Skip to contents

pressR 0.3.0

This is the first release of the successor line developed as 0.2.0.9000; version 0.2.0 itself was never released. It prepares the package for a first CRAN submission:

  • The maintainer address is now , the documentation language is declared as American English (Language: en-US) and the repository URLs use the canonical CTTIR organization name.

  • The README and vignettes no longer show CRAN status and download badges, which pointed to pages that do not exist before acceptance.

  • Unused suggested packages (‘gganimate’, ‘gifski’, ‘vdiffr’, ‘covr’, ‘pkgdown’) are no longer listed in Suggests.

  • Standardize the named v5 package logo across README, documentation pages and available application headers, with matching light and dark artwork.

This is an explicitly versioned successor. Version 0.1.0 and its manuscript contract remain preserved at commit 5f5b370 and in the baseline audit records.

  • Recognize Novel Pliance content through automatic import and temporary upload filenames; reject malformed numeric fields and contradictory sampling rates.
  • Make generic ASCII and CSV imports strict. Missing sensor values no longer become zero; mismatched widths are no longer padded or truncated. Exported timestamps and sensor IDs round-trip without default-rate substitution. Inferred layouts now align named wide columns and reject incompatible long-form sensor identities. Optional Pliance channel order uses exact field matching.
  • Refuse numeric text outside the supported nonzero magnitude range 1e-290 to 1e290, and conversions with gross disagreement or underflow. These checks prevent observed parser corruption at extreme values and cost additional import time; exact binary replay uses version 2 bundles.
  • Validate stored layout and trial dimensions before analysis. Use scaled arithmetic when finite measurements would otherwise overflow or underflow, and report unrepresentable results explicitly. Missing saddle regions yield an unknown bridge result instead of a negative finding.
  • Add checksum-verified bundles preserving typed metadata and full layout context. Version 2 additionally preserves exact binary measurements and timing; the CSV remains an inspection/exchange view. Version 1 reading retains its decimal precision limitations. Bundle input must come from a trusted source.
  • Preserve existing CSV/report outputs after reported replacement failures, including on Windows; retain and report a backup if restoration fails. Refuse symbolic-link destinations using the portable fs dependency. Archive member aliases no longer silently overwrite an earlier selected member.
  • Correct force-sensor units, unknown calibration, grid-coordinate plots, mask indexing, and all-zero summaries. Coordinate exports use x, y and an explicit unit; rollover uses cop_x, cop_y and coordinate_unit. Interpolation exposes x_grid/y_grid, not fictitious millimetres. COP paths use adjacent valid frames; mean velocity is weighted by observed time. Duplicate COP timestamps are rejected.
  • Preserve declared cohort factor levels and reject conflicting contracts.
  • Separate corrected temporal analysis, historical-method re-execution and measurement-bound reference replay. Corrected sampling and numerical plateau handling intentionally change some historical cycles. Exact replay requires saved cycle boundaries; legacy execution alone cannot promise it.
  • Correct attribution of the six saddle observations to Nyikos (2005), and withdraw unsupported legacy numbers. Reference overlays require source-compatible units, species, gait and mask definitions; reports enable no overlay by default.
  • Add HR/RR workbook import with original cells and source positions, image metadata registration without invented temperatures, bounded archive extraction, source inventories and explicit-key modality linking. Identical files retain identities.
  • Add Shiny comparison uploads, temporal profile selection, real report downloads, selected-mask CSV/report exports, failed-upload state clearing and a configurable 128 MiB upload default. Application launch restores previous options.
  • Render reports in isolated temporary directories and replace completed outputs only after success. Add unit/availability tables and provenance. Treat recording names and metadata as literal report text. Foot report plots and timing tables use separate readable panels; missing provenance is explicit.
  • Add jsonlite, digest and fs imports; readxl, archive and fixture-only writexl are optional dependencies. The successor API snapshot is reviewed independently of the frozen 0.1.0 contract. Declare Shiny 1.6.0, bslib 0.5.0 and testthat 3.2.0 minimums for the application accessibility and test mocking APIs used by this version.

pressR 0.1.0

Initial release

Parsers

Sensor Layouts

S3 classes

  • pr_layout, pr_trial, pr_dataset, pr_cop, pr_mask.

Export and reporting

Shiny application