stats19 3.4.0 2025-10

Major updates to deal with new file names and column names in updated files hosted by the Department for Transport (#268)
Refactored download logic to no longer use .zip files, which are no longer served by the DfT. The package now downloads .csv files directly.
Switched from download.file() to curl::curl_download() for more robust downloads (#258).
Improved documentation around setting a permanent download directory using the STATS19_DOWNLOAD_DIRECTORY environment variable (#211).
Promoted essential packages for data download and formatting (dplyr, lubridate, jsonlite) from Suggests to Imports.
Replaced reshape2 with tidyr for data manipulation (#276).
Added support for downloading the last 5 years of data using year = "5 years" (#261).
The get_stats19_adjustments() function now returns a message explaining that adjustments are included in the main casualty dataset, as the separate adjustments file is no longer provided by the DfT (#266).
Added a new vignette that reproduces the DfT’s pedestrian factsheet (#240, #277).

stats19 3.3.1 2025-01

Updates so package functions fail gracefully when input data is not as expected, e.g. due to URL changes (#252)

stats19 now relies on the stats19_variables object to format the different tables columns (#245) (credit @layik), fixing an issue in which ages were removed from the casualties table, fixing (#235)
If year is less than 2018 the package auto-downloads the full dataset (#239)

Update documentation to account for the shift in table names, replacing accidents with collisions and casualty with casualties (#232)

Minor update to increase default timeout in get_stats19() to 10 minutes (#226)

Major update so the package works with the new csv files (up to 2022)
Deprecation of read_accidents in favour of read_collisions and using consistent collision instead of accidents.
Other minor improvements

Major changes to the datasets provided by the DfT have led to major changes to the package. See (#212) for details.
To reduce code complexity the package no longer supports reading in multiple years
This puts the onus on the user of the package to understand the input data, rather than relying on clever coding to join everything together. Note: you can easily join different years, e.g. with the command purrr::map_dfr().

New function get_ULEZ() to get data on vehicles from a number plate (thanks to Ivo Wengraf)
Added a test to prevent rare failures in get_stats19() when data_dir points to the working directory

Add get_stats19_adjustments() function
Use GH Actions for CI (#177)
Fixed a problem with get_stats19() and multiple years that could be linked with the same data file (#168)
Fix issues with vignettes for CRAN (#190)

Tests now pass on the development version of R (4.0.0)
The package now has a hex sticker! See https://github.com/ropensci/stats19/issues/132 for discussion
The output of formatted crash datasets gains a new column, datetime that is a properly formatted date-time (POSIXct) object in the correct timezone (Europe/London) (#146)
Enables the download of multiple years as per https://github.com/ropensci/stats19/issues/99, thanks to Layik Hama
Users can now set the default data download directory with STATS19_DOWNLOAD_DIRECTORY=/path/to/data in your .Renviron file: https://github.com/ropensci/stats19/issues/141
get_stats19() gains a new argument output_format() that enables results to be returned as an sf object or a ppp object for use the the spatstat package thanks to work by Andrea Gilardi https://github.com/ropensci/stats19/pull/136

Now enables the download of 2018 data
Various bug fixes, see https://github.com/ropensci/stats19/issues
Update website link: https://docs.ropensci.org/stats19/
New work-in-progress vignette on vehicles data: https://docs.ropensci.org/stats19/articles/stats19-vehicles.html

Major change to dl_stats19(): it is now much easier to download STATS19 data. By default ask = FALSE in get_stats19() and dl_stats19().

get_stats19() gains an ask argument (TRUE by default, set as FALSE to make road crash data access even more automated!)
The date column now is of the correct class after formatting POSIXct. See #86
Added a NEWS.md file to track changes to the package.