Category Archives: DBMS

Postgres error: Relation Does not exist

Faced an interesting error while accessing postgres table with a simple java code.

"select * from User"

A little googling revealed that if you are using multicase table name (U is caps), you will need to give the table name in quotes.

A better fix would be to always use lower case characters when declaring tables in postgres.

Find and delete blocking sessions- Oracle

Following query would give you all the session running in your oracle database with query details

select sesion.sid,
from v$sqlarea sqlarea, v$session sesion
where sesion.sql_hash_value = sqlarea.hash_value
and sesion.sql_address = sqlarea.address
and sesion.username is not null;

Now you might want to kill an unwanted session


Handling Null Values in SQL- ISNULL, NVL2, IFNULL, Coalesce

There can be situations when if a value is null in sql query, you want to replace it with something else. Different DBs provide you different options

Oracle- PLSQL
select NVL2(teamid, teamid, deptid) allocationid, empid, name from employee; //If team id is available that is returned as allocation id else deptid

select ISNULL(teamid, deptid) allocationid, empid, name from employee;

Similarly MySQL has IFNULL

Another option is COALESCE, this works for both Oracle and MS-SQL. This is slightly different from others in implementation that it makes columns to be skipped unless first non-null value is found

select COALESCE(teamid, deptid, organizationid) allocationid, empid, name from employee;

Forgot password for Oracle accounts

I have a oracle installation on my windows machine but that was done a few months back and I could not recollect passwords for default sys and sysadmin records. The following trick helped reset passwords for these accounts

1. Go to Command prompt and type sqlplus /nolog. It should open SQL/> prompt.
2. On SQL/> prompt, type connect /as dba
Now you are logged in as dba, changing password is simple matter of
SQL> alter user sys identified by “newpassword”

Union vs Union All

The simple difference between using Union and Union All is that Union selects only distinct records from both the resultsets, whereas Union All will return everything.

But today I figured out one more use of union All today. In a ms sql query, I was using union to add up resultset of multiple queries, and got this error

“The text data type cannot be selected as DISTINCT because it is not comparable.”

As I knew none of my queries will return duplicate records, I did not need distinct clause. Replacing Union with “Union All” did the trick in this case.

Create a table as result of select query

If you want to dump the result of your select query into a table, this query will help in Oracle or MySql.

Create table newTable select * from oldTable where condition=true;

In case of MsSql this will work

Select * into newTable from oldTable where condition=true;

Shared lock- Effecting Select query’s performance

A shared log in SQL query is applied by default by SELECT statements. The difference between Shared and exclusive locks is that exclusive lock will stop any update on the data being locked, whereas shared lock suggests the select query to wait if some update or insert operation is in progress for the row.

Shared lock will not stop any other request from selecting or updating the records.

For example- select * from employee;

The above statement will fetch employee data from employee table, but if at the same time some insert or update query is executing, it will impact the performance of select query as it has to wait for other operation (on the row/ page it is trying to read) to finish.

Workaround: If we are sure that other operations executing at the same time are not going to impact my select query, i.e. I am ok to read a few old records/ dirty read, Sql provides us keyword NOLOCK

select * from employee with (NOLOCK);

This will make sure our select query does not get impacted by insert and updates running at the same time, though we might have problem of dirty read.

Workaround 2: With NOLOCK we know we can have a dirty read problem, to avoid that, we can use READPAST. This will make the select query skip any rows being updated at the query time and hence avoids dirty reads, So we are achieving our goal of not waiting for insert/ update operations to be over plus. On the downside, the resultset will miss some of the records.

select * from employee with (READPAST);

concatenate strings in different rows of a recordset

How to concatenate strings in different rows of a recordset, using a common id?


I have a table (test123) like this

IdValue DataValue
1 aaaaa
2 bbbbb
1 ccccc
2 ddddd

Desired output

IdValue Data_Concatenated
1 aaaaa,ccccc
2 bbbbb,ddddd

Solution 1: Using sys_connect_by_path

SELECT idvalue ,
SUBSTR(MAX(sys_connect_by_path(datavalue, ‘,’ )),2) data_concatenated
(SELECT idvalue ,
datavalue ,
row_number() over (partition BY idvalue order by datavalue) row#
FROM test123
START WITH row# = 1
CONNECT BY prior row# = row#-1
AND prior idvalue = idvalue
GROUP BY idvalue
ORDER BY idvalue;

Solution 2:

SELECT idvalue, RTRIM (EXTRACT (XMLAGG (XMLELEMENT (“X”, dataVALUE || ‘,’)), ‘/X/text()’),’,’) VALUE
FROM test123
group by IDvalue;

Stored procedure Vs SQL in code

Originally Posted December 24, 2007

The debate is old. But every now and then, we have to make a decision between using stored procedures and writing sql queries in code. When you are implementing sql queries for your database, you can either write them down in your code (again you can have different layers in your code and you can decide where to implement your sql queries, but that is different topic) or you can store them in database itself as stored procedures.

There are two schools of thoughts, one that will recommend going for stored procedure and the second will propose no stored procedures approach. I tried to understand the difference between the two approaches.

What is a stored procedure for a database? It is same as any procedure you create in your coding language, that takes some arguments, apply some logic, execute and returns back a result (if required).

As I already mentioned the debate is really old, so instead of reinventing the wheel, I will borrow the pointers from different sources and then try to add my thoughts to that

Stored procedures are secure: Stored procedure are secured against SQL injections. Are they totally safe? Isn’t it again the responsibility of person creating the procedure to make sure it is safe. Again, if I am writing queries in application layer (code), I can make sure the code is protected against the attacks.

Separation of data logic from business logic: All your queries which are interacting with database are residing over the database engine itself. This clearly separates all database interactions and hence make design more cleaner. One alternative will be to divide yourapplication layer into different parts, and keep data handling part completely separate from the rest.

A Disadvantage: Moving stored procedure from one platform to another can be tough, eg you decided to move from MSSQL to MYSQL, you may have recode all your stored porcedures. But this has a related advantage that if the front end language is changing, say you are moving from php to java, and your database is not changing then you don???t need to change the stored procedures.

Testing the app: Testing application with stored procedures can be a pain. Yes, as the changes are to be made in multiple places, your code as well as stored procedure. Now if database is handled by a different team, every time for a small change you have to communicate the changes and then test. This will take longer time than making changes just in your code as multiple teams are involved.

Stored procedure is faster: now that is debatable statement. I found this thing mentioned at many places that stored procedures as faster as the procedure is part of database engine itself and hence have faster access to database. Also you don’t need to send the complete query but just the procedure name. But as per my exp, I never faced any differences in terms of execution time with stored procedure.

Having considered all the points above, my personal opinion would be not to use the stored procedures especially in big application and if the same database is being used by multiple applications, because you don’t have complete control over the database and every time a change is made to a procedure it has to be communicated to multiple parties. In small applications and small teams one can use stored procedures as mostly same team has control of both database engine and code layer, so it is easier to coordinate and implement the changes.

Why rownum will nor work for greater or equal to checks

Someone asked a question on a forum that why rownum=2 does not work for his query, so i though of sharing some details here.

Using rownum is a tricky affair. Safest bet is to use it only when you want to limit the number of results to be shown. For example rownum<2 or rownum<=5.

Why rownum=2 or rownum>2 will not work?

Read here –

In summary, this is how oracle execute a query

  1. The FROM/WHERE clause goes first.
  2. ROWNUM is assigned and incremented to each output row from the FROM/WHERE clause.
  3. SELECT is applied.
  4. GROUP BY is applied.
  5. HAVING is applied.
  6. ORDER BY is applied.

rownum<=2 clause will get converted to

for x in
( select * from emp )
    exit when NOT(ROWNUM <= 2)
    OUTPUT record to temp
end loop

if you change exit when NOT(ROWNUM <= 2) with rownnum=2, you can see it will fail in the first run itself

So if I cannot use rownum, what can I use. Try using row_number()

It works something like

SELECT last_name FROM
   (SELECT last_name, ROW_NUMBER() OVER (ORDER BY last_name) R FROM employees)
   WHERE R BETWEEN 51 and 100;