Changelog • wellr

wellr 0.4.0

Well IDs are now a vctrs type, created with well() or as_well(), that records the size of its plate. It is a character vector underneath, so values, printing, comparisons, %in%, joins and string functions are unchanged. It sorts in plate order and works as a discrete scale in ggplot2. Combining it with character, or with wells from a different plate, gives character.
The readers (plate_read_biotek(), plate_read_biotek_wl(), plate_read_tecan() and plate_read_meta()) return the well type, using the plate size recorded in the file, a new plate argument, or else inferring it. well_plate(), well_from_index(), well_mat_to_df() and well_mat_frames_to_df() also return it, and well_format() keeps it.
well_to_index() and well_from_index() take the plate size from the well IDs when plate isn’t given.
New well_plate_size() and well_plate_dims() give the plate of well IDs. 48 well plates are now supported.
New well_row() and well_col() give the rows and columns of well IDs as factors whose levels are all the rows or columns of the plate.
well_plot() draws every well of the plate, with wells that have no data in grey, in every facet. It gains a plate argument.

well_df_to_mat() uses plate when given, which was always overwritten. Wells missing from the data are NA, instead of the values shifting into the wrong wells and being recycled. Duplicated wells, and wells outside the plate, are now errors.
well_mat_to_df() uses the shape of the matrix, instead of assuming a 96 well plate and mislabelling the wells of other plates.
well_plot() uses the colour argument for the well borders, which was ignored.
well_join() is vectorised, and errors for rows outside 1 to 26 or strings that aren’t rows, instead of giving IDs like “NA05”.
well_from_index() errors for indices outside the plate.
Errors for invalid arguments now name the argument and say what’s wrong.
Fixed deprecation warnings from readr 2.2.0, dplyr 1.2.0 and tidyselect 1.2.0.

wellr now requires R 4.1.0, readr 2.0.0, ggplot2 3.4.0 and rlang 1.1.0, and no longer depends on scales.