Skip to contents

Version stamps of derived outputs

Source: R/stamps.R
write_stamp.Rd

Derived outputs (cleaned texts, matches, LLM caches, dictionaries applied) depend on a method, its version and its inputs. A stamp is a small text file saved next to an output (<output>.stamp) that records them:

  • write_stamp() saves the stamp of an output;

  • read_stamp() reads it;

  • check_stamp() compares it with the current method, version and inputs, and refuses (by default) an output built under another version, so that a stale cache is never used by mistake.

Usage

write_stamp(path, method, version, inputs = NULL, notes = NULL)

read_stamp(path)

check_stamp(
  path,
  method,
  version,
  inputs = NULL,
  action = c("abort", "warn", "none")
)

Arguments

path

Character. Path of the output file.

method

Character. Name of the method (e.g. "generic_cleaning").

version

Character. Version of the method (e.g. "v3"). Bump it whenever a rule changes.

inputs

NULL, or a fingerprint of the inputs from content_hash().

notes

Character or NULL. Free text saved with the stamp.

action

Character. What check_stamp() does when the stamp does not match: "abort", "warn", or "none" (only return FALSE).

Value

write_stamp(): the path of the stamp, invisibly. read_stamp(): a named list, or NULL when there is no stamp. check_stamp(): TRUE when the stamp matches, FALSE otherwise (invisibly).

Details

The stamp also records the date, the R version and the version of coRpusdb. It is written in Debian control format (one field: value per line), readable by eye and by any language.

See also

Other versions: content_hash()

Examples

out <- tempfile(fileext = ".rds")
saveRDS(data.frame(doc_id = "a:1", text = "cleaned"), out)
write_stamp(out, method = "generic_cleaning", version = "v3")
read_stamp(out)
#> $method
#> [1] "generic_cleaning"
#> 
#> $version
#> [1] "v3"
#> 
#> $created_at
#> [1] "2026-11-08 17:50:33 CET"
#> 
#> $r_version
#> [1] "4.6.2"
#> 
#> $coRpusdb_version
#> [1] "0.2.0"
#> 
check_stamp(out, method = "generic_cleaning", version = "v3")
check_stamp(out, method = "generic_cleaning", version = "v4", action = "warn")
#> Warning: /tmp/RtmpKqItGK/file79980262156cf.rds was not built by the current method.
#> ✖ version v3, expected v4