library(warbleR)

##
# This is an examle that shows how to:
# 1. choose bird call samples from xeno-canto by different selection types
# 2. download the files
# 3. turn the mp3 files into wavs for further processing
# 4. fine tune the spectrogram creator and create spectrograms of each wav file
# 5. fine tune the auto_detection of warbleR and create detections
# 6. mash it all together and print spectrograms with detections
##

# Create a new folder inside a new tempdir and set it to your working dir
wd <- file.path(tempdir(), "xeno-canto_example")
dir.create(wd)
setwd(wd)

##
# Do the queries
##

# Query xeno-canto for all Phaethornis recordings (e.g., by genus)
Phae <- query_xc(qword = "Phaethornis", download = FALSE)
# Check out the structure of resulting the data frame
str(Phae)
# Query xeno-canto for all Phaethornis longirostris recordings
Phae.lon <- query_xc(qword = "Phaethornis longirostris", download = FALSE)
# Check out the structure of resulting the data frame
str(Phae.lon)

##
# filter xeno-canto recordings by quality, signal type and locality
##

# How many recordings are available for Phaethornis longirostris?
nrow(Phae.lon)
# How many signal types exist in the xeno-canto metadata?
unique(Phae.lon$Vocalization_type)
# How many recordings per signal type?
table(Phae.lon$Vocalization_type)

##
# Filter the metadata to select the signals we want to retain
##

# First by quality
Phae.lon <- Phae.lon[Phae.lon$Quality == "A", ]
nrow(Phae.lon)
# Then by signal type
Phae.lon.song <- Phae.lon[grep("song", Phae.lon$Vocalization_type, ignore.case = TRUE), ]
nrow(Phae.lon.song)
# Finally by locality
Phae.lon.LS <- Phae.lon.song[grep("La Selva Biological Station, Sarapiqui, Heredia", Phae.lon.song$Locality, ignore.case = FALSE), ]
nrow(Phae.lon.LS)
# Check resulting data frame, 3 recordings remain
str(Phae.lon.LS)
# check the location
map_xc(Phae.lon.LS, img = FALSE)


##
# Once we're sure the recordings fit, it's time to download the files, also save
# the metadata as .csv file
##

# Download sound files
query_xc(X = Phae.lon.LS)
# Save the metadata object as a .csv file
write.csv(Phae.lon.LS, "Phae_lon.LS.csv", row.names = FALSE)

##
# xeno-canto maintains recordings in mp3 format due to file size restrictions.
# However, we require wav format for all downstream analyses
##

# here we are downsampling the original sampling rate of 44.1 kHz
# to speed up downstream analyses in the vignette series
mp32wav(samp.rate = 22.05)
# Use checkwavs to see if wav files can be read
check_wavs()

##
# Make long spectrograms of whole recordings
##

# Create a vector of all the recordings in the directory
wavs <- list.files(pattern = "wav$")
# Print this object to see all sound files
wavs
# How long are these files? this will determine number of pages returned by full_spectrograms
duration_wavs(wavs)
# ovlp = 10 to speed up function
# tiff image files are better quality and are faster to produce
full_spectrograms(flist = wavs, ovlp = 10, it = "tiff")
# We can zoom in on the frequency axis by changing flim,
# the number of seconds per row, and number of rows
full_spectrograms(flist = wavs, flim = c(2, 10), sxrow = 6, rows = 15, ovlp = 10, it = "tiff")

##
# Once satisfied with the argument settings we can make long spectrograms for all the sound files.
##

# Make long spectrograms for the xeno-canto sound files
full_spectrograms(flim = c(2, 10), ovlp = 10, sxrow = 6, rows = 15, it = "jpeg", flist = wavs)
# Concatenate full_spectrograms image files into a single PDF per recording
# full_spectrograms images must be jpegs to do this
full_spectrogram2pdf(keep.img = FALSE, overwrite = TRUE)

##
# Automatically detect signals with auto_detect
##

# Select a subset of sound files
# Reinitialize the wav object
wavs <- list.files(pattern = ".wav$", ignore.case = TRUE)
# Set a seed so we all have the same results
set.seed(1)
sub <- wavs[sample(1:length(wavs), 3)]

##
# Run auto_detec() on subset of recordings
##

# play around with the auto detection, setting the values accordingly.
# Phaethornis longirostris songs have frequencies between 2 and 10 kHz and durations between 0.05 and 0.5 s.
Phae.ad <- auto_detec(
  path = wd,
  threshold = 20, # amplitude threshold in %
  ssmooth = 900, # amplitude envelope with sum smooth
  bp = c(2, 10), # bandpass filter (between 2 and 10 kHz)
  wl = 300, # window for ffilter bandpass
  parallel = 6*2 # how many cores shall be used in parallel (*2 due to hyper threading)
)

# Let's look at the number of selections per sound file
table(Phae.ad$sound.files)

# When we're done, create an image with all detections using full spectrograms
full_spectrograms(
                  flim = c(2, 10),
                  ovlp = 10,
                  sxrow = 6,
                  rows = 15,
                  it = "jpeg",
                  flist = wavs,
                  X = auto_detec(
                    path = wd,
                    threshold = 20,
                    ssmooth = 900,
                    bp = c(2, 10),
                    wl = 300,
                    parallel = 6*2)
)

# combine the image into a single pdf per species like before
full_spectrogram2pdf(keep.img = FALSE, overwrite = TRUE)