In presto, I have rows of the form
name, var1, var2, var3
Foo, A, B, C
And because I need to group by var1, and also by var2 and var3 (each separately), I want to transform each row into three rows of the form:
name, key
Foo, var1=A
Foo, var2=B
Foo, var3=C
So that I can then just group by key. Presto doesn't have an UNPIVOT function, so any advice would be appreciated!
You can increase the number of rows by using a cross join to a subquery of the needed number of rows e.g.
select
t.name
, case when n.n = 1 then t.var1
when n.n = 2 then t.var2
when n.n = 3 then t.var3
end as key
from sourcetbl t
cross join (
select 1 as n union all
select 2 as n union all
select 3 as n
) n
and then a case expression may be used to reduce the number of columns based on the supplied "row number" (n.n in my example)
Related
SELECT DISTINCT
Member_ID,
CASE
WHEN a.ASTHMA_MBR = 1 THEN 'ASTHMA'
WHEN a.COPD_MBR = 1 THEN 'COPD'
WHEN a.HYPERTENSION_MBR = 1 THEN 'HYPERTENSION'
END AS DX_FLAG
So a member may have more than one, but my statement is only returning one of them.
I'm using Teradata and trying to convert multiple columns of boolean data into one column. The statement is only returning one condition when members may have 2 or more. I tried using Select instead of Select Distinct and it made no difference.
This is a kind of UNPIVOT:
with base_data as
( -- select the columns you want to unpivot
select
member_id
,date_col
-- the aliases will be the final column value
,ASTHMA_MBR AS ASTHMA
,COPD_MBR AS COPD
,HYPERTENSION_MBR AS HYPERTENSION
from your_table
)
,unpvt as
(
select member_id, date_col, x, DX_FLAG
from base_data
-- now unpivot those columns into rows
UNPIVOT(x FOR DX_FLAG IN (ASTHMA, COPD, HYPERTENSION)
) dt
)
select member_id, DX_FLAG, date_col
from unpvt
-- only show rows where the condition is true
where x = 1
I have a table with values "1", "0" or "". The table has four columns: p, q, r and s.
I need help creating a case statement that returns values when the attribute is equal to 1.
For ID 5 the case statement should return "p s".
For ID 14 the case statement should return "s".
For ID 33 the case statement should return 'p r s". And so on.
Do I need to come with a case statement that has every possible combination? Or is there a simpler way. Below is what I have come up with thus far.
case
when p = 1 and q =1 then "p q"
when p = 1 and r =1 then "p r"
when p = 1 and s =1 then "p s"
when r = 1 then r
when q = 1 then q
when r = 1 then r
when s = 1 then s
else ''
end
One solution could be this which uses a case for each attribute to return the correct value, surrounded by a trim to remove the trailing space.
with tbl(id, p, q, r, s) as (
select 5,1,0,0,1 from dual union all
select 14,0,0,0,1 from dual
)
select id,
trim(regexp_replace(case p when 1 then 'p' end ||
case q when 1 then 'q' end ||
case r when 1 then 'r' end ||
case s when 1 then 's' end, '(.)', '\1 '))
from tbl;
The real solution would be to fix the database design. This design technically violates Boyce-Codd 4th normal form in that it contains more than 1 independent attribute. The fact an ID "has" or "is part of" attribute p or q, etc should be split out. This design should be 3 tables, the main table with the ID, the lookup table containing info about attributes that the main ID could have (p, q, r or s) and the associative table that joins the two where appropriate (assuming an ID row could have more than one attribute and an attribute could belong to more than one ID), which is how to model a many-to-many relationship.
main_tbl main_attr attribute_lookup
ID col1 col2 main_id attr_id attr_id attr_desc
5 5 1 1 p
14 5 4 2 q
14 4 3 r
4 s
Then it would be simple to query this model to build your list, easy to maintain if an attribute description changes (only 1 place to change it), etc.
Select from it like this:
select m.ID, m.col1, listagg(al.attr_desc, ' ') within group (order by al.attr_desc) as attr_desc
from main_tbl m
join main_attr ma
on m.ID = ma.main_id
join attribute_lookup al
on ma.attr_id = al.attr_id
group by m.id, m.col1;
You can use concatenations with decode() functions
select id, decode(p,1,'p','')||decode(q,1,'q','')
||decode(r,1,'r','')||decode(s,1,'s','') as "String"
from t;
Demo
If you need spaces between letters, consider using :
with t(id,p,q,r,s) as
(
select 5,1,0,0,1 from dual union all
select 14,0,0,0,1 from dual union all
select 31,null,0,null,1 from dual union all
select 33,1,0,1,1 from dual
), t2 as
(
select id, decode(p,1,'p','')||decode(q,1,'q','')
||decode(r,1,'r','')||decode(s,1,'s','') as str
from t
), t3 as
(
select id, substr(str,level,1) as str, level as lvl
from t2
connect by level <= length(str)
and prior id = id
and prior sys_guid() is not null
)
select id, listagg(str,' ') within group (order by lvl) as "String"
from t3
group by id;
Demo
in my opinion, its a bad practice to use columns for relationships.
you should have two tables, one that's called arts and another that is called mapping art looks like this:
ID - ART
1 - p
2 - q
3 - r
4 - 2
...
and mapping maps your base-'ID's to your art-ids and looks like this
MYID - ARTID
5 - 1
5 - 4
afterwards, you should make use of oracles pivot operator. its more dynamically
I just had this idea of how can i loop in sql?
For example
I have this column
PARAMETER_VALUE
E,C;S,C;I,X;G,T;S,J;S,F;C,S;
i want to store all value before (,) in a temp column also store all value after (;) into another column
then it wont stop until there is no more value after (;)
Expected Output for Example
COL1 E S I G S S C
COL2 C C X T J F S
etc . . .
You can get by using regexp_substr() window analytic function with connect by level <= clause
with t1(PARAMETER_VALUE) as
(
select 'E,C;S,C;I,X;G,T;S,J;S,F;C,S;' from dual
), t2 as
(
select level as rn,
regexp_substr(PARAMETER_VALUE,'([^,]+)',1,level) as str1,
regexp_substr(PARAMETER_VALUE,'([^;]+)',1,level) as str2
from t1
connect by level <= regexp_count(PARAMETER_VALUE,';')
)
select listagg( regexp_substr(str1,'([^;]+$)') ,' ') within group (order by rn) as col1,
listagg( regexp_substr(str2,'([^,]+$)') ,' ') within group (order by rn) as col2
from t2;
COL1 COL2
------------- -------------
E S I G S S C C C X T J F S
Demo
Assuming that you need to separate the input into rows, at the ; delimiters, and then into columns at the , delimiter, you could do something like this:
-- WITH clause included to simulate input data. Not part of the solution;
-- use actual table and column names in the SELECT statement below.
with
t1(id, parameter_value) as (
select 1, 'E,C;S,C;I,X;G,T;S,J;S,F;C,S;' from dual union all
select 2, ',U;,;V,V;' from dual union all
select 3, null from dual
)
-- End of simulated input data
select id,
level as ord,
regexp_substr(parameter_value, '(;|^)([^,]*),', 1, level, null, 2) as col1,
regexp_substr(parameter_value, ',([^;]*);' , 1, level, null, 1) as col2
from t1
connect by level <= regexp_count(parameter_value, ';')
and id = prior id
and prior sys_guid() is not null
order by id, ord
;
ID ORD COL1 COL2
--- --- ---- ----
1 1 E C
1 2 S C
1 3 I X
1 4 G T
1 5 S J
1 6 S F
1 7 C S
2 1 U
2 2
2 3 V V
3 1
Note - this is not the most efficient way to split the inputs (nothing will be very efficient - the data model, which is in violation of First Normal Form, is the reason). This can be improved using standard instr and substr, but the query will be more complicated, and for that reason, harder to maintain.
I generated more input data, to illustrate a few things. You may have several inputs that must be broken up at the same time; that must be done with care. (Note the additional conditions in CONNECT BY). I also illustrate the handling of NULL - if a comma comes right after a semicolon, that means that the "column 1" part of that pair must be NULL. That is shown in the output.
I am using WinSQL to run a query on a table to count the number of occurrences of literal strings. When trying to do a count on a specific set of strings, I still want to see if some values return a count of 0. For example:
select letter, count(*)
from table
where letter in ('A', 'B', 'C')
group by letter
Let's say we know that 'A' occurs 3 times, 'B' occurs 0 times, and 'C' occurs 5 times. I expect to have a table returned as such:
letter count
A 3
B 0
C 5
However, the table never returns a row with a 0 count, which results like so:
letter count
A 3
C 5
I've looked around and saw some articles mentioning the use of joins, but I've had no luck in correctly returning a table that looks like the first example.
You can create an in-line table containing all letters that you look for, then LEFT JOIN your table to it:
select t1.col, count(t2.letter)
from (
select 'A' AS col union all select 'B' union all select 'C'
) as t1
left join table as t2 on t1.col = t2.letter
group by t1.col
on many platforms you can now use the values statement instead of union all to create your "in line" table - like this
select t.letter, count(mytable.letter)
from ( values ('A'),('B'),('C') ) as t(letter)
left join mytable on t.letter = mytable.letter
group by t.letter
I'm not that familiar with WinSQL, but it's not pretty if you don't have the values that you want in the left most column in a table somewhere. If you did, you could use a left join and a conditional. Without it, you can do something like this:
SELECT all_letters.letter, IFNULL(letter_count.letter_count, 0)
FROM
(
SELECT 'A' AS letter
UNION
SELECT 'B' AS letter
UNION
SELECT 'C' AS letter
) all_letters
LEFT JOIN
(SELECT letter, count(*) AS letter_count
FROM table
WHERE letter IN ('A', 'B', 'C')
GROUP BY letter) letter_count
ON all_letters.letter = letter_count.letter
Given the following SQL Server table with a single char(1) column:
Value
------
'1'
'2'
'3'
How do I obtain the following results in T-SQL?
Result
------
'1+2+3'
'1+3+2'
'2+1+3'
'2+3+1'
'3+2+1'
'3+1+2'
This needs to be dynamic too, so if my table only holds rows '1' and '2' I'd expect:
Result
------
'1+2'
'2+1'
It seems like I should be able to use CROSS JOIN to do this, but since I don't know how many rows there will be ahead of time, I'm not sure how many times to CROSS JOIN back on myself..?
SELECT a.Value + '+' + b.Value
FROM MyTable a
CROSS JOIN MyTable b
WHERE a.Value <> b.Value
There will always be less than 10 (and really more like 1-3) rows at any given time. Can I do this on-the-fly in SQL Server?
Edit: ideally, I'd like this to happen in a single stored proc, but if I have to use another proc or some user defined functions to pull this off I'm fine with that.
This SQL will compute the permutations without repetitions:
WITH recurse(Result, Depth) AS
(
SELECT CAST(Value AS VarChar(100)), 1
FROM MyTable
UNION ALL
SELECT CAST(r.Result + '+' + a.Value AS VarChar(100)), r.Depth + 1
FROM MyTable a
INNER JOIN recurse r
ON CHARINDEX(a.Value, r.Result) = 0
)
SELECT Result
FROM recurse
WHERE Depth = (SELECT COUNT(*) FROM MyTable)
ORDER BY Result
If MyTable contains 9 rows, it will take some time to compute, but it will return 362,880 rows.
Update with explanation:
The WITH statement is used to define a recursive common table expression. In effect, the WITH statement is looping multiple times performing a UNION until the recursion is finished.
The first part of SQL sets the starting records. Assuming 3 rows named 'A', 'B', and 'C' in MyTable, this will generate these rows:
Result Depth
------ -----
A 1
B 1
C 1
Then the next block of SQL performs the first level of recursion:
SELECT CAST(r.Result + '+' + a.Value AS VarChar(100)), r.Depth + 1
FROM MyTable a
INNER JOIN recurse r
ON CHARINDEX(a.Value, r.Result) = 0
This takes all of the records generated so far (which will be in the recurse table) and joins them to all of the records in MyTable again. The ON clause filters the list of records in MyTable to only return the ones that do not exist already in this row's permutation. This would result in these rows:
Result Depth
------ -----
A 1
B 1
C 1
A+B 2
A+C 2
B+A 2
B+C 2
C+A 2
C+B 2
Then the recursion loops again giving these rows:
Result Depth
------ -----
A 1
B 1
C 1
A+B 2
A+C 2
B+A 2
B+C 2
C+A 2
C+B 2
A+B+C 3
A+C+B 3
B+A+C 3
B+C+A 3
C+A+B 3
C+B+A 3
At this point, the recursion stops because the UNION does not create any more rows because the CHARINDEX will always be 0.
The last SQL filters all of the resulting rows where the computed Depth column matches the # of records in MyTable. This throws out all of the rows except for the ones generated by the last depth of recursion. So the final result will be these rows:
Result
------
A+B+C
A+C+B
B+A+C
B+C+A
C+A+B
C+B+A
You can do this with a recursive CTE:
with t as (
select 'a' as value union all
select 'b' union all
select 'c'
),
const as (select count(*) as cnt from t),
cte as (
select cast(value as varchar(max)) as value, 1 as level
from t
union all
select cte.value + '+' + t.value, 1 + level
from cte join
t
on '+'+cte.value+'+' not like '%+'+t.value+'+%' cross join
const
where level <= const.cnt
)
select cte.value
from cte cross join
const
where level = const.cnt;