached across the ribbons behind each label and read as a floating box. The hit area stays clickable but is never painted; hover now reads on the label and the bar outline.
• Each node group carried both a <title> and a data-detail, so the browser drew its own tooltip on top of the styled one. The <title> is gone; aria-label still names each node for assistive technology.

Icons

Not implemented, but the source is recorded in FEATURE_BACKLOG.md: PhyloPic resolves NCBI taxids directly and walks up the tree when a clade has no silhouette of its own, via GET https://api.phylopic.org/resolve/ncbi.nlm.nih.gov/taxid/{taxid}?embed_primaryImage=true. Two things need settling first: licences differ per image (CC0, CC BY 4.0, CC BY-SA 3.0, and at least one CC BY-NC-SA 3.0 among the groups shown first), so each icon needs its own credit and the non-commercial ones need replacing; and the SVGs should be vendored rather than hot-linked, to keep the Space free of external requests.

Tests

tests/test_taxonomy.py gains coverage for the glosses, the single tooltip, and the curated taxids still matching their scientific names in the snapshot. A length check already caught one gloss that would have been clipped in the chart. All 8 tests pass.

🤖 Generated with Claude Code