coalesce to follow coercion rules?

[rgknight]

coalesce doesn't coerce an integer to a double

coalesce(as.integer(1), 0)

The conversation rules here seem to suggest that integer to double should be allowable? r-lib/vctrs#7

bind_rows does coerce integers to doubles, which is what initially confused me about which coercions allowed.

df1 <- data.frame(a = as.integer(1))
df2 <- data.frame(a = 0)
tmp <- bind_rows(df1, df2)
class(tmp$a)

coalesce also doesn't follow the factor conversion rules outlined at the link

# Factor + Character doesn't work
a <- c("first", "second", NA) %>% as.factor()
b <- "third"
coalesce(a, b)
# The warning message is also confusing here: Error: Vector 1 has type 'character' not 'integer'

# Factor + factor with different levels returns a warning but doesn't do anything
b <- as.factor(b)
coalesce(a, b)

[krlmlr]

Thanks, confirmed. We should be more consistent here.

[hadley]

This will be implemented in vctrs, not dplyr.