diff options
| author | Pjotr <wrk@bosa.lan> | 2026-11-07 17:10:28 +0000 |
|---|---|---|
| committer | Pjotr <wrk@bosa.lan> | 2026-11-07 17:10:28 +0000 |
| commit | 18d18b1c125c06dd67dcf54ccbd89eb128643737 (patch) | |
| tree | efefbdc30000ff459a46457239bc9fb01035d642 /docs/www | |
| parent | 1f61f8f86443de6353295f7363abbe1c986d863f (diff) | |
| download | genenetwork4-18d18b1c125c06dd67dcf54ccbd89eb128643737.tar.gz | |
genotypes: expose the metadata, never the genotype calls
The listing named versions but not files; a caller could pick one and
still not see what was in it. Three token-gated routes now describe a
version completely, short of the matrix:
GET /groups/{group}/genotypes versions, files, sizes, current
GET /groups/{group}/genotypes/{version} file name, header labels,
individuals, strains, chromosomes
GET /groups/{group}/genotypes/{version}/markers markers (name, chr, cM, Mb), paged
The detail reads the file's own `@`-labels (type, mat, pat, het, unk,
scale) and its panel in column order, and reports both the whole panel
and the strains a mapping uses after the parents/F1s rule -- so a caller
can check a request before sending it and interpret a result after.
Markers come back as positional rows with `meta.columns`, paged with
`?page`/`?limit` and filterable with `?chr`, following /compute/gemma's
shape rather than inventing one.
The genotype calls stay on the server: file_meta() reads the matrix to
skip it, and nothing returns a cell.
* src/genotypes.rs (Marker, FileMeta, file_meta; GenotypeFile::file):
header labels, individuals and marker positions; unit test.
* src/routes/genotypes.rs: the two new handlers, token-gated like the
listing, plus `file` in the listing and counts in the detail.
* tests/genotypes.rs: the listing carries file names but no paths.
* README.md, docs/genotypes.md, docs/www/ai-instruction.md: what the
three routes give, and that the matrix is not among them.
Diffstat (limited to 'docs/www')
| -rw-r--r-- | docs/www/ai-instruction.md | 12 |
1 files changed, 10 insertions, 2 deletions
diff --git a/docs/www/ai-instruction.md b/docs/www/ai-instruction.md index effe1c8..bb71a98 100644 --- a/docs/www/ai-instruction.md +++ b/docs/www/ai-instruction.md @@ -275,8 +275,16 @@ That lists each `version` (e.g. `BXD.8`), its `title`, `individuals`, `markers`, whether `gemma` and `rqtl` can run on it, and which one is `current` — what a mapping gets if you name no version. Pass `group` to the mapping and the server resolves the files; add -`"genotype_version": "BXD.4"` for an older panel. Paths are deliberately -not part of the API: you never name, or see, a genotype file. +`"genotype_version": "BXD.4"` for an older panel. + +You do not have to map blind: `GET /groups/{group}/genotypes/{version}` +gives that panel's file name, its header labels (`type`, `mat`, `pat`, +…), the individuals in file order, the strains a mapping actually uses, +and per-chromosome marker counts, and +`GET /groups/{group}/genotypes/{version}/markers?page=&limit=&chr=` gives +the markers themselves (name, chr, cM, Mb). That is enough to frame or +check a mapping. The genotype calls are not served — do not look for +them, and do not try to download a genotype matrix. The values and the genotype file have to describe the **same individuals, in the same order, one for one**: one entry per panel |
