Snowflake fill null values - snowflake-cloud-data-platform

I have a table like this in Snowflake, where I have a "class" column, where all rows with level 1 have a value. I would like to have a new column "WANTED_OUTPUT", where the value from class are filled until new value occurs, then fill with that value.
I have been looking at first_value and last_value function, but I miss something there can "group" all the rows from level 1 until next level 1 together before I can use the first_value and the partition over that.
Any suggestion?
+----+-------+-------+-------+---------------+
| id | col_c | level | class | WANTED_OUTPUT |
+----+-------+-------+-------+---------------+
| a | q1 | 1 | c99 | c99 |
+----+-------+-------+-------+---------------+
| a | w2 | 2 | NULL | c99 |
+----+-------+-------+-------+---------------+
| a | g6 | 2 | NULL | c99 |
+----+-------+-------+-------+---------------+
| a | j5 | 3 | NULL | c99 |
+----+-------+-------+-------+---------------+
| a | x8 | 1 | c3 | c3 |
+----+-------+-------+-------+---------------+
| a | x9 | 2 | NULL | c3 |
+----+-------+-------+-------+---------------+
| a | h5 | 1 | c67 | c67 |
+----+-------+-------+-------+---------------+

Using FIRST/LAST_VALUE:
SELECT *,
FIRST_VALUE(class) IGNORE NULLS OVER (PARTITION BY id ORDER BY level) AS wanted
FROM tab;
Tables are unordered sets by design so stable sort is required. Level may not be enough based on provided input.
Suggestion: Adding explicit timestamp or seq id column to provide stable sort column would be preferred.
SELECT tab.*,
LAST_VALUE(class) IGNORE NULLS OVER (PARTITION BY id
ORDER BY rn
RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW
) AS wanted
FROM tab;
db<>fiddle demo

Related

What's an efficient way to count "previous" rows in SQL?

Hard to phrase the title for this one.
I have a table of data which contains a row per invoice. For example:
| Invoice ID | Customer Key | Date | Value | Something |
| ---------- | ------------ | ---------- | ------| --------- |
| 1 | A | 08/02/2019 | 100 | 1 |
| 2 | B | 07/02/2019 | 14 | 0 |
| 3 | A | 06/02/2019 | 234 | 1 |
| 4 | A | 05/02/2019 | 74 | 1 |
| 5 | B | 04/02/2019 | 11 | 1 |
| 6 | A | 03/02/2019 | 12 | 0 |
I need to add another column that counts the number of previous rows per CustomerKey, but only if "Something" is equal to 1, so that it returns this:
| Invoice ID | Customer Key | Date | Value | Something | Count |
| ---------- | ------------ | ---------- | ------| --------- | ----- |
| 1 | A | 08/02/2019 | 100 | 1 | 2 |
| 2 | B | 07/02/2019 | 14 | 0 | 1 |
| 3 | A | 06/02/2019 | 234 | 1 | 1 |
| 4 | A | 05/02/2019 | 74 | 1 | 0 |
| 5 | B | 04/02/2019 | 11 | 1 | 0 |
| 6 | A | 03/02/2019 | 12 | 0 | 0 |
I know I can do this using either a CTE like this...
(
select
count(*)
from table
where
[Customer Key] = t.[Customer Key]
and [Date] < t.[Date]
and Something = 1
)
But I have a lot of data and that's pretty slow. I know I can also use cross apply to achieve the same thing, but as far as I can tell that's not any better performing than just using a CTE.
So; is there a more efficient means of achieving this, or do I just suck it up?
EDIT: I originally posted this without the requirement that only rows where Something = 1 are counted. Mea culpa - I asked it in a hurry. Unfortunately I think that this means I can't use row_number() over (partition by [Customer Key])
Assuming you're using SQL Server 2012+ you can use Window Functions:
COUNT(CASE WHEN Something = 1 THEN CustomerKey END) OVER (PARTITION BY CustomerKey ORDER BY [Date]
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) -1 AS [Count]
Old answer before new required logic:
COUNT(CustomerKey) OVER (PARTITION BY CustomerKey ORDER BY [Date]
ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) -1 AS [Count]
If you're not using 2012 an alternative is to use ROW_NUMBER
ROW_NUMBER() OVER (PARTITION BY CustomerKey ORDER BY [Date]) - 1 AS Count

Selecting grouped rows after first two rows SQL Server

This is a bit of a tricky question/situation and my search fu failed me.
Lets say i have the following data
| UID | SharedID | Type | Date |
|-----|----------|------|-----------|
| 1 | 1 | foo | 2/4/2016 |
| 2 | 1 | foo | 2/5/2016 |
| 3 | 1 | foo | 2/8/2016 |
| 4 | 1 | foo | 2/11/2016 |
| 5 | 2 | bar | 1/11/2016 |
| 6 | 2 | bar | 2/11/2016 |
| 7 | 3 | baz | 2/1/2016 |
| 8 | 3 | baz | 2/3/2016 |
| 9 | 3 | baz | 2/11/2016 |
And I would like to ommit a variable number of leading rows (most recent date in this case) and lets say that number is 2 in this example. The resulting table would be something like this:
| UID | SharedID | Type | Date |
|-----|----------|------|-----------|
| 1 | 1 | foo | 2/4/2016 |
| 2 | 1 | foo | 2/5/2016 |
| 7 | 3 | baz | 2/1/2016 |
Is this possible in SQL? Essentially I want to filter on an unknown number of rows which uses the date column as the order by. The goal is to get the oldest types and get a list of UID's in the process.
Sure, it's possible. Use a ROW_NUMBER function to assign a value to each row, partitioning by the SharedID column so that the count restarts every time that ID changes, and select those rows with a value greater than your limit.
WITH cteNumberedRows AS (
SELECT UID, SharedID, Type, Date,
ROW_NUMBER() OVER(PARTITION BY SharedID ORDER BY Date DESC) AS RowNum
FROM YourTable
)
SELECT UID, SharedID, Type, Date
FROM cteNumberedRows
WHERE RowNum > 2;
Not sure if I understand what you mean but something like this?
SELECT * FROM MyTable t1 JOIN MyTable T2 ON t2.id NOT IN (
SELECT TOP 2 UID FROM myTable
WHERE SharedID = t1.sharedID
ORDER BY [Date] DESC
)

SQL Server : update sequence number across multiple groups

I would like to update a table:
| id | type_id | created_at | sequence |
|----|---------|------------|----------|
| 1 | 1 | 2010-04-26 | NULL |
| 2 | 1 | 2010-04-27 | NULL |
| 3 | 2 | 2010-04-28 | NULL |
| 4 | 3 | 2010-04-28 | NULL |
To this (note that created_at is used for ordering, and sequence is "grouped" by type_id):
| id | type_id | created_at | sequence |
|----|---------|------------|----------|
| 1 | 1 | 2010-04-26 | 1 |
| 2 | 1 | 2010-04-27 | 2 |
| 3 | 2 | 2010-04-28 | 1 |
| 4 | 3 | 2010-04-28 | 1 |
Same question has been raised but for SQL Server.
Link
Thanks.
You can use ROW_NUMBER() to get sequence number per type_id slice. Use a CTE to make UPDATE operation simpler:
;WITH ToUpdate AS (
SELECT id, type_id, created_at, sequence,
ROW_NUMBER() OVER (PARTITION BY type_id ORDER BY created_at) AS newSeq
FROM mytable
)
UPDATE ToUpdate
SET sequence = newSeq
Demo here

Rearranging and deduplicating SQL columns based on column data

Sorry I know that's a rubbish Title but I couldn't think of a more concise way of describing the issue.
I have a (MSSQL 2008) table that contains telephone numbers:
| CustomerID | Tel1 | Tel2 | Tel3 | Tel4 | Tel5 | Tel6 |
| Cust001 | 01222222 | 012333333 | 07111111 | 07222222 | 01222222 | NULL |
| Cust002 | 07444444 | 015333333 | 07555555 | 07555555 | NULL | NULL |
| Cust003 | 01333333 | 017777777 | 07888888 | 07011111 | 016666666 | 013333 |
I'd like to:
Remove any duplicate phone numbers
Rearrange the telephone numbers so that anything beginning with "07" is the first phone number. If there are multiple 07's, they should be in the first fields. The order of the numbers apart from that doesn't really matter.
So, for example, after processing, the table would look like:
| CustomerID | Tel1 | Tel2 | Tel3 | Tel4 | Tel5 | Tel6 |
| Cust001 | 07111111 | 07222222 | 01222222 | 012333333 | NULL | NULL |
| Cust002 | 07444444 | 07555555 | 015333333 | NULL | NULL | NULL |
| Cust003 | 07888888 | 07011111 | 016666666 | 013333 | 01333333 | 017777777 |
I'm struggling to figure out how to efficiently achieve my goal (there are 600,000+ records in the table). Can anyone help?
I've created a fiddle if it'll help anyone play around with the scenario.
You can break up the numbers into individual rows using UNPIVOT, then reorder them based on the occurence of the '07' prefix using ROW_NUMBER(), and finally recombine it using PIVOT to end up with the 6 Tel columns again.
select *
FROM
(
select CustomerID, Col, Tel
FROM
(
select *, Col='Tel' + RIGHT(
row_number() over (partition by CustomerID
order by case
when Tel like '07%' then 1
else 2
end),10)
from phonenumbers
UNPIVOT (Tel for Seq in (Tel1,Tel2,Tel3,Tel4,Tel5,Tel6)) seqs
) U
) P
PIVOT (MAX(TEL) for Col IN (Tel1,Tel2,Tel3,Tel4,Tel5,Tel6)) V;
SQL Fiddle
Perhaps using cursor to collect all customer id and sorting the fields...traditional sorting technique as we used to do in school c++ ..lolz...like to know if any other method possible.
If you dont get any then it is the last way . It will take a long time for sure to execute.

Finding the max and min date values in SQL Server tables

I have two tables:
A lookup table (tabOne):
KEY | Group | Name | Desc | Val_Key
----------------------------------------
1 | a | NameA | DescA | 10
2 | b | NameB | DescB | 20
3 | c | NameC | DescC | 30
4 | d | NameD | DescD | 40
5 | e | NameE | DescE | 50
6 | f | NameF | DescF | 60
A second table containing readings (tabTwo):
KEY | Date | Reading | Val_Key
----------------------------------------
1 | Date | Read | 10
2 | Date | Read | 20
3 | Date | Read | 40
4 | Date | Read | 40
5 | Date | Read | 30
6 | Date | Read | 20
7 | Date | Read | 40
8 | Date | Read | 20
9 | Date | Read | 10
10 | Date | Read | 20
11 | Date | Read | 50
12 | Date | Read | 60
What I need to do is join tabTwo with TabOne and create a column with the newest Reading and a column with the oldest reading for each item in the group column of TabOne.
At the end of the day I want a table that look as follow:
KEY | Group | Name | Desc | Val_Key | LastReading | FirstReading |
-------------------------------------------------------------------------
1 | a | NameA | DescA | 10 | | |
2 | b | NameB | DescB | 20 | | |
3 | c | NameC | DescC | 30 | | |
4 | d | NameD | DescD | 40 | | |
5 | e | NameE | DescE | 50 | | |
6 | f | NameF | DescF | 60 | | |
Thanks!
Freddie
If this is Sql Server 2005 or newer, outer apply will help:
select TabOne.*,
last.Reading LastReading,
first.Reading FirstReading
from TabOne
outer apply
(
select top 1
Reading
from TabTwo
where TabTwo.Val_Key = TabOne.val_Key
order by TabTwo.Date desc
) last
outer apply
(
select top 1
Reading
from TabTwo
where TabTwo.Val_Key = TabOne.val_Key
order by TabTwo.Date asc
) first
Live test is # Sql Fiddle.
#Nikola Markovinović's solution can be made more universally applicable if the subqueries are moved directly to the main query's SELECT clause, which is possible each of them retrieves only one value and is, therefore, valid as a scalar expression:
SELECT
t1.[KEY],
t1.[Group],
t1.Name,
t1.[Desc],
t1.Val_Key,
(
SELECT TOP 1 Reading
FROM TabTwo
WHERE Val_Key = t1.Val_Key
ORDER BY Date DESC
) AS LastReading,
(
SELECT TOP 1 Reading
FROM TabTwo
WHERE Val_Key = t1.Val_Key
ORDER BY Date ASC
) AS FirstReading
FROM TabOne t1
If you needed e.g. dates along the way, you would probably have to stick to Nikola's solution. There is an alternative to it, but it's more cumbersome (albeit more standard too): it would involve grouping TabTwo's data by Val_Key to get earliest/latest dates per Val_Key, then joining back to TabTwo to access entire rows corresponding to the found dates to finally pull the necessary columns, and ultimately joining both result sets to TabOne to get the final column set.

Resources