scount(), sadd_count(), tally() and add_tally() are now based on collapse::fcount() with some changes to better match the arguments and behavior of the corresponding{dplyr} functions. However, much work was needed to align these speedy functions to the behavior of their tidy counterparts, in part because the tidy functions sort of the frequencies columns and the collapse::fcount() function that is used sorts on the categories labels... and it does not sort on it by default.The alternate assignments %<-% and %->% do not evaluate collect() in a try() construct, silently swallowing any error any more. A collect.default() method is defined to make sure to return intact any object that cannot be collected without error.
Default print() methods are restored for data.frame and data.table objects.
Since {dplyr} functions now return the same object as it is receiving (tibble -> tibble, but data.frame -> data.frame and data.table -> data.table), the use of lazy_dt() is mandatory to benefit of {dtplyr} translation into {data.table} statements and its speed performance, examples are adapted accordingly.
Bug correction in sunite(), the argument col= was setting the name of the column always to "col".
is_dtx() and similar functions now also detect spec_tbl_df objects from {readr} and groupedData of {datasets} are true data frames, even with strict = TRUE. Consequently, default_dtx() converts also these objects.
as_dtx() and similar functions get labels and units from the corresponding attributes of groupedData objects and apply them to each vector in the resulting data frame.
The tidy functions are reworked to use the original function name, not the one with a "t" prefix.
The functions that list family functions now start with list_, like list_tidy_functions(), list_speedy_functions() or list_fstat_functions().
dtx_rows() and other similar functions are added to avoid using dtx(tribble(...)).
Addition of "tidy" 't' functions and "speedy" 's' functions. The tidy functions are mostly renamed versions of {dplyr} or {tidyr} functions, while the speedy functions have a similar interface but are from {collapse} or {data.table} for a big part of them. Arguments are homogenized and sometimes reduced to restrict to features that are compatibles in the two groups.
The function n() is not working in ssummarise(). Use fn(var) instead, and you can also use it in tsummarise().
Addition of "fast" functions from {collapse} + fn() and fna().
The %xf% operators where xcan be -, /, etc. for a more readable alternative to TRA= in the fast functions.
More robust strict is_dt_() functions for grouped data.
dtf(), creating a column with name .rowname (or the value in getOption("SciViews.dtx.rownames")) will transform it into the row names of the resulting data.frame object.When group_by() is used, collect() creates a grouped_df. This special object is now recognized and converted by as_dtx() and co into ungrouped data frame.
collect_dtx()/dtf()/dtt()/dtbl() added to force the right data frame class (with collect_dtbl() giving the same result as collect(), except that grouped data are also ungrouped).
The most important functions are added to manage data frames (being data.frame, data.table, or tibble tbl_df objects) in a more coherent way, with the possibility to define one's preferred object.
The dtx(), as_dtx() and default_dtx() functions can be used in packages to automatically convert data frames into the preferred object without forcing it to the user.
The alternate assignment operators %<-% and %->% collect result from {dtplyr} (or {dbplyr}) and convert it into the preferred data frame object. Also, the multiple or deconstructing assignment implemented in {zeallot} is also usable here in addition to the conversion into the preferred data frame.
Row names were not printed when a data.frame is using {pillar}. It is now a column named and with class <rownames>.
as_matrix() applied to a tbl_df object did not honor .rownames. Corrected.