errors 0.4.2

• Add support for PDG rounding rules (@davidchall #59 addressing #45).
• Improve formatting of null values (#48).

errors 0.4.1

• Switch from size (deprecated in ggplot2 v3.4.0) to linewidth aesthetic
  in geom_errors() (#55).
• Implement methods for dealing with missing values in errors objects,
  and fix na.rm behavior for summary methods (#56).
• Performance improvements for data.frame methods.

errors 0.4.0

• Add new geom_errors() function to automatically add errorbars to ggplot2
  plots for variables of class errors (#52).
• Some fixes in plot.errors, improved units compatibility.
• Implement errors methods for duplicated, anyDuplicated and unique.
• Fix operators %/% and %% (#54).

errors 0.3.6

• Fix check tolerance when correlations are approx. 1 (#42).

errors 0.3.5

• Internal performance improvements (#39).

errors 0.3.4

• vctrs methods are now implemented for compatibility with dplyr 1.0 (#38).
• Implement prettier str print (#36).
• Implement drop method for data frames.

errors 0.3.3

• Fix min, max and range (#35).
• Rewrite plot to fix issues with limits and labels (#35).

errors 0.3.2

Minor changes and fixes

• Add CITATION for the R Journal paper.
• Add the paper as a vignette.

errors 0.3.1

New features

• Add a new dataset from the Annex H of the Guide to the Expression of
  Uncertainty in Measurement (GUM) and associated example (f6e5461, see ?GUM.H.3).
• Provide methods for logical values to automatically reinterpret NA as
  NA_real_ (#30).
• Add delayed S3 registration mechanism for R >= 3.6.0 (955e7ad).

Minor changes and fixes

• Move GUM examples to ?`errors-package` and provide new examples for
  ?correl (aed40eb).
• Simplify object ID handling (#27).
• Fix formatting in presence of missing values (as part of #30).
• Fix Summary.errors to deal with named arguments (dd4607b).

errors 0.3.0

New features

• Implement as.list for errors objects (7aff546).
• Implement support for correlations between variables (#26). The correl and covar methods set and retrieve pairwise correlations or covariances, respectively, between errors objects, thereby completing the full first-order Taylor series method of propagation of uncertainty.
• Add a new dataset from the Annex H of the Guide to the Expression of Uncertainty in Measurement (GUM) (as part of #26; see ?GUM.H.2 and examples under ?correl).
• Implement all.equal for errors objects (as part of #26), which returns TRUE for different (different internal identifier) objects with the same quantity values and uncertainty. To compare whether two variables are exactly the same object, identical should be used instead.

Minor changes and fixes:

• Fix encoding issues for the plus-minus symbol (#24).
• Fix pillar representation (14df89b).
• Fix coercion issues (0e04519).
• Improve documentation (as part of #26).