Skip to contents

metRscreen 0.1.1 (development version)

Version 0.1.1, on the development branch. The stable version on master is 0.0.1.0000. Check which one you have with packageVersion("metRscreen").

New features

  • Collaborative screening. collab.names now switches on a full collaborative mode (on master it only recorded a screener’s name against each decision in one shared file):
    • each screener gets their own files (<screen.file>_<name>_Screened.csv and <screen.file>_<name>_history.rds), and you switch screener in the app;
    • other screeners’ decisions are hidden by default, so screening stays independent, and can be shown with a switch;
    • a combined <screen.file>_Collab_Summary.csv flags agreements and conflicts;
    • screener names and reject reasons are remembered for the project, so later sessions can leave collab.names empty or add new screeners;
    • decisions from an earlier single-screener session can be claimed by a screener.
  • Double screening with the new collab.split = 2 argument: every paper is screened by exactly two screeners, spread evenly and at random across the team. The split is saved and reused, so it never reshuffles part-way through. Each screener sees only their own papers, and can browse all papers in a view-only mode.
  • The left and right arrow keys move between papers.
  • A shortcuts bar above the abstract lists the keyboard shortcuts.
  • New ? wildcard in keywords, matching exactly one letter or number (wom?n matches woman and women). It replaces the accidental regex meaning ? had before, where colou?r meant an optional “u”.
  • Hovering over a keyword box shows how to separate words and use the * and ? wildcards.
  • A light/dark mode switch next to the Help button. The app starts in the same mode as your computer, and decision and keyword colours are lightened in dark mode so they stay readable.
  • Each keyword box now also has its own underline (solid, double, dotted, dashed, wavy), so matches can be told apart without relying on colour. A sample next to each box shows its style, and a switch turns the underlines off for colour only.
  • Font size, light/dark mode and the underline switch are remembered between sessions on the same computer (saved in metRscreen’s folder under tools::R_user_dir("metRscreen", "config")).
  • The list of screened papers moves below the abstract when there isn’t room for both side by side.

Bug fixes

  • A relative path for screen.file (e.g. "data/refs.csv") did not work, because the app runs with its own folder as the working directory. Relative paths, ~ and here::here() paths now all work.
  • Calling metRscreen() without screen.file printed an error but carried on. It now stops.
  • Screening history was loaded from any .rds file in the same folder, so a different project’s history could be picked up. Only this file’s own _history.rds is now used.
  • The app failed to start on R versions before 4.4.0, because it used %||%, which base R only has from 4.4.0.
  • .ris import:
    • abstracts wrapped over several lines kept only their first line;
    • the A1 (author) and N2 (abstract) tags were ignored;
    • empty tags such as N1 - were not recognised in some exports;
    • dates such as 2020/// or 2020/05/01 are now reduced to the year;
    • a .ris that had already been converted is no longer converted again over the existing .csv, so papers can’t change under a session that has started (delete the .csv to import the .ris again).
  • Keyboard shortcuts (y, n, m) fired with Ctrl/Cmd held down, clashing with browser and system shortcuts; did not work with Caps Lock on; and stopped working after ticking a reject reason. They now also respect greyed-out buttons.
  • In the list of screened papers, a reference button could jump to the wrong paper whenever papers earlier in the file had not been screened yet.
  • The reference buttons were re-wired after every decision, so one click could run several times as a session went on.
  • If the reference file no longer matched a saved single-screener session (e.g. a paper had been added), the app warned you to revert but then overwrote _Screened.csv and _history.rds with nothing. These files are now left untouched.
  • Keyword highlighting:
    • * now matches zero or more letters or numbers, so parent* also highlights “parent”;
    • a word straight after a highlighted word was skipped (e.g. only the first and third of “parents parents parents”);
    • keywords such as “span”, “style” or “the” could match inside the HTML added for other keywords and break the display;
    • keywords containing characters such as +, ( or . caused an error or matched the wrong thing; they are now matched literally;
    • terms separated by a comma without a space were not split, although the help said to use no space; spaces are now optional;
    • accented letters (e.g. café) did not match wildcards.
  • Decision, progress and keyword colours were hard to read on a white background (e.g. the green “Accept” and orange keywords); they are now darker in light mode so all text meets the 4.5:1 contrast guideline.
  • The keyword tip boxes are now solid and high-contrast in both modes, with the examples shown as code.
  • Going back to a screened paper showed an empty comment box and no ticked reasons, but a new decision silently kept the old comment. The box and reasons now show what was saved for that paper, and emptying the box removes the comment.
  • Holding down y, n or m repeated the key, recording the same decision on each paper that came up. One press is now one decision.

Other changes

  • Documentation for screen.file, collab.names and the new collab.split argument expanded, and the README now covers collaborative screening, the files it produces and working as a team.