[ARROW-9389] [C++] Can't call isin/match through CallFunction

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.0.0
• Component/s: C++
• Labels: compute pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/25470
• Assignee: Wes McKinney
• Reporter: Neal Richardson
• Created: 09/Jul/20 17:05
• Updated: 11/Jan/23 08:06
• Resolved: 11/Jul/20 23:16

Description

From R:

library(arrow)
a <- Array$create(1:4)
b <- Array$create(c(2L, 4L, 3L))
arrow:::call_function("isin", a, b)

says that "isin" takes only 1 argument, not 2, which doesn't make sense. In C++ scalar_set_lookup.cc, I see auto isin = std::make_shared<ScalarFunction>("isin", Arity::Unary());, which is the source of that validation I guess, but the kernel in api_scalar.cc has signature Result<Datum> IsIn(const Datum& values, const Datum& value_set, ExecContext* ctx).

If I actually call "isin" with one argument, i.e. arrow:::call_function("isin", a), it segfaults.

Changing the definition to Arity::Binary(), it accepts 2 arguments, but it errors with NotImplemented: Function isin has no kernel matching input types (array[int32], array[int32])

Issue Links

• links to: GitHub Pull Request #7715