Skip to contents

Works out the hash of the file a wave was read from and of the samples it holds, with its length, format and size, and keeps them in the metadata slot of the wave as calculatedMetadata, beside the audioMetadata read from the file. It only calculates: nothing is compared with what the source of a recording says of it.

Usage

calculateMetadata(w)

Arguments

w

A Wave, WaveMC, TaggedWave or TaggedWaveMC object. A Wave or WaveMC object is tagged first.

Value

A TaggedWave or TaggedWaveMC object holding the fields in w@metadata$calculatedMetadata.

Details

The fields are those audioBlast calculates of the recordings it holds, under the same names without their calculated_ prefix, with the hash of the samples added:

  • hash: the SHA-256 hash of the bytes of the file, in lower case, as audioBlast and the hash column of a RecordingSet hold it.

  • hash_method: the digest hash is, "sha256".

  • wave_hash: a SHA-256 digest of every sample the wave holds, as tagApply() takes of the waves it records under options(sonicscrewdriver.provenance="full"). Two files holding the same audio under different tags have the same wave hash and different file hashes.

  • wave_hash_method: "full", as every sample is read.

  • frames: the number of sample frames the wave holds.

  • duration: frames divided by sample_rate, in seconds.

  • sample_rate: in Hz.

  • channels: the number of channels.

  • bit_depth: the bits a sample takes in the file, or NA for a lossy format such as MP3, which stores no samples as such.

  • bit_rate: the bits a second of audio takes in the file.

  • codec: the format the audio is in, as FFmpeg names it, e.g. "pcm_s16le" or "mp3".

  • bytes: the size of the file.

  • at: when they were calculated, in UTC.

  • by: the version of sonicscrewdriver that calculated them.

The file is the one readAudio() read the wave from, which it records on the wave. The file is hashed when this is called, so it should be called before the file is changed or deleted. A wave read from part of a file has the hash and bytes of the whole file, and the frames and duration of the part. A wave that was not read from a file has NA for hash, hash_method, bit_rate, codec and bytes, and its own bit depth.

For a wave file, codec, bit_depth and bit_rate are read from its format chunk. For any other file they are asked of the av package where it is installed, and are otherwise NA, but for the codec of an MP3 file.

The fields describe the file as it was read. They are carried unchanged onto the waves tagApply() makes from the wave, and are not calculated again when a process changes its samples, as the record of each process holds a digest of the wave it made.

Hashing every sample takes about a second per minute of stereo 48 kHz audio.

Examples

f <- system.file("extdata/AUDIOMOTH.WAV", package="sonicscrewdriver")
w <- calculateMetadata(readAudio(f))
w@metadata$calculatedMetadata$hash
#> [1] "b572a554a0b1531be3a6bc637c37d68e0fb50d7152bb492b2c7e91f5a6a86bec"
w@metadata$calculatedMetadata$duration
#> [1] 5