Golden Files¶

Golden files are Parquet (and Newick/FASTA) snapshots of phyloseq's built-in reference datasets. Tests use them as the ground truth for numerical comparisons without requiring an R installation at test time.

Datasets¶

Name	Samples	Taxa	Has tree	Has refseq
`GlobalPatterns`	26	19,216	yes	no
`enterotype`	280	553	no	no
`esophagus`	3	58	yes	no
`soilrep`	56	16,825	no	no

File layout¶

tests/golden/
├── PROVENANCE.json           # R/package versions + generation timestamp
├── GlobalPatterns/
│   ├── otu_table.parquet
│   ├── sample_data.parquet
│   ├── tax_table.parquet
│   ├── phy_tree.nwk
│   ├── taxa_sums.parquet
│   └── sample_sums.parquet
├── enterotype/
│   ├── otu_table.parquet
│   ├── sample_data.parquet
│   ├── tax_table.parquet
│   ├── taxa_sums.parquet
│   └── sample_sums.parquet
├── esophagus/
│   └── …
└── soilrep/
    └── …

Regenerating golden files¶

You need R ≥ 4.3 and Bioconductor phyloseq ≥ 1.54. The simplest path is the provided Docker image:

docker build -f scripts/Dockerfile.golden -t pyloseq-golden .
docker run --rm -v "$(pwd)/tests/golden:/repo/tests/golden" pyloseq-golden

Or directly on a machine with R:

Rscript scripts/requirements.R   # install deps (once)
Rscript scripts/generate_golden.R

Pinning R package versions¶

PROVENANCE.json records the exact versions used. To reproduce a specific generation:

BiocManager::install("phyloseq", version = "1.54.0")

See scripts/Dockerfile.golden for the fully pinned Docker image used in CI.