rows_patch fails when y has key values that don't occur in x

[melissagwolf]

I tried to use row_patch to fill in missing values in a data frame. It failed by design, because it didn't meet the requirement for the function: "key values in y must occur in x". However, I was able to successfully execute the row patch by using the match function from base R (see below).

This isn't a typical bug report because it is clearly stated that this is a limitation of the function, but I wanted to point out that it can be done using a different function from base R. Since row_patch is currently an experimental function in dplyr, I wanted to provide feedback to hopefully improve it. row_patch is much cleaner and I would prefer to use it if possible.

It would also be nice if row_patch could preserve other variables in y that don't occur in x. match can handle this as well.

D1 <- data.frame(
  id=seq(1,3),
  x=c("cow",NA,"sheep"))

D2 <- data.frame(
  id=seq(1,4),
  x=c("cow","turtle","parrot","frog"))

#rows_patch fails
D1 %>%
  rows_patch(D2, by = "id")

#match works
na <- is.na(D1$x)

D1$x[na] <- D2$x[match(D1$id[na],D2$id)]

[hadley]

Reprex:

library(dplyr, warn.conflicts = FALSE)
D1 <- data.frame(
  id = seq(1, 3),
  x = c("cow", NA, "sheep")
)

D2 <- data.frame(
  id = seq(1, 4),
  x = c("cow", "turtle", "parrot", "frog")
)

# rows_patch fails
D1 %>%
  rows_patch(D2, by = "id")
#> Error: Attempting to patch missing rows.

# match works
na <- is.na(D1$x)

D1$x[na] <- D2$x[match(D1$id[na], D2$id)]

^{Created on 2021-09-16 by the reprex package (v2.0.0)}

[hadley]

Might be worthwhile to have @DavisVaughan think about these functions once he's done with his current work on joins.

[DavisVaughan]

Probably would be fixed by #5588 if this comment is still applicable in that PR #5588 (comment)

I think is makes sense that keys in y that do not exist in x should always be silently ignored when updating/deleting/patching (no changes needed - we do this in this PR).

And @mgirlich followed up that it might be useful to make that an optional check, rather than just silently ignoring them #5588 (comment)