Verify a data capsule

Re-checks a capsule directory against its provenance record, one row per check: the record is readable, the data file is present with its pinned SHA-256, size and row count, the recorded schema still validates, a results manifest still reproduces its recorded verdicts, a pinned script still has its digest, and the data are not synthetic. This is rmoriebricklayer::vrfcps(); see its help for every row.

Usage

morie_capsule_verify(
  dir,
  provenance_file = "data_provenance.json",
  data_file = NULL,
  manifest_file = NULL,
  script_file = NULL
)

Arguments

dir

The capsule directory, a single string.

provenance_file

The provenance record, relative to dir.

data_file

The data file, relative to dir. NULL uses the record's resource$filename.

manifest_file

A results manifest (JSON) relative to dir. NULL uses manifest.json when the capsule has one.

script_file

An analysis script relative to dir, checked against the pinned script digest. NULL uses the one the record or the manifest names.

Value

A bricklayer_capsule_check: a list with ok (logical), checks (data frame of check, ok, detail), required and missing (character).

Details

Entirely offline; nothing is written. A failed check is a row, not an error: a provenance record that does not parse as JSON is a failed provenance_readable row, as in morie (vrfcps() itself stops on one). The errors are dir not an existing directory and an argument that is not one string.

See also

Other capsule: morie_capsule_bundle(), morie_capsule_compare(), morie_capsule_drift(), morie_capsule_manifest(), morie_capsule_sign(), morie_capsule_write()

Examples

dir <- file.path(tempdir(), "capsule-verify")
dir.create(dir, showWarnings = FALSE)
write.csv(data.frame(id = 1:3), file.path(dir, "data.csv"), row.names = FALSE)
morie_capsule_write(dir, "data.csv", retrieved = "2026-10-20")
morie_capsule_verify(dir)
#> Capsule verification: intact (7 of 7 checks passed)
#>   ok    provenance_readable  /tmp/Rtmp0CK4Yk/capsule-verify/data_provenance.json
#>   ok    data_present         /tmp/Rtmp0CK4Yk/capsule-verify/data.csv
#>   ok    data_sha256          da4aaa3a009d063f36d0515f2b50a11bcc578264aa7c45504b46fedbbae4f8f7
#>   ok    data_size_bytes      11 bytes on disk
#>   ok    data_readable        /tmp/Rtmp0CK4Yk/capsule-verify/data.csv
#>   ok    data_row_count       3 rows on disk
#>   ok    data_not_synthetic   no synthetic marker: sidecar, manifest and provenance all say real data

# an edited data file fails its digest, size and row count
write.csv(data.frame(id = 1:4), file.path(dir, "data.csv"), row.names = FALSE)
chk <- morie_capsule_verify(dir)
chk$checks[!chk$checks$ok, c("check", "detail")]
#>             check
#> 3     data_sha256
#> 4 data_size_bytes
#> 6  data_row_count
#>                                                                                                                                            detail
#> 3 expected da4aaa3a009d063f36d0515f2b50a11bcc578264aa7c45504b46fedbbae4f8f7, got b24bb2c082053f4d1f5db97140babb86d4cc7e31b49be4e3c4cf94ca87a86d2b
#> 4                                                                                                                                13 bytes on disk
#> 6                                                                                                                                  4 rows on disk
unlink(dir, recursive = TRUE)