Array(LowCardinality(String)) has() function slower #10880
Description
hello,
I store the exp_id list in an Array using Array(LowCardinality(String)) for our A/B test 。
example exps array data: ['zeyudu_rec_6096_1','default','zeyudu_rec_6096_2','zeyudu_rec_6096_3',....]
My Query:
SELECT
multiIf(has(exps_lowcard, 'zeyudu_rec_6096_1'), 'control', has(exps_lowcard, 'zeyudu_rec_6096_2'), '6096_2', 'other') AS expr_model_v20,
SUM(multiIf(windowType = 'CLICK', ((((4 * hasFollow) + hasLike) + hasComment) + (4 * hasShare)) + hasCollect, NULL) * SampleWeight) AS `SUM(ces)`
FROM default.explore_feed_bitmap
GROUP BY expr_model_v20
result: 1.1E rows taken 34 seconds
I make anther test :
I manage an external exp_id dictionay and then store the exp_id list like [1001,3001,4002,6001,6002.....]
test query:
SELECT
multiIf(has(exps_array, 6001), 'control', has(exps_array, 6002), '6002', 'other') AS expr_model_v20,
SUM(multiIf(windowType = 'CLICK', ((((4 * hasFollow) + hasLike) + hasComment) + (4 * hasShare)) + hasCollect, NULL) * SampleWeight) AS `SUM(ces)`
FROM default.explore_feed_bitmap
GROUP BY expr_model_v20
result: 1.1E rows taken 3.2 seconds ,
clickhouser version: 19.17.4.11
can you help check this issue. I think the LowCardinality Array should has the close performance with an Array stores Int。
If you need more information, I will update .
thanks
johnny