Oracle Group by NULL中的SQL返回多行

Question

I have a sql statement which is trying to retrieve the rows having a count greater than 1 for all the similar rows. In all those rows, one field appears to be holding null value.

Case 1 :

Let's say the table name is ABC

---

BUSINESS_UNIT, INVOICE, FLAG

A 1 (Null)

A 1 (Null)

---

SQL Statement : SELECT BUSINESS_UNIT, INVOICE, FLAG from TABLE ABC group by BUSINESS_UNIT, INVOICE, FLAG having COUNT(*) > 1

I am expecting it to return no rows as both the flags are basically blank. But it returns the

---

BUSINESS_UNIT, INVOICE, FLAG

A 1 (Null)

---

Case 2 :

TABLE ABC

---

BUSINESS_UNIT, INVOICE, FLAG

A 1 (Null)

A 1 (Null)

B 1 1

B 1 (Null)

---

SQL Statement : SELECT BUSINESS_UNIT, INVOICE, FLAG from TABLE ABC group by BUSINESS_UNIT, INVOICE, FLAG having COUNT(*) >1

I am expecting the result to be

---

BUSINESS_UNIT, INVOICE, FLAG

B 1 1

---

This is for Oracle 11g. Can someone please help on this and let me know if we can use any of the delivered functions in Oracle?

---

After editing the original sql statements.

I should actually have removed flag from the Group by and select and added to the count in order to identify multiple flags for similar rows.

---

Thanks!

Answer 1

well, null is also a value, so if you group by a nullable field, null values will be distinct from non null values.

If you don't wanna null values, just add a where clause

SELECT INVOICE 
from TABLE ABC 
where flag is not null
group by BUSINESS_UNIT, INVOICE, FLAG 
having COUNT(*) >1

Answer 2

This is your first query:

SELECT BUSINESS_UNIT, INVOICE, FLAG
from TABLE ABC
group by BUSINESS_UNIT, INVOICE, FLAG
having COUNT(*) > 1;

The count(*) is returning the number of rows in each group. One of the groups is flag and the NULL values are all grouped together. If you want to count non-NULL values of flag , then you can use:

SELECT BUSINESS_UNIT, INVOICE, FLAG
from TABLE ABC
group by BUSINESS_UNIT, INVOICE, FLAG
having COUNT(flag) > 1;

Or filter as Raphaël Althaus suggests.

Your second query is:

SELECT INVOICE
from TABLE ABC
group by BUSINESS_UNIT, INVOICE, FLAG
having COUNT(*) > 1;

You seem to magically want three columns in the output when there is only one column in the select . Yes, you can have more columns in the group by than in the select , if you really want.

Your final question is:

Can someone please help on this and let me know if we can use any of the delivered functions in Oracle?

Of course you can. Oracle has a very powerful set of functions. Most of them are consistent with the ANSI standard and behave the same way on Oracle and on other databases. You just need to learn how to use them correctly.

Answer 3

Besides adding a WHERE-clause as suggested by Raphaël Althaus you might also change count(*) to count(flag):

SELECT INVOICE 
from TABLE ABC 
group by BUSINESS_UNIT, INVOICE, FLAG 
having COUNT(flag) >1

Answer 4

I think I have found what I was intending to look for.

SELECT BUSINESS_UNIT, INVOICE from TABLE ABC group by BUSINESS_UNIT, INVOICE having COUNT(DISTINCT Flag) >1

Case 1: It returns no rows

Case 2 : It returns 1 row

---

BUSINESS_UNIT, INVOICE

B 1

---

Thanks Raphael, Dnoeth and Gordon for your help on this!