The manifest written by lnk_scenario_export() is the only index of a
published tree: the bucket is not listable, so readers go through it. Each
part's Hive fields (config, species, code) come from its key, the one
place they are written, and its href resolves relative to the manifest.
Value
A data frame, one row per part: dataset, config, species,
code, drainage, wsgs, key, href, rows, bytes, md5, and
version_id (the S3 object version a publish recorded, NA in a tree
built locally or published before link#334; lnk_db_load() fetches that
version, so a reader pinned to a manifest reads the parts it lists).
Attributes: manifest (where it was read), id, built, provenance,
datasets (each dataset's column contract as the export wrote it),
configs (per config: the log rows of each build, config hashes, the
builds' tokens, discharge, bundles) and files (the provenance and
bundle files).
Details
manifest may be a local path, an https:// URL (an anonymous read, for a
public prefix) or an s3:// URL. An s3:// read goes through the aws
command line with the caller's credentials, which a private prefix needs:
habitat/ is staged private (link#308).
See also
Other export:
lnk_db_load(),
lnk_downstream_files(),
lnk_downstream_load(),
lnk_scenario_export()
