Check Whether Referenced Clips Still Exist on Disk

Reconciles a table of clip references (for example a near-duplicate removal plan produced from BirdNET embeddings) against a clip library laid out as one sub-directory per class, i.e. `dir/<species>/<clip>`. Embeddings can drift out of sync with the library over time, so a clip named in the reference may no longer be present on disk. Use this before deleting or moving files so you never act on a stale path.

Usage

check_clips_on_disk(df, dir, species_col = "species", clip_col = "clip")

Arguments

df

A data frame containing at least the species and clip columns named by `species_col` and `clip_col`.

dir

Character. Path to the library root, whose immediate sub-directories are the class folders. Must exist.

species_col

Character. Name of the column holding the class / species-directory name. Default `"species"`.

clip_col

Character. Name of the column holding the clip file name. Default `"clip"`.

Value

`df` unchanged except for two appended columns:

`path`

Full expected path, `file.path(dir, species, clip)`.

`exists`

Logical; `TRUE` when that file is present on disk.

Row order is preserved. A message reports how many of the referenced clips were found.

Details

Existence is tested with [file.exists()] on the composed path; no hashing or content comparison is done (see [find_duplicate_wavs()] for content-level duplicate detection). Only the exact `species`/`clip` combination is checked — a clip that was moved to a different class folder counts as missing.

See also

Other audio: analyze_call_properties(), biggest_files(), convert_to_wav(), find_duplicate_wavs(), print.training_dataset_summary(), quarantine_clips(), training_dataset_summary()

Examples

if (FALSE) { # \dontrun{
# plan produced by birdnetEmbed::near_duplicate_removal_plan()
checked <- check_clips_on_disk(plan, "OneDrive-UNSW/call_library/reallybig")

# Only remove files that are both flagged and actually present:
to_delete <- checked[checked$action == "remove" & checked$exists, "path"]
} # }