I need the following, Can anyone please help me do it.
Rank Cust_Type Cust_Name Revenue
1 Top A 10000
2 Top B 9000
3 Top C 8000
1 Bottom X 5000
2 Bottom Y 6000
3 Bottom Z 7000
I need separate ranks for Top and Bottom Cust_Type and all this is in MySQL.
This is a bit tricky. You may want to use variables, such as in the following example:
SELECT (
CASE cust_type
WHEN #curType
THEN #curRow := #curRow + 1
ELSE #curRow := 1 AND #curType := cust_type END
) + 1 AS rank,
cust_type,
cust_name,
revenue
FROM sales,
(SELECT #curRow := 0, #curType := '') r
ORDER BY cust_type DESC, revenue DESC;
The (SELECT #curRow := 0, #curType := '') r part allows the variable initialization without requiring a separate SET command.
Test case:
CREATE TABLE sales (cust_type varchar(10), cust_name varchar(10), revenue int);
INSERT INTO sales VALUES ('Top', 'A', 10000);
INSERT INTO sales VALUES ('Top', 'B', 9000);
INSERT INTO sales VALUES ('Top', 'C', 8000);
INSERT INTO sales VALUES ('Bottom', 'X', 5000);
INSERT INTO sales VALUES ('Bottom', 'Y', 6000);
INSERT INTO sales VALUES ('Bottom', 'Z', 7000);
Result:
+------+-----------+-----------+---------+
| rank | cust_type | cust_name | revenue |
+------+-----------+-----------+---------+
| 1 | Top | A | 10000 |
| 2 | Top | B | 9000 |
| 3 | Top | C | 8000 |
| 1 | Bottom | Z | 7000 |
| 2 | Bottom | Y | 6000 |
| 3 | Bottom | X | 5000 |
+------+-----------+-----------+---------+
6 rows in set (0.00 sec)
Another test case:
CREATE TABLE sales (cust_type varchar(10), cust_name varchar(10), revenue int);
INSERT INTO sales VALUES ('Type X', 'A', 7000);
INSERT INTO sales VALUES ('Type X', 'B', 8000);
INSERT INTO sales VALUES ('Type Y', 'C', 5000);
INSERT INTO sales VALUES ('Type Y', 'D', 6000);
INSERT INTO sales VALUES ('Type Y', 'E', 4000);
INSERT INTO sales VALUES ('Type Z', 'F', 4000);
INSERT INTO sales VALUES ('Type Z', 'G', 3000);
Result:
+------+-----------+-----------+---------+
| rank | cust_type | cust_name | revenue |
+------+-----------+-----------+---------+
| 1 | Type Z | F | 4000 |
| 2 | Type Z | G | 3000 |
| 1 | Type Y | D | 6000 |
| 2 | Type Y | C | 5000 |
| 3 | Type Y | E | 4000 |
| 1 | Type X | B | 8000 |
| 2 | Type X | A | 7000 |
+------+-----------+-----------+---------+
7 rows in set (0.00 sec)
You can obviously order the cust_type in ascending order instead of descending. I used descending just to have Top before Bottom in the original test case.
I found a problem with the solution using CASE, #curRow, and #curType. It depends on the execution plan MySQL uses to process the query. For example, it shows up if you add a join to the query. Then there is no guarantee that the rank is going to be computed correctly.
Making a slight change to the answer:
CREATE TABLE sales (cust_type_id int, cust_name varchar(10), revenue int);
CREATE TABLE cust_type (cust_type_id int, type_name varchar(10));
INSERT INTO cust_type VALUES (1, 'Bottom');
INSERT INTO cust_type VALUES (2, 'Top');
INSERT INTO sales VALUES (2, 'A', 10000);
INSERT INTO sales VALUES (2, 'B', 9000);
INSERT INTO sales VALUES (2, 'C', 8000);
INSERT INTO sales VALUES (1, 'X', 5000);
INSERT INTO sales VALUES (1, 'Y', 6000);
INSERT INTO sales VALUES (1, 'Z', 7000);
If I query only the sales table I get the rank in the correct order, but if I join to the cust_type table the rank values are no longer correct
SELECT (
CASE s.cust_type_id
WHEN #curType
THEN #curRow := #curRow + 1
ELSE #curRow := 1 AND #curType := s.cust_type_id END
) AS rank,
t.type_name,
s.cust_name,
s.revenue
FROM sales s,
cust_type t,
(SELECT #curRow := 0, #curType := 0) r
WHERE s.cust_type_id = t.cust_type_id
ORDER BY t.type_name DESC, s.revenue DESC;
Result:
+------+-----------+-----------+---------+
| rank | type_name | cust_name | revenue |
+------+-----------+-----------+---------+
| 1 | Top | A | 10000 |
| 2 | Top | B | 9000 |
| 3 | Top | C | 8000 |
| 3 | Bottom | Z | 7000 |
| 2 | Bottom | Y | 6000 |
| 1 | Bottom | X | 5000 |
+------+-----------+-----------+---------+
MySQL is running the initial query into a temp table and then the ORDER BY is executing against the temp table after a rank was already computed.
This is similar to Thomas's answer, but a little simpler:
SELECT (SELECT COUNT(Cust_Type) FROM sales
WHERE Cust_Type = S.Cust_Type AND Revenue >= S.Revenue) AS Rank,
Cust_Type,
Cust_Name,
Revenue
FROM sales AS S
ORDER BY Cust_Type DESC, Rank;
I'm not sure how the performance compares to the Daniel's solution, particularly on very large data sets, or if you have to use complex joins.
What is not exactly clear is how the items should be ranked (I assumed by Revenue) or whether you are only pulling a certain number of values (e.g. the top 3 and the bottom 3) so I assumed you wanted all values. Given those assumptions,
Select Cust_Name, Cust_Type
, (Select Count(*)
From Table As T1
Where T1.Revenue > T.Revenue ) + 1 As Rank
From Table As T
Where Cust_Type = 'Top'
Union All
Select Cust_Name, Cust_Type
, (Select Count(*)
From Table As T1
Where T1.Revenue < T.Revenue ) + 1 As Rank
From Table As T
Where Cust_Type = 'Bottom'
If you were trying to do this in a single non-union query you could do:
Select Cust_Name, Cust_Type
, Case Z.Cust_Type
When 'Top' Then Z.TopRank
Else Z.BottomRank
End As Rank
From (
Select Cust_Name, Cust_Type
, (Select Count(*)
From Table As T1
Where T1.Revenue > T.Revenue ) + 1 As TopRank
, (Select Count(*)
From Table As T1
Where T1.Revenue < T.Revenue ) + 1 As BottomRank
From Table As T
) As Z
This works for me by keeping ranking of sales revenue and ordering separate .
SELECT
(Select count(s1.revenue)+1 from sales s1 where s.cust_type_id = s1.cust_type_id and s1.revenue > s.revenue)
As rank,
t.type_name,
s.cust_name,
s.revenue
FROM sales s LEFT JOIN
cust_type t USING(cust_type_id)
Group by t.type_name,s.cust_name,s.revenue DESC order by s.revenue DESC;
For JOIN tables problem, I found a solution.
I create a temporary table, this way MySQL maintain the order of value that I want to rank.
DROP TEMPORARY TABLE IF EXISTS tmp_mytable;
CREATE TEMPORARY TABLE tmp_mytable ENGINE = MEMORY
SELECT mytable.id AS id,
mytable.login AS Login,
cliente.myrank_id AS id_myrank,
mytable.rankvalue AS rankvalue
FROM mytable
INNER JOIN myjoin ON (mytable.id_myjoin = myjoin.id)
ORDER BY 3, 4 DESC;
SELECT id, login, IFNULL(id_myrank, 0) AS id_myrank, rankvalue,
#rank := IF(#prev_born = IFNULL(id_myrank, 0), #rank + 1, 1) AS ranking,
#prev_Born := IFNULL(id_myrank, 0) AS fake_field
FROM tmp_mytable, (select #prev_born := 0, #rank := 0) r
-- HAVING ranking < 20;
*PS: I tried with View creation, but insn't work too.
Here are solutions for MySQL 5.7 and MySQL 8.0
MySQL 5.7 solution
SELECT cust_type,
cust_name,
revenue,
#new_rank := IF(#current_type = cust_type, #new_rank + 1, 1),
#current_type := cust_type
FROM rank_test,
(SELECT #new_rank := null, #current_type := null) r
ORDER BY cust_type DESC, revenue DESC;
MySQL 8.0 solution
SELECT cust_type,
cust_name,
revenue,
RANK() OVER (
PARTITION BY
cust_type
ORDER BY
revenue DESC
) new_rank
FROM rank_test;
Here is schema to try on dbfiddle
CREATE TABLE rank_test (
cust_type varchar(20),
cust_name varchar(20),
revenue numeric(19, 6)
);
INSERT INTO rank_test VALUES ('TOP', 'A', 10000);
INSERT INTO rank_test VALUES ('TOP', 'B', 9000);
INSERT INTO rank_test VALUES ('TOP', 'C', 8000);
INSERT INTO rank_test VALUES ('BOTTOM', 'A', 10000);
INSERT INTO rank_test VALUES ('BOTTOM', 'B', 9000);
INSERT INTO rank_test VALUES ('BOTTOM', 'C', 8000);
Related
I have this table format:
| productid | price |
| ----------| -------------- |
| 1 | 10 |
| 2 | 20 |
| 3 | 30 |
| 4 | 40 |
Let's say I want to select all rows where:
(productid is 1 and price is over 50) OR
(productid is 2 and price is over 100) OR
(productid is 3 and price is over 20)
Is there a better generic way to achieve it (something with like arrays with indexes or something) other that do one at a time like:
select * from table where (productid = 1 and price > 50) OR
(productid = 2 and price > 100) OR
(productid = 3 and price > 20)
I would use a values clause:
select *
from the_table t
join (
values (1, 50),
(2, 100),
(3, 20)
) as p(productid, price)
on t.productid = p.productid
and t.price > p.price;
with conditions as (
select a[1]::int as product_id, a[2]::int as min_price
from (select regexp_split_to_array(unnest(string_to_array('1,50;2,100;3,20', ';')),',')) as dt(a)
)
select t.* from my_table t
inner join conditions c on t.product = c.product_id
and t.price >= c.min_price
test data:
drop table if exists my_table;
create temp table if not exists my_table(product int, price int, note text);
insert into my_table
select 1,10, 'some info 1:10' union all
select 1,20, 'some info 1:20' union all
select 1,50, 'some info 1:50' union all
select 2,20, 'some info 2:10' union all
select 2,100, 'some info 2:100:1' union all
select 2,100, 'some info 2:100:2' union all
select 3,30, 'some info 3:30' union all
select 4,40, 'some info 4:40';
result:
1 50 "some info 1:50"
2 100 "some info 2:100:1"
2 100 "some info 2:100:2"
3 30 "some info 3:30"
unnest(string_to_array('1,50;2,100;3,20', ';'))-- split CSV to rows
regexp_split_to_array(...., ',') -- split CSV to columns
I am having trouble trying to find an appropriate query(SQL-SERVER) for selecting records with condition however, the table I will be using has more than 100,000 rows and more than 20 columns.
So I need a code that satisfies the following condition:
1.)If [policy] and [plan] column is unique between rows then I will select that record
2.)If [policy] and [plan] return 2 or more rows then I will select the record which 'code' column isn't 999
3.)In some cases the unwanted rows may not have '999' in [code] column but may be other specifics
In other words, I would like to get row number 1,2,4,5,7.
Here is an example of what the table looks like
row #|policy|plan|code
-----------------------
1 | a | aa |111
-----------------------
2 | b | bb |112
-----------------------
3 | b | bb |999
-----------------------
4 | c | cc |111
-----------------------
5 | c | cc |112
-----------------------
6 | c | cc |999
-----------------------
7 | d | dd |999
-----------------------
I'm expecting to see something like
row #|policy|plan|code
-----------------------
1 | a | aa |111
-----------------------
2 | b | bb |112
-----------------------
4 | c | cc |111
-----------------------
5 | c | cc |112
-----------------------
7 | d | dd |999
-----------------------
Thank you in advance
This sounds like a prioritization query. You an use row_number():
select t.*
from (select t.*,
row_number() over (partition by policy, plan
order by code
) as seqnum
from t
) t
where seqnum = 1;
The expected output makes this a bit clearer:
select t.*
from (select t.*,
rank() over (partition by policy, plan
order by (case when code = 999 then 1 else 2 end) desc
) as seqnum
from t
) t
where seqnum = 1;
The OP wants all codes that are not 999 unless the only codes are 999. So, another approach is:
select t.*
from t
where t.code <> 999
union all
select t.*
from t
where t.code = 999 and
not exists (select 1
from t t2
where t2.policy = t.policy and t2.plan = t.plan and
t2.code <> 999
);
May be you want this (eliminate the last row if more than one)?
select t.*
from (select t.*
, row_number() over (partition by policy, plan
order by code desc
) AS RN
, COUNT(*) over (partition by policy, plan) AS RC
from t
) t
where RN > 1 OR RN=RC;
Output:
row policy plan code RN RC
1 1 a aa 111 1 1
2 2 b bb 112 2 2
3 5 c cc 112 2 3
4 4 c cc 111 3 3
5 7 d dd 999 1 1
CREATE TABLE #Table2
([row] int, [policy] varchar(1), [plan] varchar(2), [code] int)
;
INSERT INTO #Table2
([row], [policy], [plan], [code])
VALUES
(1, 'a', 'aa', 111),
(2, 'b', 'bb', 112),
(3, 'b', 'bb', 999),
(4, 'c', 'cc', 111),
(5, 'c', 'cc', 112),
(6, 'c', 'cc', 999),
(7, 'd', 'dd', 999)
;
with cte
as
(
select *,
row_number() over (partition by policy, [plan]
order by code
) as seqnum
from #Table2
)
select [row], [policy], [plan], [code] from cte where seqnum=1
I am looking to find instances of GROUPID where all price values are 0. The following is a simplified version of what I am looking at
--------------------------------
| Groupid | Price | Customer|
--------------------------------
| 001 | 9 | 4 |
| 001 | 0 | 4 |
| 002 | 4 | 4 |
| 002 | 4 | 4 |
| 003 | 0 | 4 |
| 003 | 0 | 4 |
| 004 | 4 | 4 |
| 004 | 7 | 4 |
--------------------------------
I am attempting to use the following query to find all GROUPID where both PRICE values for that particular group = 0.
SELECT * FROM MYTABLE WHERE GROUPID IN
(SELECT TB1.GROUPID FROM MYTABLE TB1 JOIN MYTABLE TB2 ON TB1.GROUPID = TB2.GROUPID
AND TB1.PRICE = 0 AND TB2.PRICE = 0)
and CUSTOMER = 4
ORDER BY GROUPID;
This query returns:
| Groupid | Price | Customer|
--------------------------------
| 001 | 9 | 4 |
| 001 | 0 | 4 |
| 003 | 0 | 4 |
| 003 | 0 | 4 |
--------------------------------
In my case, I only need it to return GROUPID 003.
I'd also like to ask for assistance in modifying the query to return all non 0 equal PRICE values within a groupid. It doesn't have to be in the same query as above. For example the return would be:
| Groupid | Price | Customer|
--------------------------------
| 002 | 4 | 4 |
| 002 | 4 | 4 |
Any help would be appreciated. Thank you for your time.
If all the prices are zero, then look at the minimum and maximum price for the groupid:
select groupid
from mytable t
group by groupid
having min(price) = 0 and max(price) = 0;
I should point out that no self-join is required for this.
Try this:
SELECT * FROM MYTABLE as m1 where Price = 0
and not exists (
select 1
from MYTABLE as m2
where m2.Groupid = m1.Groupid
and m2.Price <> 0
)
You can list the the rows, whose group_id has no rows with non-zero price
select groupid, price, customer from mytable t
where not exists (
select 1 from mytable where group_id = t.group_id
and price != 0
);
I would do it by counting the number of distinct prices over each group (and I'm assuming that each group will have the same customer) and then filtering on the price(s) you're interested in.
For example, for prices that are 0 for all rows in each groupid+customer:
WITH mytable AS (SELECT '001' groupid, 9 price, 4 customer FROM dual UNION ALL
SELECT '001' groupid, 0 price, 4 customer FROM dual UNION ALL
SELECT '002' groupid, 4 price, 4 customer FROM dual UNION ALL
SELECT '002' groupid, 4 price, 4 customer FROM dual UNION ALL
SELECT '003' groupid, 0 price, 4 customer FROM dual UNION ALL
SELECT '003' groupid, 0 price, 4 customer FROM dual UNION ALL
SELECT '004' groupid, 4 price, 4 customer FROM dual UNION ALL
SELECT '004' groupid, 7 price, 4 customer FROM dual)
SELECT groupid,
customer,
price
FROM (SELECT groupid,
customer,
price,
COUNT(DISTINCT price) OVER (PARTITION BY groupid, customer) num_distinct_prices
FROM mytable)
WHERE num_distinct_prices = 1
AND price = 0;
GROUPID CUSTOMER PRICE
------- ---------- ----------
003 4 0
003 4 0
Just change the and price = 0 to and price != 0 if you want the groups which have the same non-zero price for all rows. Or simply remove that predicate altogether.
EDIT
Gordon's is the best solution for the first part:
SELECT groupid
FROM mytable t
GROUP BY GroupID
HAVING (MAX(price) = 0 and MIN(price) = 0)
And for the second part:
SELECT groupid
FROM mytable t
GROUP BY GroupID
HAVING MIN(price) <> 0 AND (MAX(price) = MIN(price))
My original one:
SELECT groupid
FROM mytable t
GROUP BY GroupID
HAVING SUM(Price) =0
This assumes, there are no negative prices.
To the second part of your question:
SELECT groupid
FROM mytable t
WHERE Price > 0
GROUP BY GroupID, Price
HAVING COUNT(price) > 1
In your sample data, you have only one customer. I assume if you had more than one customer, you would still want to return the rows where the groupid has the same price, across all rows and all customers. If so, you could use the query below. It is almost the same as Boneist's - I just use min(price) and max(price) instead of count(distinct), and I don't include customer in partition by.
If the price may be NULL, it will be ignored in the computation of max price and min price; if all the NON-NULL prices are equal for a groupid, all the rows for that group will be returned. If price can be NULL and this is NOT the desired behavior, that can be changed easily - but the OP will have to clarify.
The query below retrieves all the cases when there is a single price for the groupid. To retrieve only the groups where the price is 0 (an additional condition), add and price = 0 to the WHERE clause of the outer query. I added more test data to illustrate some of the cases the query covers.
with
test_data ( groupid, price, customer ) as (
select '001', 9, 4 from dual union all
select '001', 0, 4 from dual union all
select '002', 4, 4 from dual union all
select '002', 4, 4 from dual union all
select '003', 0, 4 from dual union all
select '003', 0, 4 from dual union all
select '004', 4, 4 from dual union all
select '004', 7, 4 from dual union all
select '002', 4, 8 from dual union all
select '005', 2, 8 from dual union all
select '005', null, 8 from dual
),
prep ( groupid, price, customer, min_price, max_price) as (
select groupid, price, customer,
min(price) over (partition by groupid),
max(price) over (partition by groupid)
from test_data
)
select groupid, price, customer
from prep
where min_price = max_price
;
GROUPID PRICE CUSTOMER
------- --------- ---------
002 4 8
002 4 4
002 4 4
003 0 4
003 0 4
005 8
005 2 8
This may be what you want:
SELECT * FROM MYTABLE
WHERE GROUPID NOT IN (
SELECT GROUPID
FROM MYTABLE
WHERE Price <> 0)
and just change the last line for the other query:
SELECT * FROM MYTABLE
WHERE GROUPID NOT IN (
SELECT GROUPID
FROM MYTABLE
WHERE Price = 0)
I would do it very similarly to what Gordon posted
SELECT groupId
FROM MyTable
GROUP BY groupId
HAVING SUM(price) = 0
I have been trying to do the bellow query but honestly it's driving me crazy.
I have 2 Tables on MS SQL CE 4.0
Table 1 Name: Items
ID
Item_Code
Logged_by
Description
ID | Item_Code | Logged_by | Description
1 | A | Pete | just an A
2 | B | Mary | Seams like a B
3 | C | Joe | Obviously this is a C
4 | D | Pete | This is another A
Table 2 Name: Item_Comments
ID
Item_Code
Comment
Date
ID | Item_Code | Comment | Date
1 | B | Done | 2014/08/08
2 | A | Nice A | 2014/08/08
3 | B | Send 1 More | 2014/08/09
4 | C | Done | 2014/08/10
5 | D | This is an A | 2014/08/10
6 | D | Opps Sorry | 2014/08/11
The wanted result: I'm looking to join the most recent comment from Item_Comments to the Items Table
ID | Item_Code | Logged_by | Description | Comment
1 | A | Pete | just an A | Nice A
2 | B | Mary | Seams like a B | Send 1 More
3 | C | Joe | Obviously this is a C | Done
4 | D | Pete | This is another A | Opps Sorry
I did this query but I'm getting all the information =( mixed.
SELECT *
FROM Items t1
JOIN
(SELECT Item_Code, Comment, MAX(date) as MyDate
FROM Item_Comments
Group By Item_Code, Comment, Date
) t2
ON Item_Code= Item_Code
ORDER BY t1.Item_Code;
Do you know any way to do this ?
Try:
select x.*, z.comment
from items x
join (select item_code, max(date) as latest_dt
from item_comments
group by item_code) y
on x.item_code = y.item_code
join item_comments z
on y.item_code = z.item_code
and y.latest_dt = z.date
Fiddle test: http://sqlfiddle.com/#!6/d387f/8/0
You were close with your query but in your inline view aliased as t2 you were grouping by comment, leaving the max function to not actually aggregate anything at all. In t2 you should have just selected item_code and max(date) and grouped only by item_code, then you can use that to join into item_comments (y and z in my query above).
This is a second way of doing this using a subquery, however I would stick to the above (a join w/ an inline view):
select i.*, c.comment
from items i
join item_comments c
on i.item_code = c.item_code
where c.date = (select max(x.date)
from item_comments x
where x.item_code = c.item_code)
order by i.id
Fiddle test: http://sqlfiddle.com/#!6/d387f/11/0
Note if you run this inside piece you get every single record:
SELECT Item_Code, Comment, MAX(date) as MyDate
FROM Item_Comments
Group By Item_Code, Comment, Date
You want only the most recent comment. Assuming this is SQL Server 2008 or earlier, this get's you the most recent date for each Item_Code:
SELECT Item_Code, MAX(date) as MyDate
FROM Item_Comments
Group By Item_Code
Now you need to join that back and look up the comment on that date:
SELECT C.*
FROM Item_Comments C
INNER JOIN
(SELECT Item_Code, MAX(date) as MyDate
FROM Item_Comments
Group By Item_Code
) t2
ON C.Item_Code= t2.Item_Code
AND C.date = t2.MyDate
Now you can use that to join back to your original table:
SELECT t1.*, LatestComment.*
FROM Items t1
INNER JOIN
(
SELECT C.*
FROM Item_Comments C
INNER JOIN
(SELECT Item_Code, MAX(date) as MyDate
FROM Item_Comments
Group By Item_Code
) t2
ON C.Item_Code= t2.Item_Code
AND C.date = t2.MyDate
) LatestComment
On LatestComment.Item_Code = t1.Item_Code
Depending on the actual database you are using, this can get much simpler. Thats why you need to tag your database and version.
Try this,
create table items (id int, item_code char(1), logged_by varchar(10), description varchar(30));
insert into items values (1, 'A', 'Pete', 'just an A');
insert into items values (2, 'B', 'Mary', 'Seams like a B');
insert into items values (3, 'C', 'Joe', 'Obviously this is a C');
insert into items values (4, 'D', 'Pete', 'This is another A');
create table item_comments (id int, item_code char(1), comment varchar(20), date date);
insert into item_comments values (1, 'B', 'Done', '2014/08/08');
insert into item_comments values (2, 'A', 'Nice A', '2014/08/08');
insert into item_comments values (3, 'B', 'Send 1 More', '2014/08/09');
insert into item_comments values (4, 'C', 'Done', '2014/08/10');
insert into item_comments values (5, 'D', 'This is an A', '2014/08/10');
insert into item_comments values (6, 'D', 'Opps Sorry', '2014/08/11');
select * from items;
select * from item_comments;
select * from (select i.logged_by,i.id,i.item_code,i.description,ic.comment
,row_number() over(partition by i.id order by i.id )as Rnk
from items i inner join item_comments ic
on i.item_code=ic.item_code and i.id in(1,3)) x
where x.Rnk=1
union
select * from (select i.logged_by,i.id,i.item_code,i.description,ic.comment
,row_number() over(partition by i.id order by i.id )as Rnk
from items i inner join item_comments ic
on i.item_code=ic.item_code and i.id in(2,4)
) x where x.Rnk=2 order by item_code
I have a table with an id as the primary key, and a description as another field.
I want to first select the records that have the id<=4, sorted by description, then I want all the other records (id>4), sorted by description. Can't get there!
select id, descr
from t
order by
case when id <= 4 then 0 else 1 end,
descr
select *, id<=4 as low from table order by low, description
You may want to use an id <= 4 expression in your ORDER BY clause:
SELECT * FROM your_table ORDER BY id <= 4 DESC, description;
Test case (using MySQL):
CREATE TABLE your_table (id int, description varchar(50));
INSERT INTO your_table VALUES (1, 'c');
INSERT INTO your_table VALUES (2, 'a');
INSERT INTO your_table VALUES (3, 'z');
INSERT INTO your_table VALUES (4, 'b');
INSERT INTO your_table VALUES (5, 'g');
INSERT INTO your_table VALUES (6, 'o');
INSERT INTO your_table VALUES (7, 'c');
INSERT INTO your_table VALUES (8, 'p');
Result:
+------+-------------+
| id | description |
+------+-------------+
| 2 | a |
| 4 | b |
| 1 | c |
| 3 | z |
| 7 | c |
| 5 | g |
| 6 | o |
| 8 | p |
+------+-------------+
8 rows in set (0.00 sec)
Related post:
Using MySql, can I sort a column but have 0 come last?
select id, description
from MyTable
order by case when id <= 4 then 0 else 1 end, description
You can use UNION
SELECT * FROM (SELECT * FROM table1 WHERE id <=4 ORDER by description)aaa
UNION
SELECT * FROM (SELECT * FROM table1 WHERE id >4 ORDER by description)bbb
OR
SELECT * FROM table1
ORDER BY
CASE WHEN id <=4 THEN 0
ELSE 1
END, description