Changes in version 0.3.5                        

  - QC report: self-contained HTML per participant (or a zip of all)
    with summary, exclusions, the signal at each step, every trial and
    parameters.

  - BIDS-like export: derivatives/shinygazer/sub-*/eye/ with processed
    time series, events, blinks, binned data, trial quality and JSON
    metadata.

  - Study file: save all settings to .json and restore them later.

  - Optional resampling onto a uniform time grid at each participant's
    own rate (dropped samples inserted as missing).

  - Baseline value: choose the median (default) or mean of the baseline
    period (step 4). The methods paragraph names the statistic and, for
    the mean, cites Mathôt & Vilotijević (2022). Requires gazeR
    >= 0.2.10.

                        Changes in version 0.3.4                        

  - Behavioural exclusions report percentages: the step 6 preview shows
    the overall and per-reason percentage of trials excluded and the
    range across participants; log lines and the methods paragraph
    include percentages.

                        Changes in version 0.3.3                        

  - Participants are processed one at a time and then combined, so
    memory use peaks at about one participant's data instead of the
    whole study (helps the browser version). Results are identical to
    processing everyone together.
  - Intermediate columns (interp, pup, index) are no longer kept in the
    processed data.
  - Requires gazeR >= 0.2.9, where smoothing is done within each trial
    (it previously blended the first and last samples of a trial with
    the neighbouring trial).

                        Changes in version 0.3.2                        

  - Exclusion threshold slider is now a percentage (0-100%) instead of a
    proportion (0-1), matching every table in the app.
  - Files with no event messages (e.g. resting state): time-locking and
    event-based baseline are switched off automatically, with a message,
    instead of every trial being excluded as "no event".
  - Clear errors when time-locking or an event baseline has no event
    selected, and when all trials are excluded (lists the actual
    reasons).
  - Trial viewer: blink and missing-data bands have a minimum visible
    width on long trials (e.g. minutes of resting state).
  - Browser version: gazeR and shinygazer are bundled from r-universe;
    lme4 is no longer bundled (gazeR >= 0.2.8 makes it optional).

                        Changes in version 0.3.1                        

  - Methods paragraph (Export tab): ready-to-adapt text describing the
    run, with gazeR citation, every pre-processing setting, exclusions
    and the final sample, plus references for the methods used. Copy or
    download as Markdown.

  - Available from r-universe (https://jgeller112.r-universe.dev). The
    browser version installs gazeR and shinygazer from there instead of
    building its own WebAssembly packages.

  - Requires gazeR >= 0.2.7, where saccades is optional; shiny gazeR no
    longer depends on saccades at all (install and browser build).

                        Changes in version 0.3.0                        

  - Each participant's own sampling rate is used: read from the file
    header for EyeLink .asc files, otherwise estimated from the time
    stamps. If rates differ (e.g. some participants recorded at 500 Hz),
    step 1 lists them and offers to upsample everyone to the highest
    rate with gazer::upsample_gaze() (recommended), or to process each
    participant at their own rate. Requires gazeR >= 0.2.6.
  - Download raw combined, processed or binned data for all participants
    or one, as a single CSV or one CSV per participant (Export and Data
    tabs).

                        Changes in version 0.2.0                        

  - The app can run in the browser with shinylive. A GitHub Actions
    workflow builds WebAssembly versions of saccades, gazeR and
    shinygazer and deploys the app to GitHub Pages.

  - Requires gazeR >= 0.2.5, which now includes read_asc() and fixes to
    smooth_interpolate_pupil() (cubic + interpolate-first + Hanning) and
    lowpass_filter() (edge artifacts). The copies and workarounds
    bundled in shinygazer 0.1.0 have been removed;
    shinygazer::read_asc() is now a re-export of gazer::read_asc().

  - Optional low-pass filter (zero-phase Butterworth,
    gazer::lowpass_filter()) in step 3, applied per trial after
    interpolation.

                        Changes in version 0.1.0                        

  - First release: a Shiny app for gazeR's pupil pre-processing pipeline
    (run_app()), with EyeLink, SR Data Viewer, Tobii (Titta) and Pupil
    Labs Neon input; tracker or noise-based blink detection; behavioural
    joins and exclusions; a trial viewer with missing-data shading;
    quality reports; and time-course, raincloud and pupil-behaviour
    plots with within- or between-subject error.
  - read_asc(): fast single-file reader for EyeLink .asc files (bundled
    until it is part of gazeR).
  - Works around a bug in gazeR <= 0.2.4 where
    smooth_interpolate_pupil() with cubic interpolation,
    interpolate-first and Hanning smoothing filtered the un-interpolated
    signal.