Sources are kept by content, not by date: {prefix}/{provider}/{dataset}/{md5}/{file}.
A provider that re-publishes a corrected file under the same name gets a second
object beside the first, never an overwrite, and a file read by many renders is
stored once.
Usage
source_archive_path(md5, file, provider, dataset, prefix = "archive/sources")
Arguments
- md5
hexadecimal MD5 of the file
- file
the file's base name
- provider, dataset
the ingest's provider and dataset
- prefix
archive prefix within the bucket (default "archive/sources")
Value
object paths, relative to the bucket