The tidy approach discourages row names but they still happen, should we provide a way to match by row names ? base::merge allows it using by = "row.names" or by = 0.
I don't like the former since it's confused with column names, the latter is not explicit and is coerced to character on a by vector so I don't love it either.
We might use by = "*row*".
I thought we might have a helper by_row() which is either recognised by NSE or is just NA with a class and printing method, but used in vectors the attribute/class will be lost by c() and the NSE handling would look awful.
So I think "row" it is.
The tidy approach discourages row names but they still happen, should we provide a way to match by row names ? base::merge allows it using
by = "row.names"orby = 0.I don't like the former since it's confused with column names, the latter is not explicit and is coerced to character on a
byvector so I don't love it either.We might use
by = "*row*".I thought we might have a helper
by_row()which is either recognised by NSE or is just NA with a class and printing method, but used in vectors the attribute/class will be lost byc()and the NSE handling would look awful.So I think "row" it is.