inst/shiny/pitch_app/server.R

In soundgen: Sound Synthesis and Acoustic Analysis

# pitch_app()
#
# To do: maybe analyze one bit at a time like in formant_app
#
# # tip: to read the output, do smth like:
# a = read.csv('~/Downloads/output.csv', stringsAsFactors = FALSE)
# as.numeric(unlist(strsplit(a$pitch, ',')))

# shinyBS needs to be included as a dependency (instead of just "import"):
# see https://stackoverflow.com/questions/52649138/including-shinybs-in-a-package

server = function(input, output, session) {
  # make overlaid plots resizable (js fix)
  shinyjs::js$inheritSize(parentDiv = 'specDiv')
  shinyjs::js$inheritSize(parentDiv = 'oscDiv')

  myPars = reactiveValues(
    print = FALSE,       # if TRUE, some functions print a message to the console when called
    debugQn = FALSE,     # for debugging - click "?" to step into the code
    zoomFactor = 2,     # zoom buttons change time zoom by this factor
    zoomFactor_freq = 1.5,  # same for frequency
    out = NULL,          # for storing the output
    drawSpec = TRUE,
    shinyTip_show = 1000,      # delay until showing a tip (ms)
    shinyTip_hide = 0,         # delay until hiding a tip (ms)
    slider_ms = 50,            # how often to update play slider
    scrollFactor = .75,        # how far to scroll on arrow press/click
    wheelScrollFactor = .1,    # how far to scroll on mouse wheel (prop of xlim)
    cursor = 0,
    pitchCert_mult = NULL,     # old pitch prior
    initDur = 1500,            # initial duration to plot (ms)
    spec_xlim = c(0, 1500),
    play = list(on = FALSE)
  )
  tooltip_options = list(delay = list(show = 1000, hide = 0))

  # clean-up of www/ folder: remove all files except temp.wav
  # if (!dir.exists("www")) dir.create("www")  # otherwise trouble with shinyapps.io
  if (file.exists('www/temp.csv')) {
    showModal(modalDialog(
      title = "Unsaved data",
      "Found unsaved data from a prevous session. Append to the new output?",
      easyClose = TRUE,
      footer = tagList(
        actionButton("discard", "Discard"),
        actionButton("append", "Append")
      )
    ))
  }
  observeEvent(input$discard, {
    file.remove('www/temp.csv')
    removeModal()
  })
  observeEvent(input$append, {
    myPars$out = read.csv('www/temp.csv')
    removeModal()
  })

  files = list.files('www/', pattern = '.wav')
  files = files[files != 'temp.wav']
  for (f in files){
    file.remove(paste0('www/', f))
  }

  reset = function() {
    if (myPars$print) print('Resetting...')
    myPars$pitch = NULL       # pitch contour
    myPars$pitchCands = NULL  # matrix of pitch candidates
    myPars$bp = NULL          # selected points (under brush)
    myPars$manual = data.frame(frame = NA, freq = NA)[-1, ]  # manually added pitch anchors
    myPars$manualUnv = numeric()                             # manually unvoiced frames
    myPars$drawSpec = FALSE   # prevent the spectrogram from being redrawn needlessly
    # (only draw it after extracting it)
    myPars$spec = NULL
    myPars$selection = NULL
    myPars$cursor = 0
    myPars$spectrogram_brush = NULL
  }

  resetSliders = function() {
    if (myPars$print) print('Resetting sliders...')
    sliders_to_reset = names(input)[which(names(input) %in% rownames(defaults_analyze))]
    for (v in sliders_to_reset) {
      new_value = defaults_analyze[v, 'default']
      try(updateSliderInput(session, v, value = new_value))
      try(updateNumericInput(session, v, value = new_value))
      updateSelectInput(session, 'wn', selected = 'gaussian')
      updateCheckboxGroupInput(session, 'pitchMethods', selected = c('dom', 'autocor'))
      updateCheckboxGroupInput(session, 'summaryFun', selected = c('mean', 'sd'))
      updateTextInput(session, 'summaryFun_text', value = '')
      updateSelectInput(session, 'pathfinding', selected = 'fast')
      updateSliderInput(session, 'spec_ylim', value=c(0, defaults_analyze['spec_ylim','default']))
      updateRadioButtons(session, 'spec_colorTheme', selected='bw')
      updateSelectInput(session, 'osc', selected = 'linear')
    }
  }
  observeEvent(input$reset_to_def, resetSliders())


  observeEvent(input$loadAudio, {
    if (myPars$print) print('Loading audio...')
    reset()  # also triggers done()

    # work only with audio files
    idx_audio = which(apply(matrix(input$loadAudio$type), 1, function(x) {
      grepl('audio', x, fixed = TRUE)
    }))
    if (length(idx_audio) > 0) {
      myPars$fileList = input$loadAudio[idx_audio, ]
      myPars$n = 1   # file number in queue
      myPars$nFiles = nrow(myPars$fileList)  # number of uploaded files in queue
      # set up a list for storing manual anchors for each of uploaded files
      myPars$history = vector('list', length = myPars$nFiles)
      names(myPars$history) = myPars$fileList$name
      for (i in 1:length(myPars$history)) {
        myPars$history[[i]] = list(
          detailed = NULL, summary = NULL,
          manual = NULL, manualUnv = NULL)
      }
    }

    # if there is a csv among the uploaded files, use the annotations in it
    ext = substr(input$loadAudio$name,
                 (nchar(input$loadAudio$name) - 2),
                 nchar(input$loadAudio$name))
    old_out_idx = which(ext == 'csv')[1]  # grab the first csv, if any
    if (!is.na(old_out_idx)) {
      user_ann = read.csv(input$loadAudio$datapath[old_out_idx], stringsAsFactors = FALSE)
      oblig_cols = c('file', 'time', 'pitch')
      if (nrow(user_ann) > 0 &
          !any(!oblig_cols %in% colnames(user_ann))) {
        if (is.null(myPars$out)) {
          myPars$out = user_ann[, oblig_cols]
        } else {
          # Note: rbind_fill ~ merge(all = TRUE), except that merge removes
          # duplicates, which is highly undesirable in this case because it
          # removes a row from the output when we go back and forth between files
          myPars$out = soundgen:::rbind_fill(myPars$out, user_ann[, oblig_cols])
          # remove duplicate rows
          myPars$out = unique(myPars$out)
        }
        # save f0 contours from the csv as manual anchors
        for (i in 1:nrow(user_ann)) {
          time_i = suppressWarnings(as.numeric(unlist(strsplit(user_ann$time[i], ','))))
          pitch_i = suppressWarnings(as.numeric(unlist(strsplit(user_ann$pitch[i], ','))))

          frame_i = round(time_i / input$step)
          step_i = time_i[2] - time_i[1]
          if (step_i != input$step) pitch_i = pitch_i[!duplicated(frame_i)]
          manual_i = data.frame(
            frame = frame_i, freq = pitch_i
          )
          myPars$history[[user_ann$file[i]]] = list(
            manual = manual_i[!is.na(manual_i$freq), ],
            manualUnv = manual_i$frame[is.na(manual_i$freq)]
          )
        }
      }
    }

    choices = as.list(myPars$fileList$name)
    names(choices) = myPars$fileList$name
    updateSelectInput(session, 'fileList',
                      choices = as.list(myPars$fileList$name))
    # readAudio(1)  # read the first sound in queue
  })

  observeEvent(input$showpanel, {
    if(input$showpanel == TRUE) {
      shinyjs::removeCssClass("Main", "col-sm-12")
      shinyjs::addCssClass("Main", "col-sm-9")
      shinyjs::show(id = "Sidebar")
      shinyjs::enable(id = "Sidebar")
    }
    else {
      shinyjs::removeCssClass("Main", "col-sm-9")
      shinyjs::addCssClass("Main", "col-sm-12")
      shinyjs::hide(id = "Sidebar")
    }
  })

  readAudio = function(i) {
    # reads an audio file with tuneR::readWave
    if (myPars$print) print('Reading audio...')
    temp = myPars$fileList[i, ]
    myPars$myAudio_filename = temp$name
    myPars$myAudio_path = temp$datapath
    myPars$myAudio_type = temp$type

    extension = substr(myPars$myAudio_filename,
                       nchar(myPars$myAudio_filename) - 2, nchar(myPars$myAudio_filename))
    if (extension == 'wav' | extension == 'WAV') {
      myPars$temp_audio = tuneR::readWave(temp$datapath)
    } else if (extension == 'mp3' | extension == 'MP3') {
      myPars$temp_audio = tuneR::readMP3(temp$datapath)
    } else {
      warning('Input not recognized: must be a wav or mp3 file')
    }

    myPars$myAudio = as.numeric(myPars$temp_audio@left)
    myPars$ls = length(myPars$myAudio)
    myPars$samplingRate = myPars$temp_audio@samp.rate
    myPars$maxAmpl = 2 ^ (myPars$temp_audio@bit - 1)
    if (input$normalizeInput) {
      myPars$myAudio = myPars$myAudio / max(abs(myPars$myAudio)) * myPars$maxAmpl
    }
    updateSliderInput(session, 'spec_ylim', max = myPars$samplingRate / 2 / 1000)  # check!!!
    myPars$dur = round(length(myPars$temp_audio@left) / myPars$temp_audio@samp.rate * 1000)
    myPars$myAudio_list = list(
      sound = myPars$myAudio,
      samplingRate = myPars$samplingRate,
      scale = myPars$maxAmpl,
      timeShift = 0,
      ls = length(myPars$myAudio),
      duration = myPars$dur / 1000
    )
    myPars$time = seq(1, myPars$dur, length.out = myPars$ls)
    myPars$spec_xlim = c(0, min(myPars$initDur, myPars$dur))

    # update info - file number ... out of ...
    updateSelectInput(session, 'fileList',
                      label = NULL,
                      selected = myPars$fileList$name[myPars$n])
    file_lab = paste0('File ', myPars$n, ' of ', myPars$nFiles)
    output$fileN = renderUI(HTML(file_lab))

    # if we've already worked with this file in current session,
    # re-load the manual anchors
    hist = myPars$history[[myPars$myAudio_filename]]
    if (!is.null(hist$manual)) myPars$manual = hist$manual
    if (!is.null(hist$manualUnv)) myPars$manualUnv = hist$manualUnv
  }

  extractSpectrogram = observe({
    # Instead of re-loading the file every time, could save the spectrogram
    # matrix and re-draw manually with soundgen:::filled.contour.mod
    if (!is.null(myPars$myAudio)) {
      if (myPars$print) print('Extracting spectrogram...')
      if (input$specType == 'reassigned') {
        temp_spec = soundgen:::.spectrogram(
          myPars$myAudio_list,
          dynamicRange = input$dynamicRange,
          windowLength = input$reass_windowLength,
          step = input$reass_step,
          specType = 'reassigned',
          wn = input$wn,
          zp = 2 ^ input$zp,
          output = 'all',
          plot = FALSE
        )
        myPars$spec = temp_spec$processed
        myPars$reassigned = temp_spec$reassigned
      } else {
        myPars$spec = soundgen:::.spectrogram(
          myPars$myAudio_list,
          dynamicRange = input$dynamicRange,
          windowLength = input$windowLength,
          step = input$step,
          specType = input$specType,
          wn = input$wn,
          zp = 2 ^ input$zp,
          contrast = input$specContrast,
          brightness = input$specBrightness,
          blur = c(input$blur_freq, input$blur_time),
          output = 'processed',
          plot = FALSE
        )
      }
    }
  })

  writeAudioFile = observeEvent(myPars$temp_audio, {
    if (myPars$print) print('Writing audio file...')
    # Method: saves a temporary audio file in 'www/'. This is a workaround since
    # html tag for some reason cannot play myPars$myAudio_path (although feeding
    # it to spectrogram works - so probably only works within R). Alternatives:
    # soundgen::play() or shinyFiles library

    # first remove the previous sound file to avoid cluttering up the www/ folder
    if (!is.null(myPars$myfile)){
      file.remove(paste0('www/', myPars$myfile))
    }
    randomID = paste(sample(c(letters, 0:9), 8, replace = TRUE), collapse = '')
    myPars$myfile = paste0(randomID, '.wav')
    # this is the new sound file. NB: has to be saved in www/ !!!
    seewave::savewav(myPars$temp_audio,
                     f = myPars$samplingRate,
                     filename = paste0('www/', myPars$myfile))
    output$htmlAudio = renderUI(
      tags$audio(src = myPars$myfile, type = myPars$myAudio_type,
                 id = 'myAudio',
                 style = "display: none; transform: scale(0.75); transform-origin: 0 0;")
    )
  })


  # Updating spec / osc stuff to speed up plotting
  observe({
    if (!is.null(myPars$myAudio)) {
      if (myPars$print) print('Scaling audio...')
      if (input$osc == 'dB') {
        myPars$myAudio_scaled = osc(
          myPars$myAudio,
          dynamicRange = input$dynamicRange,
          dB = TRUE,
          maxAmpl = myPars$maxAmpl,
          plot = FALSE,
          returnWave = TRUE)
        myPars$ylim_osc = c(-2 * input$dynamicRange, 0)
      } else {
        myPars$myAudio_scaled = myPars$myAudio
        myPars$ylim_osc = c(-myPars$maxAmpl, myPars$maxAmpl)
      }
    }
  })

  observe({
    # Cut just the part of spec currently needed for plotting
    # (faster than plotting a huge matrix with xlim/ylim)
    if (!is.null(myPars$spec) & !is.null(myPars$myAudio_scaled)) {
      if (myPars$print) print('Trimming the spec & osc')
      # spec
      x = as.numeric(colnames(myPars$spec))
      idx_x = which(x >= (myPars$spec_xlim[1] / 1.05) &
                      x <= (myPars$spec_xlim[2] * 1.05))
      # 1.05 - a bit beyond b/c we use xlim/ylim and may get white space
      y = as.numeric(rownames(myPars$spec))
      idx_y = which(y >= (input$spec_ylim[1] / 1.05) &
                      y <= (input$spec_ylim[2] * 1.05))
      myPars$spec_trimmed = downsample_spec(
        myPars$spec[idx_y, idx_x],
        maxPoints = 10 ^ input$spec_maxPoints)
      # dim(myPars$spec_trimmed)

      # osc
      idx_s = max(1, (myPars$spec_xlim[1] / 1.05 * myPars$samplingRate / 1000)) :
        min(myPars$ls, (myPars$spec_xlim[2] * 1.05 * myPars$samplingRate / 1000))
      downs_osc = 10 ^ input$osc_maxPoints

      isolate({
        myPars$myAudio_trimmed = myPars$myAudio_scaled[idx_s]
        myPars$time_trimmed = myPars$time[idx_s]
        myPars$ls_trimmed = length(myPars$myAudio_trimmed)
        if (!is.null(myPars$myAudio_trimmed) &&
            myPars$ls_trimmed > downs_osc) {
          myseq = round(seq(1, myPars$ls_trimmed,
                            length.out = downs_osc))
          myPars$myAudio_trimmed = myPars$myAudio_trimmed[myseq]
          myPars$time_trimmed = myPars$time_trimmed[myseq]
          myPars$ls_trimmed = length(myseq)
        }
      })
      myPars$drawSpec = TRUE
    }
  })

  downsample_sound = function(x, maxPoints) {
    if (!is.null(myPars$myAudio_trimmed) &&
        myPars$ls_trimmed > (10 ^ input$osc_maxPoints)) {
      if (myPars$print) print('Downsampling osc')
      myseq = round(seq(1, myPars$ls_trimmed,
                        by = myPars$ls_trimmed / input$osc_maxPoints))
      myPars$myAudio_trimmed = myPars$myAudio_trimmed[myseq]
      myPars$ls_trimmed = length(myseq)
    }
  }

  downsample_spec = function(x, maxPoints) {
    lxy = nrow(x) * ncol(x)
    if (length(lxy) > 0 && lxy > maxPoints) {
      if (myPars$print) print('Downsampling spectrogram...')
      lx = ncol(x)  # time
      ly = nrow(x)  # freq
      downs = sqrt(lxy / maxPoints)
      seqx = round(seq(1, lx, length.out = lx / downs))
      seqy = round(seq(1, ly, length.out = ly / downs))
      out = x[seqy, seqx]
    } else {
      out = x
    }
    return(out)
  }

  # Actuall plotting of the spec / osc
  output$spectrogram = renderPlot({
    if (myPars$drawSpec == TRUE) {
      if (myPars$print) print('Drawing spectrogram...')
      par(mar = c(0.2, 2, 0.5, 2))
      # no need to save user's graphical par-s - revert to orig on exit
      if (is.null(myPars$myAudio_trimmed) | is.null(myPars$spec)) {
        plot(1:10, type = 'n', bty = 'n', axes = FALSE, xlab = '', ylab = '')
        text(
          x = 5, y = 5, cex = 3,
          labels =
             'Upload wav/mp3 file(s) to begin...\n
             Suggested max duration ~10 s')
      } else {
        if (input$specType != 'reassigned') {
          # rasterized spectrogram
          soundgen:::filled.contour.mod(
            x = as.numeric(colnames(myPars$spec_trimmed)),
            y = as.numeric(rownames(myPars$spec_trimmed)),
            z = t(myPars$spec_trimmed),
            levels = seq(0, 1, length = 30),
            color.palette = soundgen:::switchColorTheme(input$spec_colorTheme),
            log = '',
            yScale = 'linear',
            xlim = myPars$spec_xlim,
            xaxt = 'n',
            xaxs = 'i', xlab = '',
            ylab = '',
            main = '',
            ylim = input$spec_ylim
          )
        } else {
          # unrasterized reassigned spectrogram
          soundgen:::plotUnrasterized(
            myPars$reassigned,
            levels = seq(0, 1, length = 30),
            color.palette = soundgen:::switchColorTheme(input$spec_colorTheme),
            log = '',
            yScale = 'linear',
            xlim = myPars$spec_xlim,
            xaxt = 'n',
            xaxs = 'i', xlab = '',
            ylab = '',
            main = '',
            ylim = input$spec_ylim,
            cex = input$reass_cex
          )
        }

        # Add text label of file name
        ran_x = myPars$spec_xlim[2] - myPars$spec_xlim[1]
        ran_y = input$spec_ylim[2] - input$spec_ylim[1]
        text(x = myPars$spec_xlim[1] + ran_x * .01,
             y = input$spec_ylim[2] - ran_y * .01,
             labels = myPars$myAudio_filename,
             adj = c(0, 1))  # left, top
      }
    }
  })

  observe({
    myPars$specOver_opts = list(
      xlim = myPars$spec_xlim, ylim = input$spec_ylim,
      xaxs = "i", yaxs = "i",
      bty = 'n', xaxt = 'n', yaxt = 'n',
      xlab = '', ylab = '')
  })

  output$specOver = renderPlot({
    if (!is.null(myPars$spec)) {
      par(mar = c(0.2, 2, 0.5, 2))
      # bg=NA makes the image transparent

      # empty plot to enable hover/click events for the spectrogram underneath
      do.call(plot, c(list(
        x = myPars$spec_xlim,
        y = input$spec_ylim,
        type = 'n'),
        myPars$specOver_opts))

      if (!is.null(myPars$spectrogram_hover)) {
        # horizontal line
        do.call(points, c(list(
          x = myPars$spec_xlim,
          y = rep(myPars$spectrogram_hover$y, 2),
          type = 'l', lty = 3),
          myPars$specOver_opts))
        # frequency label
        do.call(text, list(
          x = myPars$spec_xlim[1],
          y = myPars$spectrogram_hover$y,
          labels = myPars$spectrogram_hover$freq,
          adj = c(0, 0)))
        # vertical line
        do.call(points, list(
          x = rep(myPars$spectrogram_hover$x, 2),
          y = input$spec_ylim,
          type = 'l', lty = 3))
        # time label
        do.call(text, list(
          x = myPars$spectrogram_hover$x,
          y = input$spec_ylim[1] + .025 * diff(input$spec_ylim),
          labels = myPars$spectrogram_hover$time,
          adj = .5))
      }

      # Add a rectangle showing the selected region
      if (!is.null(myPars$spectrogram_brush) & input$spectro_clickAct == 'select') {
        rect(
          xleft = myPars$spectrogram_brush$xmin,
          xright = myPars$spectrogram_brush$xmax,
          ybottom = input$spec_ylim[1],
          ytop = input$spec_ylim[2],
          col = rgb(.2, .2, .2, alpha = .15),
          border = NA
        )
      }

      # show prior
      if (is.list(myPars$prior)) {
        ran_x_5 = diff(range(myPars$spec_xlim)) * .05   # 5% of plot width
        points(myPars$spec_xlim[1] + myPars$prior$prob * ran_x_5,
               myPars$prior$freq / 1000, type = 'l', lty = 2)
        text(x = myPars$spec_xlim[1] + ran_x_5,
             y = ifelse(input$priorAdapt,
                        myPars$priorMean / 1000,
                        input$priorMean / 1000),
             pos = 2, labels = 'Prior', offset = 0.25)
        text(x = myPars$spec_xlim[1],
             y = input$pitchFloor / 1000,
             pos = 4, labels = 'floor', offset = 0)
        text(x = myPars$spec_xlim[1],
             y = input$pitchCeiling / 1000,
             pos = 4, labels = 'ceiling', offset = 0)
      }

      # add manual values to the list of pitch candidates for seamless plotting
      n = ncol(myPars$pitchCands$freq)
      # if (length(n>0) == 0 | length(nrow(myPars$manual)>0) == 0) browser()
      if (!is.null(myPars$pitchCands) &&
          (n > 0 & nrow(myPars$manual) > 0)) {
        temp_freq = rep(NA, n)
        temp_freq[myPars$manual$frame] = myPars$manual$freq
        temp_freq = rbind(myPars$pitchCands$freq, temp_freq)
        temp_cert = rbind(myPars$pitchCands$cert, rep(1, n))  # change 1 to input$manualCert
        temp_source = rbind(myPars$pitchCands$source, rep('manual', n))
      } else {
        temp_freq = myPars$pitchCands$freq
        temp_cert = myPars$pitchCands$cert
        temp_source = myPars$pitchCands$source
      }
      soundgen:::addPitchCands(
        pitchCands = temp_freq,
        pitchCert = temp_cert,
        pitchSource = temp_source,
        pitch = myPars$pitch,
        addToExistingPlot = TRUE,
        showLegend = TRUE,
        ylim = c(input$spec_ylim[1], input$spec_ylim[2]),
        pitchPlot = list(cex = input$spec_cex)
      )
    }
  }, bg = 'transparent')

  output$specSlider = renderPlot({
    if (!is.null(myPars$spec)) {
      par(mar = c(0.2, 2, 0.5, 2))
      # bg=NA or "transparent" makes the image transparent

      if (myPars$cursor == 0) {
        # just a transparent plot
        do.call(plot, c(list(
          x = 1, type = 'n'),
          myPars$specOver_opts))
      } else {
        # horizontal line at current play time
        do.call(plot, c(list(
          x = rep(myPars$cursor, 2),
          y = input$spec_ylim,
          type = 'l'),
          myPars$specOver_opts))
      }
    }
  }, bg = 'transparent')


  # oscillogram
  observe({
    myPars$oscOver_opts = list(
      xlim = myPars$spec_xlim,
      ylim = myPars$ylim_osc,
      xaxs = "i", yaxs = "i",
      bty = 'n', xaxt = 'n', yaxt = 'n',
      xlab = '', ylab = '')
  })

  output$oscillogram = renderPlot({
    if (!is.null(myPars$myAudio_trimmed)) {
      if (myPars$print) print('Drawing osc...')
      par(mar = c(2, 2, 0, 2))
      # plot(myPars$myAudio_trimmed, type = 'l')
      plot(myPars$time_trimmed,
           myPars$myAudio_trimmed,
           type = 'l',
           xlim = myPars$spec_xlim,
           ylim = myPars$ylim_osc,
           axes = FALSE, xaxs = "i", yaxs = "i", bty = 'o',
           xlab = 'Time, ms',
           ylab = '')
      box()
      time_location = axTicks(1)
      time_labels = soundgen:::convert_sec_to_hms(time_location / 1000, 3)
      axis(side = 1, at = time_location, labels = time_labels)
      if (input$osc == 'dB') {
        axis(side = 4, at = seq(0, input$dynamicRange, by = 10))
        mtext("dB", side = 2, line = 3)
      }
      abline(h = 0, lty = 2)
    }
  })

  output$oscOver = renderPlot({
    par(mar = c(2, 2, 0, 2))  # bg = 'transparent'
    # bg=NA makes the image transparent

    # empty plot to enable hover/click events for the spectrogram underneath
    do.call(plot, c(list(
      x = myPars$spec_xlim,
      y = myPars$ylim_osc,
      type = 'n'),
      myPars$oscOver_opts))

    # highlight voiced frames
    visible_frames = which(myPars$result$time >= myPars$spec_xlim[1] &
                             myPars$result$time <= myPars$spec_xlim[2])
    abline(v = myPars$result$time[visible_frames[1]] - input$step / 2, lty = 2, lwd = .5)
    for (f in visible_frames) {
      abline(v = myPars$result$time[f] + input$step / 2, lty = 2, lwd = .5)
      rect(
        xleft = myPars$result$time[f] - input$step / 2,
        xright = myPars$result$time[f] + input$step / 2,
        # ybottom = myPars$ylim_osc[1],
        # overlaps the osc -> problems with transparency on some platform
        ybottom = myPars$ylim_osc[2] - (myPars$ylim_osc[2] - myPars$ylim_osc[1]) / 4,
        ytop = myPars$ylim_osc[2],
        col = ifelse(f %in% myPars$voiced_frames, rgb(0, 0, 1, .25), NA),
        lwd = 0
      )
    }
  }, bg = 'transparent')

  observe({
    if (input$summaryFun_text != '') {
      myPars$summaryFun = input$summaryFun_text
    } else {
      myPars$summaryFun = input$summaryFun
    }
    # print(myPars$summaryFun)
  })

  observe(priority = 5, {
    # analyze the file (executes every time a slider with arg value is changed)
    if (!is.null(myPars$myAudio)) {
      if (myPars$print) print('Calling analyze()...')
      withProgress(message = 'Analyzing the sound...', value = 0.5, {
        temp_anal = soundgen:::.analyze(
          myPars$myAudio_list,
          windowLength = input$windowLength,
          step = input$step,
          wn = input$wn,
          zp = input$zp,
          dynamicRange = input$dynamicRange,
          silence = input$silence,
          entropyThres = input$entropyThres,
          nFormants = 0,     # disable formant tracking
          loudness = NULL,  # disable loudness analysis
          roughness = list(amRes = 0),  # no roughness analysis
          amRange = NULL,  # no AM analysis
          pitchMethods = input$pitchMethods,
          pitchFloor = input$pitchFloor,
          pitchCeiling = input$pitchCeiling,
          priorMean = isolate(input$priorMean),
          priorSD = isolate(input$priorSD),
          priorAdapt = input$priorAdapt,  # rerun analyze() if priorAdapt changes
          nCands = input$nCands,
          minVoicedCands = input$minVoicedCands,
          pitchDom = list(
            domThres = input$domThres,
            domSmooth = input$domSmooth
          ),
          pitchAutocor = list(
            autocorThres = input$autocorThres,
            # autocorSmooth = input$autocorSmooth,
            autocorUpsample = input$autocorUpsample,
            autocorBestPeak = input$autocorBestPeak
          ),
          pitchCep = list(
            cepThres = input$cepThres,
            cepZp = input$cepZp
          ),
          pitchSpec = list(
            specMethod = input$specMethod,
            specThres = input$specThres,
            specPeak = input$specPeak,
            specSinglePeakCert = input$specSinglePeakCert,
            specHNRslope = input$specHNRslope,
            specSmooth = input$specSmooth,
            specMerge = input$specMerge,
            specRatios = input$specRatios
          ),
          pitchHps = list(
            hpsThres = input$hpsThres,
            hpsNum = input$hpsNum,
            hpsNorm = input$hpsNorm,
            hpsPenalty = input$hpsPenalty
          ),
          pitchZc = list(
            zcThres = input$zcThres,
            zcWin = input$zcWin
          ),
          # we don't want analyze to waste time on pathfinding
          # b/c we do it separately in obs_pitch()
          interpolWin = 0,
          pathfinding = 'none',
          snakeStep = 0,
          snakePlot = FALSE,
          smooth = 0,
          plot = FALSE,
          returnPitchCands = TRUE
        )
        myPars$result = temp_anal$result
        myPars$summary = soundgen:::summarizeAnalyze(
          temp_anal$result,
          summaryFun = input$summaryFun
        )
        myPars$pitchCands = temp_anal$pitchCands
        myPars$spec_from_anal = temp_anal$spectrogram
        myPars$X = as.numeric(colnames(myPars$spec_from_anal))
        # add: update defaults that depend on samplingRate

        # if rerunning analyze() for the same audio, preserve the old manual values
        # (if any) and paste them back in
        isolate({
          if (!is.null(myPars$pitch) &
              nrow(myPars$manual) > 0) {
            # if the number of frames has changed (new windowLengh or step),
            # up/downsample manual pitch candidates accordingly
            len_old = length(myPars$pitch)  # !!! switch to myPars$manual
            len_new = ncol(myPars$pitchCands$freq)
            myPars$manual$frame = ceiling(myPars$manual$frame * len_new / len_old)
            # in case some manual frames merge into one, remove duplicates
            myPars$manual = myPars$manual[!duplicated(myPars$manual$frame), ]
          }
          obs_pitch()  # run pathfinder
          # if (length(myPars$pitch) != ncol(myPars$pitchCands$freq)) browser()
          # save the prior
          if (input$priorAdapt) {
            pitch_sem = HzToSemitones(myPars$pitch[!is.na(myPars$pitch)])
            if (length(pitch_sem) > 0) {
              myPars$priorMean = semitonesToHz(mean(pitch_sem))
              myPars$priorSD = semitonesToHz(sd(pitch_sem)) * 4
              myPars$pitchCert_mult = getPrior(
                priorMean = myPars$priorMean,
                priorSD = myPars$priorSD,
                pitchFloor = input$pitchFloor,
                pitchCeiling = input$pitchCeiling,
                pitchCands = myPars$pitchCands$freq,
                plot = FALSE
              )
            } else {
              myPars$priorMean = NA
            }
          } else {
            myPars$pitchCert_mult = getPrior(
              priorMean = input$priorMean,
              priorSD = input$priorSD,
              pitchFloor = input$pitchFloor,
              pitchCeiling = input$pitchCeiling,
              pitchCands = myPars$pitchCands$freq,
              plot = FALSE
            )
          }
        })
      })
    }
  })

  obs_pitch = function(updateAll = TRUE) {
    if (length(myPars$pitchCands$freq) > 0) {
      if (myPars$print) print('Looking for pitch contour with obs_pitch()')
      myPars$voicedSegments = soundgen:::findVoicedSegments(
        myPars$pitchCands$freq,
        manualV = myPars$manual$frame,
        manualTryToV = myPars$manualTryToV,
        manualUnv = myPars$manualUnv,
        shortestSyl = input$shortestSyl,
        shortestPause = input$shortestPause,
        minVoicedCands = input$minVoicedCands,
        pitchMethods = input$pitchMethods,
        step = input$step,
        samplingRate = myPars$samplingRate
      )
      if (updateAll | is.null(myPars$voicedSegments_old)) {
        # the first time we update everything
        sylToUpdate = myPars$voicedSegments
      } else {
        if (identical(myPars$voicedSegments_old, myPars$voicedSegments)) {
          # nothing changed in terms of syllable structure - only update the
          # currently edited syllable
          if (!is.null(myPars$closest_frame)) {
            syl_idx = which(myPars$closest_frame >= myPars$voicedSegments$segmentStart &
                              myPars$closest_frame <= myPars$voicedSegments$segmentEnd)
            sylToUpdate = myPars$voicedSegments[syl_idx, ]
          } else {
            sylToUpdate = data.frame()
          }
        } else {
          # some changes in terms of syllable structure - update the syllables
          # that changed
          a1 = myPars$voicedSegments_old  # may be empty
          a2 = myPars$voicedSegments  # may be empty
          if (nrow(a1) > 0) a1$included_a1 = TRUE
          if (nrow(a2) > 0) a2$included_a2 = TRUE
          res = merge(a1, a2, all = TRUE)
          sylToUpdate = na.omit(res[is.na(res$included_a1) & res$included_a2, 1:2])
        }
      }
      myPars$voicedSegments_old = myPars$voicedSegments
      # print(sylToUpdate)

      # for each syllable, impute NA's and find a nice path through pitch candidates
      if (is.null(myPars$pitch) || nrow(myPars$voicedSegments) == 0) {
        myPars$pitch = rep(NA, ncol(myPars$pitchCands$freq))
      } else if (nrow(myPars$voicedSegments) > 0) {
        myPars$voiced_frames = unlist(apply(myPars$voicedSegments, 1, function(x) x[1]:x[2]))
        myPars$unvoiced_frames = (1:ncol(myPars$pitchCands$freq)) [-myPars$voiced_frames]
        # make sure myPars$pitch is the same length as ncol(pitchCands$freq)
        if (length(myPars$pitch) != ncol(myPars$pitchCands$freq)) {
          myPars$pitch = .resample(
            list(sound = myPars$pitch),
            mult = ncol(myPars$pitchCands$freq) / length(myPars$pitch),
            lowPass = FALSE,
            plot = FALSE)
        }
        myPars$pitch[myPars$unvoiced_frames] = NA
      }
      if (nrow(sylToUpdate) > 0) {
        # if we have found at least one putatively voiced syllable
        for (syl in 1:nrow(sylToUpdate)) {
          myseq = sylToUpdate$segmentStart[syl]:sylToUpdate$segmentEnd[syl]
          manual_syl = myPars$manual[myPars$manual$frame %in% myseq, ]
          manual_syl$frame = manual_syl$frame - myseq[1] + 1  # adjust manual idx to syllable
          # compute the optimal path through pitch candidates
          myPars$pitch[myseq] = soundgen:::pathfinder(
            pitchCands = myPars$pitchCands$freq[, myseq, drop = FALSE],
            pitchCert = myPars$pitchCands$cert[, myseq, drop = FALSE],
            pitchSource = myPars$pitchCands$source[, myseq, drop = FALSE],
            step = input$step,
            manual = manual_syl,
            certWeight = input$certWeight,
            pathfinding = ifelse(input$pathfinding == 'slow',
                                 'fast',  # slow doesn't work well with manual cand-s
                                 input$pathfinding),
            interpolWin_bin = ceiling(input$interpolWin / input$step),
            interpolTol = input$interpolTol,
            interpolCert = input$interpolCert,
            snakeStep = 0,
            snakePlot = FALSE
          )
        }
      }

      ## Median smoothing of pitch contour
      if (input$smooth > 0) {
        points_per_sec = length(myPars$pitch) / myPars$dur * 1000
        # smooth of 1 means that smoothing window is ~100 ms
        myPars$smoothing_ww = round(input$smooth * points_per_sec / 10, 0)
        # the larger smooth, the heavier the smoothing (lower tolerance
        # threshold before values are replaced by median over smoothing window).
        # smooth of 1 gives smoothingThres of 4 semitones
        myPars$smoothingThres = 4 / input$smooth
        #print(myPars$pitchCands$source)
        myPars$pitch = soundgen:::medianSmoother(
          as.data.frame(myPars$pitch),
          smoothing_ww = myPars$smoothing_ww,
          smoothingThres = myPars$smoothingThres,
          inviolable = myPars$manual$frame
        )[, 1]
      }

      myPars$voiced_frames = which(!is.na(myPars$pitch))
      myPars$unvoiced_frames = (1:ncol(myPars$pitchCands$freq)) [-myPars$voiced_frames]
    }
  }
  observeEvent(
    # par-s that we don't need to use in analyze(), only in obs_pitch()
    # (b/c they do not affect the pitch candidates)
    c(input$shortestSyl, input$shortestPause,
      input$interpolWin, input$interpolTol, input$interpolCert,
      input$pathfinding, input$certWeight, input$smooth,
      input$priorMean, input$priorSD, input$priorAdapt),
    obs_pitch()
  )
  observeEvent(
    priority = 1, ignoreInit = TRUE,
    c(input$priorMean, input$priorSD, input$priorAdapt,
      input$pitchFloor, input$pitchCeiling), {
        if (!is.null(myPars$pitchCands$cert) & !input$priorAdapt) {
          if (myPars$print) print('Updating pitchCert with new prior')
          if (!is.null(myPars$pitchCert_mult)) {
            # undo the old prior, if any
            myPars$pitchCands$cert = myPars$pitchCands$cert / myPars$pitchCert_mult
          }
          # get a new prior
          myPars$pitchCert_mult = getPrior(
            priorMean = input$priorMean,
            priorSD = input$priorSD,
            pitchFloor = input$pitchFloor,
            pitchCeiling = input$pitchCeiling,
            pitchCands = myPars$pitchCands$freq,
            plot = FALSE
          )
          # update pitchCert
          myPars$pitchCands$cert = myPars$pitchCands$cert * myPars$pitchCert_mult
        }
      }
  )

  ## Clicking events
  observeEvent(input$spectrogram_click, {
    myPars$spectrogram_brush = NULL
    if (length(myPars$pitchCands$freq) > 0 & input$spectro_clickAct == 'addCand') {
      myPars$closest_frame = which.min(abs(
        as.numeric(colnames(myPars$pitchCands$freq)) - input$spectrogram_click$x))
      # create a manual pitch estimate for the closest frame with the clicked value
      new_freq = round(input$spectrogram_click$y * 1000, 3)
      # don't accept values beyond [pitchFloor, pitchCeiling]
      if (new_freq >= input$pitchFloor &
          new_freq <= input$pitchCeiling) {
        if (myPars$closest_frame %in% myPars$manual$frame) {
          myPars$manual$freq[myPars$manual$frame == myPars$closest_frame] = new_freq
        } else {
          myPars$manual = rbind(
            myPars$manual,
            data.frame(frame = myPars$closest_frame, freq = new_freq))
        }
        # just to keep things tidy
        myPars$manual = myPars$manual[order(myPars$manual$frame), ]
        # if this frame was manually flagged as unvoiced, remove this flag
        idx_rem = which(myPars$manualUnv == myPars$closest_frame)
        if (length(idx_rem) > 0) myPars$manualUnv = myPars$manualUnv[-idx_rem]
        if(input$automPathUpdate) {
          obs_pitch(updateAll = FALSE)
        } else {
          myPars$pitch[myPars$closest_frame] = new_freq
        }
      }
    } else if (input$spectro_clickAct == 'select') {
      myPars$cursor = input$spectrogram_click$x
    }
  })

  observeEvent(input$spectrogram_dblclick, {
    myPars$closest_frame = which.min(abs(as.numeric(colnames(myPars$pitchCands$freq)) -
                                           input$spectrogram_dblclick$x))
    if (length(myPars$closest_frame) > 0) {
      # remove manual anchor for this frame, if any
      idx_rem = which(myPars$manual$frame == myPars$closest_frame)
      if (length(idx_rem) > 0) myPars$manual = myPars$manual[-idx_rem, ]
      # mark the frame as unvoiced if it's not already marked as unvoiced
      if (!myPars$closest_frame %in% myPars$manualUnv)
        myPars$manualUnv = c(myPars$manualUnv, myPars$closest_frame)
      # re-run pitch contour
      if(input$automPathUpdate) {
        obs_pitch(updateAll = FALSE)
      } else {
        myPars$pitch[myPars$closest_frame] = NA
      }
    }
  })

  observeEvent(input$osc_click, {
    if (!is.null(myPars$pitchCands$freq)) {
      clicked_frame = which.min(abs(myPars$result$time - input$osc_click$x))
      if (clicked_frame %in% myPars$voiced_frames) {
        unvoiceSel(clicked_frame, bp = TRUE)
      } else {
        voiceSel(clicked_frame, bp = TRUE)
      }
    }
  })

  ## Buttons for operations with selection
  startPlay = function() {
    if (!is.null(myPars$myAudio)) {
      if (!is.null(myPars$spectrogram_brush) &&
          (myPars$spectrogram_brush$xmax - myPars$spectrogram_brush$xmin > 100)) {
        # at least 100 ms selected
        myPars$play$from = myPars$spectrogram_brush$xmin / 1000
        myPars$play$to = myPars$spectrogram_brush$xmax / 1000
      } else {
        myPars$play$from = myPars$cursor / 1000 # myPars$spec_xlim[1] / 1000
        myPars$play$to = myPars$spec_xlim[2] / 1000
      }
      myPars$play$dur = myPars$play$to - myPars$play$from
      myPars$play$timeOn = proc.time()
      myPars$play$timeOff = myPars$play$timeOn + myPars$play$dur
      myPars$cursor_temp = myPars$cursor
      myPars$play$on = TRUE
      if (myPars$print) print('Playing selection...')

      # play selection
      if (input$audioMethod == 'Browser') {
        # play with javascript
        shinyjs::js$playme_js(  # need an external js script for this
          audio_id = 'myAudio',  # defined in UI
          from = myPars$play$from,
          to = myPars$play$to)
      } else {
        # or play with R:
        playme(myPars$myAudio_path, from = myPars$play$from, to = myPars$play$to)
      }
    }
  }
  observeEvent(c(input$selection_play), startPlay())  # add , myPars$myAudio for autoplay

  stopPlay = function() {
    # browser()  # for debugging
    myPars$play$on = FALSE
    myPars$cursor = myPars$cursor_temp
    shinyjs::js$stopAudio_js(audio_id = 'myAudio')
  }
  observeEvent(input$selection_stop, stopPlay())

  observe({
    if (!is.null(myPars$play$on) && myPars$play$on) {
      time = proc.time()
      if (!is.null(myPars$slider_ms)) invalidateLater(myPars$slider_ms)
      if ((time - myPars$play$timeOff)[3] > 0) {
        myPars$play$on = FALSE
        myPars$cursor = myPars$cursor_temp  # reset to original cursor
      } else {
        myPars$cursor = as.numeric(
          myPars$play$from + time - myPars$play$timeOn
        )[3] * 1000  # [3] for "elapsed", ie "real" time
      }
    }
  })

  # HOTKEYS
  observeEvent(input$userPressedSmth, {
    button_key = substr(input$userPressedSmth, 1, nchar(input$userPressedSmth) - 8)
    # see https://keycode.info/
    if (button_key == ' ') {                  # SPACEBAR (play / stop)
      if (myPars$play$on) stopPlay() else startPlay()
    } else if (button_key == 'ArrowLeft') {    # ARROW LEFT (scroll left)
      shiftFrame('left', step = myPars$scrollFactor)
    } else if (button_key == 'ArrowRight') {    # ARROW RIGHT (scroll right)
      shiftFrame('right', step = myPars$scrollFactor)
    } else if (button_key == 'ArrowUp') {       # ARROW UP (horizontal zoom-in)
      changeZoom(myPars$zoomFactor)
    } else if (button_key %in% c('s', 'S')) {    # S (horizontal zoom to selection)
      zoomToSel()
    } else if (button_key == 'ArrowDown') {   # ARROW DOWN (horizontal zoom-out)
      changeZoom(1 / myPars$zoomFactor)
    } else if (button_key == '+') {     # + (vertical zoom-in)
      changeZoom_freq(1 / myPars$zoomFactor_freq)
    } else if (button_key == '-') {    # - (vertical zoom-out)
      changeZoom_freq(myPars$zoomFactor_freq)
    } else if (button_key == 'PageDown') {    # PageDown (next file)
      nextFile()
    } else if (button_key == 'PageUp') {  # PageUp (previous file)
      lastFile()
    } else if (button_key %in% c('u', 'U')) {  # U (unvoice)
      unvoiceSel()
    } else if (button_key %in% c('v', 'V')) {  # V (voice)
      voiceSel()
    } else if (button_key %in% c('r', 'R')) {  # R (raise by an octave)
      octaveUp()
    } else if (button_key %in% c('l', 'L')) {  # L (lower by an octave)
      octaveDown()
    } else if (button_key %in% c('p', 'P')) {  # P (set prior)
      setPrior()
    } else if (button_key %in% c('d', 'D')) {  # D (Draw pitch contour)
      obs_pitch()
    }
  })

  # UNVOICE
  observeEvent(input$selection_unvoice, {
    unvoiceSel()
  })
  unvoiceSel = function(frames = myPars$brush_sel_x, bp = myPars$bp) {
    if (myPars$print) print('Unvoicing selection...')
    if (!is.null(bp) & length(frames) > 0) {
      # NB: play forgets the previous selection, but other buttons do not,
      # hence myPars$bp instead of input$spectrogram_brush
      myPars$manualUnv = sort(unique(c(myPars$manualUnv, frames)))
      # (ie all points within selected time range, regardless of frequency - or
      # could use myPars$brush_sel_xy)
      # remove manual anchors within selection, if any
      idx_rem = which(myPars$manual$frame %in% myPars$manualUnv)
      if (length(idx_rem) > 0) myPars$manual = myPars$manual[-idx_rem, ]
      obs_pitch()
    }
  }

  # VOICE
  voiceSel = function(frames = myPars$brush_sel_x, bp = myPars$bp) {
    if (myPars$print) print('Voicing selection...')
    if (!is.null(bp) & length(frames) > 0) {
      # manually voice the selected frames
      myPars$manualTryToV = c(myPars$manualTryToV, frames)
      # remove them from the list of manually unvoiced frames
      if (length(myPars$manualUnv) > 0) {
        idx_rem = which(myPars$manualUnv %in% frames)
        if (length(idx_rem) > 0) {
          myPars$manualUnv = myPars$manualUnv[-idx_rem]
        }
      }
      obs_pitch()
    }
  }
  observeEvent(input$selection_voice, {
    voiceSel()
  })

  # SHIFT BY AN OCTAVE
  octaveUp = function() {
    if (myPars$print) print('Selection octave up...')
    if (!is.null(myPars$bp) & length(myPars$brush_sel_xy) > 0) {
      # remove previous manual cands in selection
      idx_rem = which(myPars$manual$frame %in% myPars$brush_sel_xy)
      if (length(idx_rem) > 0) myPars$manual = myPars$manual[-idx_rem, ]
      # add the new ones
      myPars$manual = rbind(myPars$manual, data.frame(
        frame = myPars$brush_sel_xy,
        freq = myPars$pitch[myPars$brush_sel_xy] * 2
      ))
      # make sure we stay within pitchFloor/pitchCeiling
      myPars$manual$freq[myPars$manual$freq < input$pitchFloor] = input$pitchFloor
      myPars$manual$freq[myPars$manual$freq > input$pitchCeiling] = input$pitchCeiling
      if (input$automPathUpdate) {
        obs_pitch()
      } else {
        myPars$pitch[myPars$brush_sel_xy] = myPars$pitch[myPars$brush_sel_xy] * 2
      }
    }
  }
  observeEvent(input$selection_octaveUp, {
    octaveUp()
  })

  octaveDown = function() {
    if (myPars$print) print('Selection octave down...')
    if (!is.null(myPars$bp) & length(myPars$brush_sel_xy) > 0) {
      # remove previous manual cands in selection
      idx_rem = which(myPars$manual$frame %in% myPars$brush_sel_xy)
      if (length(idx_rem) > 0) myPars$manual = myPars$manual[-idx_rem, ]
      # add the new ones
      myPars$manual = rbind(myPars$manual, data.frame(
        frame = myPars$brush_sel_xy,
        freq = myPars$pitch[myPars$brush_sel_xy] / 2
      ))
      # make sure we stay within pitchFloor/pitchCeiling
      myPars$manual$freq[myPars$manual$freq < input$pitchFloor] = input$pitchFloor
      myPars$manual$freq[myPars$manual$freq > input$pitchCeiling] = input$pitchCeiling
      if (input$automPathUpdate) {
        obs_pitch()
      } else {
        myPars$pitch[myPars$brush_sel_xy] = myPars$pitch[myPars$brush_sel_xy] / 2
      }
    }
  }
  observeEvent(input$selection_octaveDown, {
    octaveDown()
  })

  # PRIOR
  observe({
    if (input$priorAdapt) {
      pitch_sem = HzToSemitones(myPars$pitch[!is.na(myPars$pitch)])
      if (length(pitch_sem) > 0) {
        myPars$priorMean = semitonesToHz(mean(pitch_sem))
        myPars$priorSD = semitonesToHz(sd(pitch_sem)) * 4
        myPars$prior = getPrior(
          priorMean = myPars$priorMean,
          priorSD = myPars$priorSD,
          pitchFloor = input$pitchFloor,
          pitchCeiling = input$pitchCeiling,
          len = 100,
          plot = FALSE
        )
      } else {
        myPars$priorMean = NA
        myPars$prior = NA
      }
    } else {
      myPars$prior = getPrior(
        priorMean = input$priorMean,
        priorSD = input$priorSD,
        pitchFloor = input$pitchFloor,
        pitchCeiling = input$pitchCeiling,
        len = 100,
        plot = FALSE
      )
    }
  })
  setPrior = function() {
    if (myPars$print) print('Setting prior...')
    if (!is.null(input$spectrogram_brush)) {
      pr = c(input$spectrogram_brush$ymin, input$spectrogram_brush$ymax) * 1000
      # don't push down below pitchFloor
      pr[pr < input$pitchFloor] = input$pitchFloor
      # but update pitchCeiling if prior is higher
      if (any(pr > input$pitchCeiling)) {
        updateSliderInput(session, 'pitchCeiling', value = round(max(pr)))
      }
      meanPr = round(mean(pr))
      sdPr = round((HzToSemitones(pr[2]) - HzToSemitones(mean(pr))) / 2, 1)
      updateSliderInput(session, 'priorMean', value = meanPr)
      updateSliderInput(session, 'priorSD', value = sdPr)
    }
  }
  observeEvent(input$selection_setPrior, {
    setPrior()
  })

  observeEvent(input$button_pathUpdate, {
    obs_pitch()
  })

  # HOVER
  observeEvent(input$spectrogram_hover, {
    if (!is.null(input$spectrogram_hover) & !is.null(myPars$spec)) {
      myPars$spectrogram_hover = input$spectrogram_hover
      cursor_hz = myPars$spectrogram_hover$y * 1000
      cursor_notes = soundgen::HzToNotes(cursor_hz)
      myPars$spectrogram_hover$freq = paste0(
        round(myPars$spectrogram_hover$y * 1000), ' Hz (',
        cursor_notes, ')')
      myPars$spectrogram_hover$time = paste0(
        round(myPars$spectrogram_hover$x), ' ms'
      )
    }
  })
  shinyjs::onevent('dblclick', id = 'specOver', {
    # NB: more flexible and less mafan than juggling with the observer of
    # input$spectrogram_hover
    myPars$spectrogram_hover = NULL
    shinyjs::js$clearBrush(s = '_brush')
  } )

  # BRUSH
  brush = observeEvent(input$spectrogram_brush, {
    if (myPars$print) print('Running brush()...')
    if (!is.null(input$spectrogram_brush)) {
      myPars$spectrogram_brush = input$spectrogram_brush
    }
    myPars$pitch_df = data.frame(
      time = as.numeric(colnames(myPars$pitchCands$freq)),
      freq = myPars$pitch / 1000
    )
    myPars$bp = brushedPoints(myPars$pitch_df,
                              brush = input$spectrogram_brush,
                              xvar = 'time', yvar = 'freq',
                              allRows = TRUE)
    # selected pitch points
    myPars$brush_sel_xy = which(myPars$bp[, 'selected_'] == TRUE)
    # selected frames (along x axis)
    myPars$brush_sel_x = which(myPars$pitch_df$time > input$spectrogram_brush$xmin &
                                 myPars$pitch_df$time < input$spectrogram_brush$xmax)
  })

  # ZOOM
  changeZoom_freq = function(coef) {
    # midpoint = mean(input$spec_ylim)
    # halfRan = diff(input$spec_ylim) / 2 / coef
    # newLow = max(0, midpoint - halfRan)
    # newHigh = min(myPars$samplingRate / 2, midpoint + halfRan)
    updateSliderInput(session, 'spec_ylim', value = c(0, input$spec_ylim[2] * coef))
  }
  observeEvent(input$zoomIn_freq, changeZoom_freq(1 / myPars$zoomFactor_freq))
  observeEvent(input$zoomOut_freq, changeZoom_freq(myPars$zoomFactor_freq))

  changeZoom = function(coef, toCursor = FALSE) {
    if (myPars$print) print('Zooming')
    # intelligent zoom-in a la Audacity: midpoint moves closer to seletion/cursor
    if (!is.null(myPars$cursor) & toCursor) {
      if (!is.null(myPars$spectrogram_brush)) {
        midpoint = 3/4 * mean(c(myPars$spectrogram_brush$xmin,
                                myPars$spectrogram_brush$xmax)) +
          1/4 * mean(myPars$spec_xlim)
      } else {
        if (myPars$cursor > 0) {
          midpoint = 3/4 * myPars$cursor + 1/4 * mean(myPars$spec_xlim)
        } else {
          # when first opening a file, zoom in to the beginning
          midpoint = mean(myPars$spec_xlim) / coef
        }
      }
    } else {
      midpoint = mean(myPars$spec_xlim)
    }
    halfRan = diff(myPars$spec_xlim) / 2 / coef
    newLeft = max(0, midpoint - halfRan)
    newRight = min(myPars$dur, midpoint + halfRan)
    myPars$spec_xlim = c(newLeft, newRight)
    # use user-set time zoom in the next audio
    if (!is.null(myPars$spec_xlim)) myPars$initDur = diff(myPars$spec_xlim)
  }
  observeEvent(input$zoomIn, changeZoom(myPars$zoomFactor, toCursor = TRUE))
  observeEvent(input$zoomOut, changeZoom(1 / myPars$zoomFactor))
  zoomToSel = function() {
    if (!is.null(myPars$spectrogram_brush)) {
      myPars$spec_xlim = round(c(myPars$spectrogram_brush$xmin,
                                 myPars$spectrogram_brush$xmax))
    }
  }
  observeEvent(input$zoomToSel, {
    zoomToSel()
  })

  shiftFrame = function(direction, step = 1) {
    ran = diff(myPars$spec_xlim)
    shift = ran * step
    if (direction == 'left') {
      newLeft = max(0, myPars$spec_xlim[1] - shift)
      newRight = newLeft + ran
    } else if (direction == 'right') {
      newRight = min(myPars$dur, myPars$spec_xlim[2] + shift)
      newLeft = newRight - ran
    }
    myPars$spec_xlim = c(newLeft, newRight)
    # update cursor when shifting frame, but not when zooming
    myPars$cursor = myPars$spec_xlim[1]
  }
  observeEvent(input$scrollLeft, shiftFrame('left', step = myPars$scrollFactor))
  observeEvent(input$scrollRight, shiftFrame('right', step = myPars$scrollFactor))

  moveSlider = observe({
    if (myPars$print) print('Moving slider')
    width = round(diff(myPars$spec_xlim) / myPars$dur * 100, 2)
    left = round(myPars$spec_xlim[1] / myPars$dur * 100, 2)
    shinyjs::js$scrollBar(  # need an external js script for this
      id = 'scrollBar',  # defined in UI
      width = paste0(width, '%'),
      left = paste0(left, '%')
    )
    myPars$cursor = myPars$spec_xlim[1]
  })

  observeEvent(input$scrollBarLeft, {
    if (!is.null(myPars$spec)) {
      spec_span = diff(myPars$spec_xlim)
      scrollBarLeft_ms = input$scrollBarLeft * myPars$dur
      myPars$spec_xlim = c(max(0, scrollBarLeft_ms),
                           min(myPars$dur, scrollBarLeft_ms + spec_span))
    }
  }, ignoreInit = TRUE)

  observeEvent(input$scrollBarMove, {
    direction = substr(input$scrollBarMove, 1, 1)
    if (direction == 'l') {
      shiftFrame('left', step = myPars$scrollFactor)
    } else if (direction == 'r') {
      shiftFrame('right', step = myPars$scrollFactor)
    }
  }, ignoreNULL = TRUE)

  observeEvent(input$scrollBarWheel, {
    direction = substr(input$scrollBarWheel, 1, 1)
    if (direction == 'l') {
      shiftFrame('left', step = myPars$wheelScrollFactor)
    } else if (direction == 'r') {
      shiftFrame('right', step = myPars$wheelScrollFactor)
    }
  }, ignoreNULL = TRUE)

  observeEvent(input$zoomWheel, {
    direction = substr(input$zoomWheel, 1, 1)
    if (direction == 'l') {
      changeZoom(1 / myPars$zoomFactor)
    } else if (direction == 'r') {
      changeZoom(myPars$zoomFactor, toCursor = TRUE)
    }
  }, ignoreNULL = TRUE)

  # # step-overlap
  # observeEvent(input$overlap, {
  #   # change step if overlap changes, but don't change step if windowLength changes
  #   step = round(input$windowLength * (1 - input$overlap / 100))
  #   if (input$step != step)
  #     updateNumericInput(session, 'step', value = step)
  # }, ignoreInit = TRUE)
  # observeEvent(c(input$step, input$windowLength), {
  #   # change overlap if step or windowLength change
  #   overlap = (1 - input$step / input$windowLength) * 100
  #   if (input$overlap != overlap)
  #     updateSliderInput(session, 'overlap', value = overlap)
  # })


  # SAVE OUTPUT
  done = function() {
    # meaning we have finished editing pitch contour for a sound - prepares
    # the output
    if (myPars$print) print('Running done()...')
    if (!is.null(myPars$myAudio_path) &&
        !is.null(myPars$result) && !
        is.null(myPars$pitch)) {
      new = data.frame(
        file = basename(myPars$myAudio_filename),
        time = paste(round(myPars$X), collapse = ', '),
        pitch = paste(round(myPars$pitch), collapse = ', '),
        stringsAsFactors = FALSE
      )
      result_new = soundgen:::updateAnalyze(
        result = myPars$result,
        pitch_true = myPars$pitch,
        spectrogram = myPars$spec_from_anal,
        windowLength = input$windowLength,
        harmHeight_pars = list(
          harmThres = defaults_analyze['harmThres', 'default'],
          harmTol = defaults_analyze['harmTol', 'default'],
          harmPerSel = defaults_analyze['harmPerSel', 'default']),
        flux_pars = list(thres = 0.15),
        smooth = input$smooth,
        smoothing_ww = myPars$smoothing_ww,
        smoothingThres = myPars$smoothingThres
      )
      summary_new = soundgen:::summarizeAnalyze(
        result_new,
        summaryFun = isolate(myPars$summaryFun))
      new = cbind(new$file,
                  summary_new,
                  new[, c('time', 'pitch')])
      colnames(new)[1] = 'file'  # otherwise misnamed
      new$file = as.character(new$file)  # otherwise becomes a factor
      if (is.null(myPars$out)) {
        myPars$out = new
      } else {
        idx = which(myPars$out$file == new$file)
        myPars$out = soundgen:::rbind_fill(myPars$out, new)
        if (length(idx) == 1) myPars$out = myPars$out[-idx, ]
        myPars$out = myPars$out[order(myPars$out$file), ]
      }
    }
    cols_order = c(colnames(myPars$out)[!colnames(myPars$out) %in% c('time', 'pitch')],
                   'time', 'pitch')
    myPars$out = myPars$out[, cols_order]

    # add manual corrections to the history list
    if (!is.null(myPars$myAudio_filename)) {
      myPars$history[[myPars$myAudio_filename]]$manual = myPars$manual
      myPars$history[[myPars$myAudio_filename]]$manualUnv = myPars$manualUnv
    }

    # save a backup
    if (!is.null(myPars$out)) {
      write.csv(myPars$out, 'www/temp.csv', row.names = FALSE)
      myPars$result$pitch = myPars$pitch
      myPars$history[[myPars$myAudio_filename]]$detailed = myPars$result
      myPars$history[[myPars$myAudio_filename]]$summary = myPars$out
      myPars$history[[myPars$myAudio_filename]]$pitch = myPars$pitch
      my_pitch <<- myPars$history
    }
  }

  observeEvent(input$fileList, {
    done()
    myPars$n = which(myPars$fileList$name == input$fileList)
    reset()
    readAudio(myPars$n)
  }, ignoreInit = TRUE)

  nextFile = function() {
    if (!is.null(myPars$myAudio_path)) {
      # done()
      if (myPars$n < myPars$nFiles) {
        myPars$n = myPars$n + 1
        updateSelectInput(session, 'fileList',
                          selected = myPars$fileList$name[myPars$n])
        # ...which triggers observeEvent(input$fileList)
      }
    }

  }
  observeEvent(input$nextFile, nextFile())

  lastFile = function() {
    if (!is.null(myPars$myAudio_path)) {
      # done()
      if (myPars$n > 1) {
        myPars$n = myPars$n - 1
        updateSelectInput(session, 'fileList',
                          selected = myPars$fileList$name[myPars$n])
      }
    }
  }
  observeEvent(input$lastFile, lastFile())

  output$saveRes = downloadHandler(
    filename = function() 'output.csv',
    content = function(filename) {
      done()  # finalize the last file
      write.csv(myPars$out, filename, row.names = FALSE)
      my_pitch <<- myPars$history
      if (file.exists('www/temp.csv')) file.remove('www/temp.csv')
      # offer to close the app
      showModal(modalDialog(
        title = "Terminate the app?",
        easyClose = TRUE,
        footer = tagList(
          actionButton("terminate_no", "Keep working"),
          actionButton("terminate_yes", "Terminate")
        )
      ))
    }
  )
  observeEvent(input$terminate_no, {
    removeModal()
  })
  observeEvent(input$terminate_yes, {
    stopApp(returnValue = myPars$history)
  })

  observeEvent(input$about, {
    if (myPars$debugQn) {
      browser()  # back door for debugging)
    } else {
      showNotification(
        ui = paste0("Manual pitch editor: soundgen ", packageVersion('soundgen'), ". Left-click to add/correct a pitch anchor, double-click to remove/unvoice the frame. More info: ?pitch_app and http://cogsci.se/soundgen.html"),
        duration = 10,
        closeButton = TRUE,
        type = 'default'
      )
    }
  })

  ## TOOLTIPS - have to be here instead of UI b/c otherwise problems with regulating delay
  # (see https://stackoverflow.com/questions/47477237/delaying-and-expiring-a-shinybsbstooltip)
  # STFT
  shinyBS::addTooltip(session, id='reset_to_def', title = 'Reset all settings to default values', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='audioMethod', title = "Play audio with javascript (recommended in Firefox, doesn't work in Chrome) or with R (browser-independent, but then the cursor doesn't move, and you can't stop playback)", placement="right", trigger="hover", options = list(delay = list(show = 1000, hide = 0)))
  shinyBS::addTooltip(session, id='windowLength', title = 'Length of STFT window, ms. Larger values improve frequency resolution at the expense of time resolution', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='step', title = 'Step between analysis frames, ms (alternative to "overlap")', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specType', title = 'Spectrogram type, argument "specType" in spectrogram(). Affects pitch tracking', placement="right", trigger="hover", options = list(delay = list(show = 1000, hide = 0)))
  # shinyBS::addTooltip(session, id='overlap', title = 'Overlap between analysis frames, % (alternative to "step")', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='dynamicRange', title = 'Dynamic range of spectrogram, dB', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='zp', title = 'Zero padding of STFT window (improves frequency resolution): 8 means 2^8 = 256, etc.', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='wn', title = 'Type of STFT window', placement="right", trigger="hover", options = tooltip_options)

  # voicing
  shinyBS::addTooltip(session, id='silence', title = 'Frames with RMS below silence threshold are not analyzed', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='entropyThres', title = 'Frames with Weiner entropy above entropy threshold are ignored when searching for pitch candidates', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='nCands', title = 'Maximum number of pitch candidates to use per method', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='minVoicedCands', title = 'Minimum number of pitch candidates that have to be defined to consider a frame voiced', placement="right", trigger="hover", options = tooltip_options)

  # priors
  shinyBS::addTooltip(session, id='pitchFloor', title = 'No candidates below this absolute threshold', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='pitchCeiling', title = 'No candidates above this absolute threshold', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='priorMean', title = 'Candidates close to this value are prioritized (how close is determined by priorSD)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='priorAdapt', title = 'Adds a second pass for finding the optimal pitch contour, with prior determined by the initial pitch estimates', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='priorSD', title = 'Determines the width of expected pitch range (standard deviation of gamma distribution around priorMean)', placement="right", trigger="hover", options = tooltip_options)

  # trackers
  shinyBS::addTooltip(session, id='domThres', title = 'Dominant frequency is defined as the lowest bin in a spectrum smoothed and normalized to range from 0 to 1 that it at least "domThres" high (1 = absolute maximum, ie peak frequency)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='domSmooth', title = 'Width of smoothing interval for finding the lowest dominant frequency band (low values = no smoothing)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='autocorThres', title = 'Voicing threshold for autocorrelation algorithm', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='autocorSmooth', title = 'Width of smoothing interval (in bins) for finding peaks in the autocorrelation function', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='autocorUpsample', title = 'Upsamples acf to this resolution (Hz) to improve accuracy in high frequencies', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='autocorBestPeak', title = 'Amplitude of the lowest best candidate relative to the absolute max of the acf', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='cepThres', title = 'Voicing threshold for cepstral algorithm', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='cepZp', title = 'Length of cepstral window after zero padding: 8 means 2^8 = 256, etc.', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specType', title = '"commonFactor" = greatest common factor of putative harmonics, "BaNa" = ratio of putative harmonics', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specThres', title = 'Voicing threshold for Ba-Na algorithm', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specPeak', title = 'Minimum amplitude of harmonics considered pitch candidates', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specHNRslope', title = '0 = same threshold regardless of HNR; positive = lower threshold in noisy sounds', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specSmooth', title = 'Width of window for detecting harmonics in the spectrum, Hz', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specMerge', title = 'Pitch candidates within specMerge semitones are merged with boosted certainty', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specSinglePeakCert', title = 'If pitch is calculated based on a single harmonic ratio (as opposed to several ratios converging on the same candidate), its certainty is taken to be specSinglePeakCert', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specRatios', title = 'For method = "commonFactor", the number of integer fractions to consider', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='hpsNum', title = 'How many times to downsample and then multiply the spectra', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='hpsThres', title = 'How high a spectral peak has to be to be considered a pitch candidate, ~0 to 1', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='hpsNorm', title = 'Rather arbitrary normalization of certainty in hps candidates intended to make them more comparable to other pitch tracking methods (0 = no boost in certainty, 2 = default quadratic)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='hpsPenalty', title = 'HPS performs worse at low frequencies (relative to windowLength), so low-frequency pitch candidates are penalized (0 = no penalization, ~10-20 = a lot)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='zcThres', title = 'Minimum certainty in zero-crossing pitch candidates, ~0 to 1', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='zcWin', title = 'Confidence in zero-crossing candidates depends on how stable the estimates are across zcWin glottal cycles', placement="right", trigger="hover", options = tooltip_options)

  # pathfinder
  shinyBS::addTooltip(session, id='summaryFun', title = "The function(s) used to summarize output", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='summaryFun_text', title = "If specified, overrides the options above. For short column names, define and name your function in R prior to starting pitch_app", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='automPathUpdate', title = "Update the optimal pitch contour automatically every time an anchor changes? Turn off to avoid delays when editing a long audio", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='pathfinding', title = "Method of finding the optimal path through pitch candidates: 'none' = best candidate per frame, 'fast' = simple heuristic, 'slow' = annealing (initial analysis only)", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='certWeight', title = 'Specifies how much we prioritize the certainty of pitch candidates vs. pitch jumps', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='shortestSyl', title = 'Shorter voiced segments (ms) will be treated as voiceless or merged with longer segments', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='shortestPause', title = "The smallest gap between voiced syllables (ms) that means they shouldn't be merged", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='smooth', title = 'Amount of median smoothing', placement="right", trigger="hover", options = tooltip_options)

  # smoothing
  shinyBS::addTooltip(session, id='interpolWin', title = "If no pitch candidates are found within ±interpolTol of the median 'best guess' over ±interpolWin, this median is added as an interpolated candidate", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='interpolTol', title = "Tolerated deviance from 'best guess' before adding an interpolated candidate: proportion of best guess frequency", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='interpolCert', title = "Certainty assigned to interpolated pitch candidates", placement="right", trigger="hover", options = tooltip_options)

  # spectrogram
  shinyBS::addTooltip(session, id='spec_ylim', title = "Range of displayed frequencies, kHz", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='spec_maxPoints', title = 'The number of points to plot in the spectrogram (smaller = faster, but low resolution)', placement="below", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='spec_cex', title = "Magnification coefficient controlling the size of points showing pitch candidates", placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specContrast', title = 'Regulates the contrast of the spectrogram', placement="below", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='specBrightness', title = 'Regulates the brightness of the spectrogram', placement="below", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='blur_freq', title = 'Gaussian filter of frequency: >0 = blur, <0 = unblur (sharpen)', placement="below", trigger="hover", options = list(delay = list(show = 1000, hide = 0)))
  shinyBS::addTooltip(session, id='blur_time', title = 'Gaussian filter of time: >0 = blur, <0 = unblur (sharpen)', placement="below", trigger="hover", options = list(delay = list(show = 1000, hide = 0)))

  # oscillogram
  shinyBS::addTooltip(session, id='osc', title = 'The type of oscillogram to show', placement="below", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='osc_height', title = 'The height of oscillogram, pixels', placement="below", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='osc_maxPoints', title = 'The number of points to plot in the oscillogram (smaller = faster, but low resolution)', placement="below", trigger="hover", options = tooltip_options)

  # action buttons
  shinyBS:::addTooltip(session, id='lastFile', title='Save and return to the previous file (PageUp)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS:::addTooltip(session, id='nextFile', title='Save and proceed to the next file (PageDown)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS:::addTooltip(session, id='selection_play', title='Play selection (SPACEBAR)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='selection_unvoice', title = 'Unvoice selection (U)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='selection_voice', title = 'Voice selection (V) - obviouslly, pitch estimates may be totally incorrect', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='selection_octaveUp', title = 'Raise pitch for selection by an octave (R)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='selection_octaveDown', title = 'Lower pitch for selection by an octave (L)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='selection_setPrior', title = 'Set a prior on expected pitch values corresponding to the selected frequency range (P)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='button_pathUpdate', title = 'Draw / refresh pitch contour (D) (only needed if "Out/Path/Update path automatically" is turned off)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='saveRes', title = 'Download results (see ?pitch_app for recovering unsaved data after a crash)', placement="right", trigger="hover", options = tooltip_options)

  # navigation / zoom
  shinyBS::addTooltip(session, id='zoomIn_freq', title = 'Zoom in frequency (+)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='zoomOut_freq', title = 'Zoom out frequency (-)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='scrollLeft', title = 'Scroll left (arrow LEFT)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='zoomOut', title = 'Zoom out time (arrow DOWN)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='zoomToSel', title = 'Zoom to selection (S)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='zoomIn', title = 'Zoom in time (arrow UP)', placement="right", trigger="hover", options = tooltip_options)
  shinyBS::addTooltip(session, id='scrollRight', title = 'Scroll right (arrow RIGHT)', placement="right", trigger="hover", options = tooltip_options)
}