Problem:
Count distinct values in an array filtered by another array on same row (and agg higher).
Explanation:
Using this data:
In the Size D70, there are 5 pcs available (hqsize), but shops requests 15. By using the column accumulatedNeed, the 5 first stores in the column shops should receive items (since every store request 1 pcs). That is [4098,4101,4109,4076,4080].
It could also be that the values in accumulatedNeed would be [1,4,5,5,5,...,15], where shop 1 request 1 pcs, shop2 3 pcs, etc. Then only 3 stores would get.
In the size E75 there is enough stock, so every shop will receive (10 shops):
Now i want the distinct list of shops from D70 & E75, which would be be final result:
[4098,4101,4109,4076,4080,4062,4063,4067,4072,4075,4056,4058,4059,4061] (14 unique stores) (4109 is only counted once)
Wanted result:
[4098,4101,4109,4076,4080,4062,4063,4067,4072,4075,4056,4058,4059,4061]. (14 unique stores)
I'm totally open to structure the data otherwise if better.
The reason why it can't be precalculated is that the result depends on which shops that are filtered on.
Additional issue
The answer below from Vdimir is good and I've used it as basics for the final solution, but the solution does not cover (partial fullfillment).
If the stock number is in the runningNeed array we are all goodt, but remainers are not handled.
If you got:
select 5 as stock,[2,2,3,3] as need, [1,2,3,4] as shops, arrayCumSum(need) as runningNeed,arrayMap(x -> (x <= stock), runningNeed) as mask
You will get:
This is not correct since the 3rd shop should have 1 from stock (5-2-2 = 1)
I can't seem to get my head around how to make an array with "stock given", which in this case would be [2,2,1,0]
I use this query to create table with data similar to your screenshot:
CREATE TABLE t
(
Size String,
hqsize Int,
accumulatedNeed Array(Int),
shops Array(Int)
) engine = Memory;
INSERT INTO t VALUES ('D70', 5, [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15], [4098,4101,4109,4076,4080,4083,4062,4063,4067,4072,4075,4056,4057,4058,4059]),('E75', 43, [1,2,3,4,5,6,7,8,9,10], [4109,4062,4063,4067,4072,4075,4056,4058,4059,4061]);
Find which shops that can receive enough items:
SELECT arrayMap(x -> (x <= hqsize), accumulatedNeed) as mask FROM t;
┌─mask────────────────────────────┐
│ [1,1,1,1,1,0,0,0,0,0,0,0,0,0,0] │
│ [1,1,1,1,1,1,1,1,1,1] │
└─────────────────────────────────┘
Filter not fulfilled shops according to this mask:
Note that shops and accumulatedNeed have to have equals sizes.
SELECT arrayFilter((x,y) -> y, shops, mask) as fulfilled_shops, arrayMap(x -> (x <= hqsize), accumulatedNeed) as mask FROM t;
┌─fulfilled_shops─────────────────────────────────────┬─mask────────────────────────────┐
│ [4098,4101,4109,4076,4080] │ [1,1,1,1,1,0,0,0,0,0,0,0,0,0,0] │
│ [4109,4062,4063,4067,4072,4075,4056,4058,4059,4061] │ [1,1,1,1,1,1,1,1,1,1] │
└─────────────────────────────────────────────────────┴─────────────────────────────────┘
Then you can create table with all distinct shops:
SELECT DISTINCT arrayJoin(fulfilled_shops) as shops FROM (
SELECT arrayMap(x -> (x <= hqsize), accumulatedNeed) as mask, arrayFilter((x,y) -> y, shops, mask) as fulfilled_shops FROM t
);
┌─shops─┐
│ 4098 │
│ 4101 │
│ 4109 │
│ 4076 │
│ 4080 │
│ 4062 │
│ 4063 │
│ 4067 │
│ 4072 │
│ 4075 │
│ 4056 │
│ 4058 │
│ 4059 │
│ 4061 │
└───────┘
14 rows in set. Elapsed: 0.049 sec.
Or if you need single array group it back:
SELECT groupArrayDistinct(arrayJoin(fulfilled_shops)) as shops FROM (
SELECT arrayMap(x -> (x <= hqsize), accumulatedNeed) as mask, arrayFilter((x,y) -> y, shops, mask) as fulfilled_shops FROM t
);
┌─shops───────────────────────────────────────────────────────────────────┐
│ [4080,4076,4101,4075,4056,4061,4062,4063,4109,4058,4067,4059,4072,4098] │
└─────────────────────────────────────────────────────────────────────────┘
If you need data only from D70 & E75 you can filter extra rows from table with WHERE before.
Have such table and data:
create table sensor_values(
dt DateTime default now(),
value UInt32
)
engine MergeTree()
partition by toYYYYMM(dt)
order by tuple();
insert into sensor_values(value) values (1), (2), (11), (13), (4), (17), (5), (8);
Data:
value
-----
1
2
11
13
4
17
5
8
I would like to select data in range from first bad value (11) to last bad value (17). Bad values are more than 10.
Desired range after select:
value
-----
11
13
4
17
My first thoughts were to define whether value bad or not and then to calculate (some how) accumulative sum:
value isBad cumSum
--------------------
1 0 0
2 0 0
11 1 1
13 1 2
4 0 2
17 1 3
5 0 3
8 0 3
Then I would select from min(cumSum) to max(cumSum) - 1 but I miss last bad value.
How can I get the last value included in select result?
You can try to use either the window-functions (see: runningDifference, neighbor) or array-functions:
SELECT arrayJoin(slice) as result
FROM (
SELECT
groupArray(data) AS arr,
arrayFirstIndex(x -> (x > 10), arr) AS first_index,
(length(arr) - arrayFirstIndex(x -> (x > 10), arrayReverse(arr)) + 1) AS last_index,
arraySlice(arr, first_index, last_index - first_index + 1) AS slice
FROM
(
/* test dataset */
SELECT arrayJoin([1, 2, 11, 13, 4, 17, 5, 8]) AS data
)
)
/*
┌─result─┐
│ 11 │
│ 13 │
│ 4 │
│ 17 │
└────────┘
*/
From the quantile function documentation:
We recommend using a level value in the range of [0.01, 0.99]. Don't use a level value equal to 0 or 1 – use the min and max functions for these cases.
Does this also applies for quantileExact and quantilesExact functions?
In my experiments, I've found that quantileExact(0) = min and quantileExact(1) = max, but cannot be sure about it.
That recommendation is not about accuracy but about complexity of quantile*.
quantileExact is much much heavier than max min.
See the time difference, min / max 8 times faster even on a small dataset.
create table Speed Engine=MergeTree order by X
as select number X from numbers(1000000000);
SELECT min(X), max(X) FROM Speed;
┌─min(X)─┬────max(X)─┐
│ 0 │ 999999999 │
└────────┴───────────┘
1 rows in set. Elapsed: 1.040 sec. Processed 1.00 billion rows, 8.00 GB (961.32 million rows/s., 7.69 GB/s.)
SELECT quantileExact(0)(X), quantileExact(1)(X) FROM Speed;
┌─quantileExact(0)(X)─┬─quantileExact(1)(X)─┐
│ 0 │ 999999999 │
└─────────────────────┴─────────────────────┘
1 rows in set. Elapsed: 8.561 sec. Processed 1.00 billion rows, 8.00 GB (116.80 million rows/s., 934.43 MB/s.)
It turns out it is safe to use 0 and 1 values for quantileExact and quantilesExact functions.
We have modifier [with totals] that can summarize values across all rows and get the total result with key value=0 or null or smth like this
The problem is that I don't understand how I can use these values in the next calculations
Maybe I'm using the wrong format
select processing_date,count(*)
from `telegram.message`
where processing_date>='2019-05-01'
group by processing_date with totals
The documentation says that
You can use WITH TOTALS in subqueries, including subqueries in the
JOIN clause (in this case, the respective total values are combined).
Example subqueries in the JOIN (CH tests scripts in github):
SELECT k, s1, s2
FROM
(
SELECT intDiv(number, 3) AS k, sum(number) AS s1
FROM
(
SELECT *
FROM system.numbers
LIMIT 10
)
GROUP BY k WITH TOTALS
)
ANY LEFT JOIN
(
SELECT intDiv(number, 4) AS k, sum(number) AS s2
FROM
(
SELECT *
FROM system.numbers
LIMIT 10
)
GROUP BY k WITH TOTALS
) USING (k)
ORDER BY k ASC
/* Result:
┌─k─┬─s1─┬─s2─┐
│ 0 │ 3 │ 6 │
│ 1 │ 12 │ 22 │
│ 2 │ 21 │ 17 │
│ 3 │ 9 │ 0 │
└───┴────┴────┘
Totals:
┌─k─┬─s1─┬─s2─┐
│ 0 │ 45 │ 45 │
└───┴────┴────┘
*/
As a workaround, you can combine results of several totals using client libraries.
Using "with rollup" instead of "with totals" decides problems with format