How to achieve NOT EXISTS with the ABAP CDS? - abap

I would like to select all EKPO records that don't have an existing posting in MSEG table.
In ABAP SQL this can be done like below:
SELECT ebeln, ebelp FROM ekpo INTO TABLE #DATA(orders)
WHERE NOT EXISTS ( SELECT ebeln FROM mseg
WHERE ebeln = ekpo~ebeln
AND ebelp = ekpo~ebelp ).
The only solution I found is to create 2 CDS views, the first one to select all the orders with a record in MSEG and the second one being the negation of the first. But I would expect a cleaner solution so I wanted to ask here.

Here's how we do it:
define view my_view as
select from ekpo
association[0..1] to mseg
on mseg.ebeln = ekpo.ebeln
and mseg.ebelp = ekpo.ebelp
{
ebeln,
ebelp
}
where mseg.mandt is null
If no entry matching the criteria exists in mseg, all fields of the association will be null. Otherwise, mseg.mandt will never be null.

Related

Select from where in ( ) table?

I want a select with dynamic where conditions in ABAP Syntax.
An SQL Statement would look like this:
SELECT * FROM MCH1 WHERE MATNR IN (...) AND CHARG IN (...)
My approach was to add 2 structures ZMATN_STR and ZCHARG_STR to the dictionary with associated components as line (MATNR, CHARG).
Then create 2 table types with associated line types.
Now im stucked in ABAP because I don't know how to write the where clause.
That's what I have so far:
SELECT *
FROM
mch1
FOR ALL ENTRIES IN #matnrs
WHERE
matnr = #matnrs-matnr
INTO TABLE #DATA(lt_result).
It works for either matnr or charg but not with both of them.
Additional Info
This select happens in a function module where 2 import parameter exists (the 2 table types) - so I cannot just write where in ('xxx', 'yyy')
data lr_matnr type range of matnr.
data lr_charg type range of MCH1-charg.
"Fill lr_matnr and lr_charg...
SELECT * FROM MCH1 WHERE MATNR IN #lr_matnr AND CHARG IN #lr_charg
INTO TABLE #data(lt_result).

Struggling with a complicated query on row-based Field/Value table

Bare with me for a little bit of setup here please.
I have a table MAIN that has a Field/Value representation that looks like this:
I have another table called STORE_FLAG:
I am trying to write a parameterized query for which I will be given one FIELD_ID and one or more IDs from the STORE_FLAG table.
What I need to do is select from the MAIN table ROW_IDs where:
for the given FIELD_ID, the VALUE = 'YES' AND
for the given STORE_FLAG_IDS, ANY of those FIELD_IDs correspond to a VALUE = 'x' in the MAIN table.
Not that this would be a good idea, but I cannot pivot the whole table into a column-based table to then do a traditional where clause.
Example:
Given a Field_Id = 1 and a list of StoreIds = (30,50). I would want to return row_ids 1 and 2. This is because row_id 1 and 2 have a field_id 1 with value 'YES' AND at least one of the field_ids 3 and 5 have a value 'x'. But row_id 3 has a value of null for both field_id 3 and 5 and row_id 4 has a field_id 1 with value = 'NO'.
I was thinking something like this:
SELECT DISTINCT ROW_ID FROM MAIN
WHERE (FIELD_ID = :providedFieldId OR FIELD_ID IN (SELECT FIELD_ID FROM STORE_FLAG WHERE ID IN :providedStoreIdList))
AND (FIELD_VALUE = 'YES' OR FIELD_VALUE = 'x')
which (I think) works, but feels naïve to me..? I feel like there is some sort of super duper grouping way to do this, but I can't wrap my head around it. Any suggestions would be really appreciated.
here is a way to do this
select distinct m.row_id
from main m
where m.field_id=:providedFieldId
and m.field_value='YES'
and exists (select 1
from STORE_FLAG sf
join main m2
on sf.field_id=m2.field_id
where sf.id in ('30','50') /* you need to bind the values from :providedStoreIdList using a table function*/
and m2.field_value='x'
and m2.row_id=m.row_id
)
link on how to bind an in list
https://asktom.oracle.com/pls/apex/f?p=100:11:0::::P11_QUESTION_ID:110612348061
Your provided solution /query will not work as you say. Because in your last line of query [AND (FIELD_VALUE = 'YES' OR FIELD_VALUE = 'x')] conflict with your requirement . Using your query, you will get ROW_ID if any one is true either FIELD_VALUE = 'YES' OR FIELD_VALUE = 'x'. Which is wrong. You can see below query-
SELECT SUB_QUERY.ROW_ID FROM
(
select DISTINCT MAIN.ROW_ID,MAIN.FIELD_VALUE from STORE_FLAG
RIGHT OUTER JOIN MAIN ON STORE_FLAG.FIELD_ID=MAIN.FIELD_ID
WHERE ((STORE_FLAG.ID IN ('202','203') AND MAIN.FIELD_VALUE='x')
OR (MAIN.FIELD_ID ='1' AND MAIN.FIELD_VALUE='YES'))
) SUB_QUERY
GROUP BY SUB_QUERY.ROW_ID
HAVING (LISTAGG(SUB_QUERY.FIELD_VALUE, ',') WITHIN GROUP (ORDER BY SUB_QUERY.ROW_ID) IN ('YES,x','x,YES'))
I think you need to run and understand my sub query part at first.

Behavior of IN with empty subquery

I want to select every customer from the table KNA1 which does not have any entry in table KNBK.
So I made a select query which should do this for me:
SELECT-OPTIONS: s_kn_nr FOR kna1-kunnr.
SELECT-OPTIONS: s_kn_okd FOR kna1-ktokd.
SELECT * FROM kna1
INTO TABLE #DATA(ls_kna1)
WHERE kunnr IN #s_kn_nr
AND ktokd IN #s_kn_okd
and kunnr not in ( select kunnr from knbk where kunnr in #s_kr_nr )
Now my question is, does the select query handle empty subselects in the same way as empty select options and just select everything?
The answer is simple: no.
Such a subquery returns a set and IN functions here exactly as the mathematical operation of belonging to set or no. So if the set is empty then no element belongs to it.

I need to create a VIEW from an existing TABLE and MAP an additional COLUMN to that VIEW

I am fairly new to SQL. What I am trying to do is create a view from an existing table. I also need to add a new column to the view which maps to the values of an existing column in the table.
So within the view, if the value in a field for Col_1 = A, then the value in the corresponding row for New_Col = C etc
Does this even make sense? Would I use the CASE clause? Is mapping in this way even possible?
Thanks
The best way to do this is to create a mapping or lookup table
For example consider the following LOOKUP table.
COL_A NEW_VALUE
---- -----
A C
B D
Then you can have a query like this:
SELECT A.*, LOOK.NEW_VALUE
FROM TABLEA AS A
JOIN LOOKUP AS LOOK ON A.COL_A = LOOK.COL_A
This is what DimaSUN is doing in his query too -- but in his case he is creating the table dynamically in the body of the query.
Also note, I'm using a JOIN (which is an inner join) so only results in the lookup table will be returned. This could filter the results. A LEFT JOIN there would return all data from A but some of the new columns might be null.
Generally, a view is an instance of a table/a replica provided that there is no alteration to the original table. So, as per your query you can manipulate the data and columns in a view by using case.
Create View viewname as
Select *,
case when column=a.value then 'C'
....
ELSE
END
FROM ( Select * from table) a
If You have restricted list of replaced values You may hardcode that list in query
select T.*,map.New_Col
from ExistingTable T
left join (
values
('A','C')
,('B','D')
) map (Col_1,New_Col) on map.Col_1 = T.Col_1
In this sample You hardcode 'A' -> 'C' and 'B' -> 'D'
In general case You better may to use additional table ( see Hogan answer )

I want to join two tables with a common column in Big query?

To join the tables, I am using the following query.
SELECT *
FROM(select user as uservalue1 FROM [projectname.FullData_Edited]) as FullData_Edited
JOIN (select user as uservalue2 FROM [projectname.InstallDate]) as InstallDate
ON FullData_Edited.uservalue1=InstallDate.uservalue2;
The query works but the joined table only has two columns uservalue1 and uservalue2.
I want to keep all the columns present in both the table. Any idea how to achieve that?
#legacySQL
SELECT <list of fields to output>
FROM [projectname:datasetname.FullData_Edited] AS FullData_Edited
JOIN [projectname:datasetname.InstallDate] AS InstallDate
ON FullData_Edited.user = InstallDate.user
or (and preferable)
#standardSQL
SELECT <list of fields to output>
FROM `projectname.datasetname.FullData_Edited` AS FullData_Edited
JOIN `projectname.datasetname.InstallDate` AS InstallDate
ON FullData_Edited.user = InstallDate.user
Note, using SELECT * in such cases lead to Ambiguous column name error, so it is better to put explicit list of columns/fields you need to have in your output
The way around it is in using USING() syntax as in example below.
Assuming that user is the ONLY ambiguous field - it does the trick
#standardSQL
SELECT *
FROM `projectname.datasetname.FullData_Edited` AS FullData_Edited
JOIN `projectname.datasetname.InstallDate` AS InstallDate
USING (user)
For example:
#standardSQL
WITH `projectname.datasetname.FullData_Edited` AS (
SELECT 1 user, 'a' field1
),
`projectname.datasetname.InstallDate` AS (
SELECT 1 user, 'b' field2
)
SELECT *
FROM `projectname.datasetname.FullData_Edited` AS FullData_Edited
JOIN `projectname.datasetname.InstallDate` AS InstallDate
USING (user)
returns
user field1 field2
1 a b
whereas using ON FullData_Edited.user = InstallDate.user gives below error
Error: Duplicate column names in the result are not supported. Found duplicate(s): user
Don't use subqueries if you want all columns:
SELECT *
FROM [projectname.FullData_Edited] as FullData_Edited JOIN
[projectname.InstallDate] as InstallDate
ON FullData_Edited.uservalue1 = InstallDate.uservalue2;
You may have to list out the particular columns you want to avoid duplicate column names.
While you are at it, you should also switch to standard SQL.