Collapse numeric in-text citations produced by bookdown, retaining links to chapter bibliography while ensuring all refs within collapsed ranges are included in the bibliography.

collapse_numeric.R

collapse_numeric <- function(index) {
  # index: path to home page of bookdown-generated HTML book (e.g. index.html)
  doc <- readLines(index)
  pages <- file.path(
    dirname(index), sub('.*data-path="(.*?)".*', '\\1', 
                        grep('data-level="\\d+"', doc, val=T))
    )
  
  lapply(pages, function(p) {
    x <- rmarkdown:::read_utf8(p)
    idx <- grep('doc-biblioref', x)
    if(length(idx) > 0) {
      for(idx_ in idx) { # lines (paras) containing cites
        m <- regmatches(x[idx_], gregexpr('<span class="citation">.*?</span>', 
                                          x[idx_], perl=TRUE))[[1]]
        # ^ find in-text citations
        m_split <- strsplit(m, ',')
        
        m_split_new <- lapply(m_split, function(x) {
          j <- as.numeric(gsub('.*>(\\d+)</a>.*', '\\1', x))
          # extract numeric component
          n <- length(j)
          if(n > 1) {
            starts <- c(1, which(diff(j) > 1) + 1)
            ends <- c(which(diff(j) > 1), n)
            paste(mapply(function(a, b) {
              if(b-a > 1) { 
                paste(x[unique(c(a, b))], collapse='-')  
              } else {
                paste(x[unique(a:b)], collapse=',') 
                # ^ don't collapse if only a single number or just two consecutive numbers
              }
            }, starts, ends), collapse=',') 
            # ^ collapse
          } else {
            x
          }
        })
        
        for(i in seq_along(m_split)) {
          x <- sub(m[i], m_split_new[[i]], x)  
        }
        rmarkdown:::write_utf8(x, p)  
      }
    }
  })
}