safejoin

[moodymudskipper]

what about a full package safejoin containing eat with variants of all dplyr joins using safe features developed for eat (safe_inner_join etc...)

[moodymudskipper]

Don't use grouping option and dots, but do the unicity checks.

Make all relevant parameters default to options so we choose the degree of safety for a whole project or chunk of code.

[moodymudskipper]

see if safe_nest_join makes sense. Actually nest_join might be the function eat should wrap instead of left_join.

[moodymudskipper]

when eating and replacing existent columns it would be nice to replace without moving the column to the end. also (maybe it is already there ?) it would be nice to have a "selective" (find a better name!) option which basically means that on a left join we take the value from rhs only when it matches, and keep existing. to keep lhs in priority the option could be named "coalesce". Or both could have names that are variations from coalesce.

[moodymudskipper]

maybe an at parameter to decide where to insert the newly created columns ?

[moodymudskipper]

2 new checks to implement :

• check that every row of is used : M like Match ? and on the right N because it's after M ? and what we called X and Y can be named U like Unique and V for same reasons ?
• L like levels : fails / warn / msg when levels are not identical on join columns