dtplyr (development version)

dtplyr 1.0.1

Better handling for .data and .env pronouns (#138).
dplyr verbs now work with NULL inputs (#129).
joins do better job at determining output variables in the presence of duplicated outputs (#128). When joining based on different variables in x and y, joins consistently preserve column from x, not y (#137).
lazy_dt() objects now have a useful glimpse() method (#132).
group_by() now has an arrange parameter which, if set to FALSE, sets the data.table translation to use by rather than keyby (#85).
rename() now works without data.table attached, as intended (@michaelchirico, #123).

Converted from eager approach to lazy approach. You now must use lazy_dt() to begin a translation pipeline, and must use collect(), as.data.table(), as.data.frame(), or as_tibble() to finish the translation and actually perform the computation (#38).

This represents a complete overhaul of the package replacing the eager evaluation used in the previous releases. This unfortunately breaks all existing code that used dtplyr, but frankly the previous version was extremely inefficient so offered little of data.table's impressive speed, and was used by very few people.
dtplyr provides methods for data.tables that warning you that they use the data frame implementation and you should use lazy_dt() (#77)
Joins now pass ... on to data.table's merge method (#41).
ungroup() now copies its input (@christophsax, #54).
mutate() preserves grouping (@christophsax, #17).
if_else() and coalesce() are mapped to data.table's fifelse() and fcoalesce() respectively (@michaelchirico, #112).

Maintenance release for CRAN checks.
inner_join(), left_join(), right_join(), and full_join(): new suffix argument which allows you to control what suffix duplicated variable names receive, as introduced in dplyr 0.5 (#40, @christophsax).
Joins use extended merge.data.table() and the on argument, introduced in data.table 1.9.6. Avoids copy and allows joins by different keys (#20, #21, @christophsax).

This is a compatibility release. It makes dtplyr compatible with dplyr 0.6.0 in addition to dplyr 0.5.0.