[IMPALA-1898] Substituting the agg output in analysis has issues with explicit select-list aliases.

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: Impala 2.1.1
• Fix Version/s: Impala 2.2.7, Impala 2.3.0
• Labels: correctness sql-language
• Target Version: Impala 2.2.7, Impala 2.3.0
• Assignee: Tim Armstrong
• Reporter: Alexander Behm
• Created: 18/Mar/15 05:43
• Updated: 01/Sep/15 19:55
• Resolved: 23/Jul/15 17:33

Description

In the query below, it appears we incorrectly substitute the "timestamp_col" SlotRef within the extract() function with the expr that has the select-list alias "timestamp_col" (i.e., the extract() function itself).
extract() returns an INT.

select extract(timestamp_col, 'hour') AS timestamp_col, string_col,
sum(double_col) AS double_total
FROM functional.alltypes
GROUP BY 1, 2
ERROR: AnalysisException: No matching function with signature: extract(INT, STRING).

Another manifestation of this same bug:

create database wes;
create table wes.airlines_parquet (cancelled tinyint) stored as parquet;
insert into wes.airlines_parquet values(1), (2), (3), (NULL), (0), (-1);
select cast(cancelled as boolean) as `cancelled`, count(*) as `count`
from wes.`airlines_parquet`
group by 1;
ERROR: AnalysisException: select list expression not produced by aggregation output (missing from GROUP BY clause?): CAST(cancelled AS BOOLEAN)