Each recording is isolated: failures are recorded without discarding successful
results. Resume requires matching input checksums, recipe, analysis parameters,
package implementation, R version and dependency versions, plus an intact
versioned output record. Legacy or damaged records are recomputed. The output
checksum detects accidental changes; it does not authenticate an untrusted
file. Failed recomputation leaves the previous result file unchanged. Input and output
paths are local; no network or hardware acquisition is performed.
TIVITA input identity includes the consumed metadata log and the names/paths
of reference images returned by the reader. Reference image bytes are not
read by the analysis and are not checksummed.
For Cubert sessions, file bytes and explicit reader arguments are tracked,
but external vendor SDK settings and assets are not enumerated. Use
resume = FALSE when that external configuration may have changed.
Arguments
- manifest
Data frame with distinct
idandpathcolumns. Optionalreader_argslist-column provides per-recording reader arguments.- recipe
A recipe from
hs_recipe().- out_dir
Directory for per-recording RDS files and
run-summary.csv.- analysis
One of
"summary","beer_lambert", or"pca".- analysis_args
Named list of arguments for the analysis function.
- resume
Logical. Reuse intact results only if their fingerprint matches.
Examples
# \donttest{
dir <- file.path(tempdir(), "hs-batch-example")
hdr <- hs_example_files(tempdir())
manifest <- data.frame(id = "example", path = hdr)
first <- hs_batch(manifest, out_dir = dir)
#> Reading ENVI file: example.dat
#> 10 cols x 10 rows x 11 bands (bsq)
#> Wavelength range: 430-510 nm
first$status
#> [1] "complete"
# A second run reuses the intact result
hs_batch(manifest, out_dir = dir)$status
#> [1] "resumed"
unlink(dir, recursive = TRUE)
# }