Oracle Best practices for changing Byte to Char on Varchar2 columns

Question

Dear Team,
Application Team wanted to change Byte to Char on Varchar2 columns to accommodate Multi byte character  on couple of production tables.
Wanted to know is it safe to have mixture of BYTE and CHAR semantics in the same table i have read on the couple of documents that It's good practice to avoid using a mixture of BYTE and CHAR semantics columns in the same table.
What happens if we have mixture of BYTE and CHAR semantics columns in the same table?
Do we need to gather stats & rebuild indexes on the table after these column changes .
Thanks in Advance !!!
SK

Answer 1

Application Team wanted to change Byte to Char on Varchar2 columns to accommodate Multi byte character  on couple of production tables.
Wanted to know is it safe to have mixture of BYTE and CHAR semantics in the same table i have read on the couple of documents that It's good practice to avoid using a mixture of BYTE and CHAR semantics columns in the same table.
No change is needed to 'accommodate Multibyte characters'. That support has NOTHING to do with whether a column is specified using BYTE or CHAR.
In 11g the limit for a VARCHAR2 column is 4000 bytes, period. If you specify CHAR and try to insert 1001 characters that each take 4 bytes you will get an exception since that would require 4004 bytes and the limit is 4000 bytes.
In practice the use of CHAR is mostly a convenience to the developer when defining columns for multibyte characters. For example for a NAME column you might want to make sure Oracle will allocate room for 50 characters REGARDLESS of the actual length in bytes.
If you provide a name of 50 one byte characters then only 50 bytes will be used. Provide a name of 50 four byte characters and 200 bytes will be used.
So if  that NAME column was defined using BYTE how would you know what length to use for the column? Fifty BYTES will seldom be long enough and 200 bytes SEEMS large since the business user wants a limit of FIFTY characters.
That is why such columns would typically use CHAR; so that the length (fifty) defined for the column matches the logical length of the number of characters.
What happens if we have mixture of BYTE and CHAR semantics columns in the same table?
Nothing happens - Oracle could care less.
Do we need to gather stats & rebuild indexes on the table after these column changes .
No - not if you by 'need' you mean simply because you made ONLY that change.
But that begs the question: if the table already exists, has data and has been in use without their being any problems then why bother changing things now?
In other words: if it ain't broke why try to fix it?
So back to your question of 'best practices'
Best practices is to set the length semantics at the database level when the database is first created and to then use that same setting (BYTE or CHAR) when you create new objects or make DDL changes.
Best practices is also to not fix things that aren't broken.
See the 'Length Semantics' section of the globalization support guide for more best practices
http://docs.oracle.com/cd/E11882_01/server.112/e10729/ch2charset.htm#i1006683