Version 0.4.1 of revtools is available on GitHub now, and should be on CRAN soon. This version fixes a few bugs in 0.4.0, but there are also a large number of improvements to parts of the package. The details are as follows:

New functions and features

  • Improved behaviour of screen_topics:
    • Clicking on topics highlights all points in that topic, and renders points from non-selected topics as grey with low opacity
    • New 'processing' tab allows manipulation of arguments passed to make_dtm
  • More consistent apps:
    • Apps that record screening information now do so in a unique column starting with 'selected_', allowing users to determine at which stage articles were excluded during the screening process. New column names are as follows:
      • screen_titles stores data in 'selected_titles'
      • screen_abstracts stores data in 'selected_abstracts'
      • screen_topics stores data in 'selected_topics'
    • All apps now use the same naming convention for selection/exclusion (previously screen_topics returned TRUE/FALSE)
    • screen_abstracts and screen_titles no longer add 'color' or 'order' columns to exported objects
  • Improvements to make_dtm:
    • Now accepts objects of class data.frame as an input
    • Flexible bigram detection c/o package 'ngram'
    • Chooses words for display by number of characters, rather than frequency of occurrence in the dataset
    • Allows user-controlled removal (the default) or retention of empty rows
  • screen_abstracts automatically moves to the next reference when a selection is made, and hides screened articles by default
  • read_bibliography now recognises ris tags from Web of Science (.ciw format)
  • add_line_breaks now available as a standalone function, as well as being an argument in format_citation
  • objects returned by find_duplicates now include information on the call that created them (via attributes)

Changed defaults

  • find_duplicates now searches for exact matches of DOIs by default, or fuzzy matching of article titles if DOIs are missing
  • make_dtm now returns an object of slam::simple_triplet_matrix rather than matrix. This can be coerced to a matrix by as.matrix

Bug fixes

  • merge_columns should return columns in the same order regardless of order that objects are provided
  • Importing multiple files at once no longer generates non-unique 'label' entries, which could cause failure of article citations in screen_topics.
  • csv files now import with correct text encoding using read_bibliography
  • screen_duplicates loads data.frames correctly from the command line
  • data.frames returned by read_bibliography always have snake case column names
  • make_dtm now removes all words of three letters or fewer
  • Improved behaviour of format_citation

In other news, this is likely to be the last version of revtools to contain it’s own import code, which is being outsourced to package ‘synthesisr’ as part of the metaverse project. The function read_bibliography will be retained, but as a wrapper to code from synthesisr. More to follow in 2020.