Join on multi valued dimension fails when using CTE
- Dominant language
- Java
- Stars
- 14.1k
- Forks
- 3.8k
- Avg merge
- 2d 58m
- Merged PRs (30d)
- 233
Description
Consider a datasource having a multi valued dimension. If there is a full join query like below it works
```
SELECT
(COALESCE(base."channel_mvd", comparison."channel_mvd")) AS "channel",
(ANY_VALUE(base."added" - comparison."added")) AS "added_delta"
FROM (
SELECT
"channel_mvd",
(SUM(added)) AS "added"
FROM "wikipedia" CROSS JOIN UNNEST(MV_TO_ARRAY("channel")) as u("channel_mvd")
WHERE ("__time" >= '2016-06-27T03:00:00.000Z' AND "__time" < '2016-06-27T06:00:00.000Z')
GROUP BY 1
) AS base
FULL JOIN (
SELECT
"channel_mvd",
(SUM(added)) AS "added"
FROM "wikipedia" CROSS JOIN UNNEST(MV_TO_ARRAY("channel")) as u("channel_mvd")
WHERE ("__time" >= '2016-06-27T00:00:00.000Z' AND "__time" < '2016-06-27T03:00:00.000Z')
GROUP BY 1
) AS comparison ON (base."channel_mvd" IS NOT DISTINCT FROM comparison."channel_mvd")
GROUP BY 1
ORDER BY "added_delta" DESC
LIMIT 250
```
added `CROSS JOIN UNNEST(MV_TO_ARRAY("channel")) as u("channel_mvd")` just to simulate column being mvd as channel is not. Anyways if any mvd is used directly in the join above then also it works.
However, if we do the same thing with CTE then it fails (assume `channel_mvd` is mvd here) -
```
WITH base AS (
SELECT
"channel_mvd",
(SUM(added)) AS "added"
FROM "wikipedia"
WHERE ("__time" >= '2016-06-27T03:00:00.000Z' AND "__time" < '2016-06-27T06:00:00.000Z')
GROUP BY 1
ORDER BY "channel_mvd" DESC
LIMIT 250
)
SELECT
(COALESCE(base."channel_mvd", comparison."channel_mvd")) AS "channel",
(ANY_VALUE(base."added" - comparison."added")) AS "added_delta"
FROM base
LEFT JOIN (
SELECT
"channel_mvd",
(SUM(added)) AS "added"
FROM "wikipedia"
WHERE ("__time" >= '2016-06-27T00:00:00.000Z' AND "__time" < '2016-06-27T03:00:00.000Z') AND (
("channel_mvd") IN (SELECT "base"."channel_mvd" FROM "base")
)
GROUP BY 1
) AS comparison ON (base."channel_mvd" IS NOT DISTINCT FROM comparison."channel_mvd")
GROUP BY 1
ORDER BY "added_delta" DESC
LIMIT 250
```
It fails with ` QueryUnsupportedException{msg=Joining against a multi-value dimension is not supported.` and if I add `CROSS JOIN UNNEST(MV_TO_ARRAY("channel")) AS u ("channel_mvd")` then it fails with `Unhandled Query Planning Failure, see broker logs for details`
Contributor guide
Research direction
Start by running the two SQL examples in the issue against a datasource with a multi-valued dimension, then inspect the broker logs for the unhandled planning failure. Compare the direct join and CTE paths; done means the CTE-based join no longer raises the multi-value-dimension or query-planning exceptions.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- java, sql
- Domain
- databases
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 45/100