[ARROW-13737] [C++] Support scalar columns in hash aggregations (was: hash_sum on scalar column segfaults)

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 6.0.0
• Component/s: C++
• Labels: kernel pull-request-available query-engine
• External issue URL: https://github.com/apache/arrow/issues/29370
• Assignee: David Li
• Reporter: Neal Richardson
• Created: 24/Aug/21 17:17
• Updated: 11/Jan/23 08:35
• Resolved: 30/Aug/21 16:24

Description

Reproducer in R. Works fine without grouping, but segfaults if you group:

tab <- Table$create(a=1:4, b=6:9)

tab %>% 
  mutate(counter = 1) %>% 
  summarize(count = sum(counter))

Table
1 rows x 1 columns
$count <double>

tab %>% 
  mutate(counter = 1) %>% 
  group_by(a) %>% 
  summarize(count = sum(counter))

 *** caught segfault ***
address 0x0, cause 'invalid permissions'

Traceback:
 1: .Call(`_arrow_ExecPlan_run`, plan, final_node)
 2: ExecPlan_run(self, node)
 3: plan$Run(final_node)
 4: do_exec_plan(.data, group_vars = dplyr::group_vars(.data))
 5: do_arrow_summarize(.data, ...)
...

Issue Links

• blocks: ARROW-13501 [R] Bindings for count aggregation
• links to: GitHub Pull Request #10994