inner_join incorrectly throws error

[lymanmark]

When using inequality joins and setting relationship to "one-to-one", an error is thrown despite the criteria for a "one-to-one" match being satisfied.

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

x <- data.frame(
  x = 543,
  y = 178,
  x_max = 567,
  x_text = "Text 1"
)

y <-
  data.frame(
    x = c(457, 450),
    y = c(114, 167),
    x_max = c(606, 660),
    y_text = c("Text 2", "Text 3")
  )

# Returns one row as expected
inner_join(
  x,
  y,
  by = join_by(
    overlaps(x, x_max, x, x_max),
    closest(y >= y)
  )
)
#>   x.x y.x x_max.x x_text x.y y.y x_max.y y_text
#> 1 543 178     567 Text 1 450 167     660 Text 3

# Throws error with multiple matches
inner_join(
  x,
  y,
  by = join_by(
    overlaps(x, x_max, x, x_max),
    closest(y >= y)
  ),
  relationship = "one-to-one"
)
#> Error in `inner_join()`:
#> ! Each row in `x` must match at most 1 row in `y`.
#> ℹ Row 1 of `x` matches multiple rows in `y`.
#> Backtrace:
#>      ▆
#>   1. ├─dplyr::inner_join(...)
#>   2. ├─dplyr:::inner_join.data.frame(...)
#>   3. │ └─dplyr:::join_mutate(...)
#>   4. │   └─dplyr:::join_rows(...)
#>   5. │     └─dplyr:::dplyr_locate_matches(...)
#>   6. │       ├─base::withCallingHandlers(...)
#>   7. │       └─vctrs::vec_locate_matches(...)
#>   8. ├─vctrs:::stop_matches_relationship_one_to_one(...)
#>   9. │ └─vctrs:::stop_matches_relationship(...)
#>  10. │   └─vctrs:::stop_matches(...)
#>  11. │     └─vctrs:::stop_vctrs(...)
#>  12. │       └─rlang::abort(message, class = c(class, "vctrs_error"), ..., call = call)
#>  13. │         └─rlang:::signal_abort(cnd, .file)
#>  14. │           └─base::signalCondition(cnd)
#>  15. └─dplyr (local) `<fn>`(`<vc______>`)
#>  16.   └─dplyr:::rethrow_error_join_relationship_one_to_one(cnd, error_call)
#>  17.     └─dplyr:::stop_join_matches_multiple(...)
#>  18.       └─dplyr:::stop_join(...)
#>  19.         └─dplyr:::stop_dplyr(...)
#>  20.           └─rlang::abort(...)

^{Created on 2023-04-26 with reprex v2.0.2}

[DavisVaughan]

Minimal reprex

library(vctrs)

a <- data_frame(x = 1, y = 5, z = 3)
b <- data_frame(x = c(1, 2), y = c(4, 3), z = c(1, 2))

vec_locate_matches(
  a,
  b,
  condition = c("<=", ">=", ">="),
  filter = c("none", "none", "max"),
  relationship = "none"
)
#>   needles haystack
#> 1       1        2

vec_locate_matches(
  a,
  b,
  condition = c("<=", ">=", ">="),
  filter = c("none", "none", "max"),
  relationship = "one-to-one"
)
#> Error in `vec_locate_matches()`:
#> ! Each value of `needles` can match at most 1 value from `haystack`.
#> ✖ Location 1 of `needles` matches multiple values.
#> Backtrace:
#>     ▆
#>  1. ├─vctrs::vec_locate_matches(...)
#>  2. └─vctrs:::stop_matches_relationship_one_to_one(...) at vctrs/R/match.R:320:2
#>  3.   └─vctrs:::stop_matches_relationship(...) at vctrs/R/match.R:491:2
#>  4.     └─vctrs:::stop_matches(...) at vctrs/R/match.R:563:2
#>  5.       └─vctrs:::stop_vctrs(...) at vctrs/R/match.R:360:2
#>  6.         └─rlang::abort(message, class = c(class, "vctrs_error"), ..., call = call) at vctrs/R/conditions.R:64:2

Notably this doesn't error, so it has something to do with nested containment groups because b in the above example isn't fully ordered on all columns.

a <- data_frame(x = 1, y = 5, z = 3)
b <- data_frame(x = c(1, 2), y = c(3, 4), z = c(1, 2))

vec_locate_matches(
  a,
  b,
  condition = c("<=", ">=", ">="),
  filter = c("none", "none", "max"),
  relationship = "one-to-one"
)
#>   needles haystack
#> 1       1        2