Where matchdb keeps its data — md_data

The build handles multi-gigabyte sources and produces a database of similar size, so nothing lives in the package repository. md_data_root() resolves the root in this order: the MATCHDB_DATA environment variable, then ~/Documents/matchdb.

Usage

md_data_root()

md_path(..., create = FALSE)

md_data_init(quiet = FALSE)

Arguments

...: Path components appended to the root.
create: Create the directory if it does not exist.
quiet: For md_data_init(), do not print the root it set up.

Value

A normalized absolute path.

Details

The layout under the root is fixed, because provenance depends on it:

sources/<source>/<version>/: one pinned, immutable copy of an upstream file. Never overwritten - a new upstream release is a new version directory. See md_source_pin().
db/: MATCHDB.duckdb, the relational source of truth.
publish/<label>/<spec_hash>/: compiled references (md_publish()).
tranches/<build>/: the database as parquet, split by layer and class (md_publish_tranches()).
logs/: the workspace log: runs-index.csv (one row per fetch, compile or publish, never rewritten), one detail file per run, and state.json (what is here now). See md_log().
work/: scratch. Safe to delete; never an input to anything.