NEWS.md
NULL POSIXct (dates) (#39 @romainfrancois)"NNW" to "up-left" etc. Compass directions still work, but the built-in dataset purpose has different names and documentation will gradually change to the new terms.behead_if() is for tiered headers within the same row or column. It takes filter functions similarly to dplyr::filter() to decide which cells to treat as headers, and can be applied more than once to the same row or column of headers until every tier has been dealt with.merge_rows() and merge_cols() combines header text when it is split over multiple cells.behead(), enhead(), partition() and rectify() give a more informative error message for non-distinct cells, for example when trying to pass cells from more than one sheet to these functions (@gregrs-uk, #15).This version makes some big breaking changes for the sake of a more intuitive grammar. It comes with much more documentation in the online book Spreadsheet Munging Strategies.
The main new workhorses:
behead() takes one level of headers from a pivot table and make it part of the data. Chain this function to gradually strip every level of header away until you have tidy data.spatter() is a data-type aware version of tidyr::spread() and is a common final step.partition() breaks up small-multiples on a single sheet, so you can handle them individually.rectify() visualises the cells in the console as they would look in a spreadsheet.The previous version can be installed as follows.
devtools::install_version("unpivotr", version = "0.3.1", repos = "http://cran.us.r-project.org")NNW() etc. has been removed in favour of the verbose join_header(), which has itself been renamed to enhead() to suggest its similarity to behead() (though they are not complements).enhead() (formerly join_header() now follows the tidyverse convention of fct for ‘factor’ and ord for ‘ordered factor’.enhead() (formerly join_header()) now uses col_names and row_names as arguments instead of colnames and rownames, for consistency with tidyr.behead() is takes one level of headers from a pivot table and make it part of the data. Think of it like tidyr::gather(), except that it works when there is more than one row of headers (or more than one column of row-headers), and it only works on tables that have first come through enhead() (formerly join_header() or tidyxl::xlsx_cells().rectify() displays cells as though in a spreadsheet, rather than in the ‘melted’ form of enhead() (formerly join_header()) and tidyxl::xlsx_cells(). This is useful for understanding the structure of a pivot table as a human, when planning how to unpivot it. A print method is available to render large datasets in the browser or the RStudio viewer pane.partition() divides a grid of cells into partitions containing individual tables. Give it the corner cells of each table on a spreadsheet.pack() packs cells values from separate columns per data type into one list-column. unpack() is the complement.isolate_sentinels() move sentinel values into a separate column, leaving NA behind (or NULL for list-columns).spatter() is like tidyr::spread(), but preserves mixed data types.enhead() (formerly join_header()) now returns a data_type column that names the column that contains the value of a cell, similar to tidyxl::xlsx_cells().enhead() (formerly join_header() now follows the tidyverse convention of fct for ‘factor’ and ord for ‘ordered factor’.enhead() (formerly join_header()) gains a drop = TRUE argument to control whether to discard cells that don’t have a matching header (e.g. ones that are left of the leftmost header in enhead(x, y, "NNW")).justify() moves one set of cells to the same positions as another set. This is useful when header cells aren’t at the corner of the cells they describle. Put the header cells into justify(), along with cells that are at the corner.purpose (built-in dataset) gains a new list-member small-multiples.tidy_table().dplyr::distinct(), which doesn’t handle list columns).dplyr/rlang combination instead of the old dplyr/lazyeval one.This release overhauls the tidy_table() function of unpivotr to preserve the original data types of table cells and to support HTML tables.
tidytable() has been renamed tidy_table(). tidytable() is an error, rather than a deprecation warning, because tidy_table() is so different from before.tidy_table() method and vignette for HTML.tidy_table() method for matrices. Convert matrices to data.frames first, choosing what to do with row and column names.tidy_table() returns only relevant columns, according to the data types of the columns in the given data frame. It uses tibble::type_sum() to determine the column type and to name the columns, so whereas characters used to be returned in a column called character, they are now returned in a column called chr. The full list of column names is in ?tidy_table and is chr, cplx, cplx, dbl, fctr, int, lgl, list. The columns fctr and list are list-columns, where each element is itself a list. This means that factors with different levels are kept separate. For HTML tables, an html column is returned containing the standalone HTML of each cell.rowname and colname arguments to tidy_table() now default to FALSE.tibble.