Difference between varchar2(4000 byte) & varchar2(4000 char

Question

Hi,
My existing database NLS parameters as follows
CHARACTER SET US7ASCII
NATIONAL CHARACTER SET AL16UTF16
created a test database to support globalization, I changed as follows
CHARACTER SET UTF8
NATIONAL CHARACTER SET UTF8
some of the table column datatypes are varchar2(4000)
I would like to know what is difference between VARCHAR2(4000 BYTE) and VARCHAR2(4000 CHAR).
Thanks

Answer 1

Indeed, VARCHAR2(x BYTE) means that the column will hold as much characters as will fit into x bytes. Depending on the character set and particular characters this may be x or less characters.
For example, a VARCHAR2(20 BYTE) column in an AL32UTF8 database can hold 20 characters from the ASCII range, 10 Latin letters with umlaut, 10 Cyryllic, 10 Hebrew, or 10 Arabic letters (2 bytes per character), or 6 Chinese, Japanese, Korean, or Devanagari (Indic) characters. Or a mixture of these characters of any total length up to 20 bytes.
VARCHAR2(x CHAR) means that the column will hold x characters but not more than can fit into 4000 bytes. Internally, Oracle will set the byte length of the column (DBA_TAB_COLUMNS.DATA_LENGTH) to MIN(x * mchw, 4000), where mchw is the maximum byte width of a character in the database character set. This is 1 for US7ASCII or WE8MSWIN1252, 2 for JA16SJIS, 3 for UTF8, and 4 for AL32UTF8.
For example, a VARCHAR2(3000 CHAR) column in an AL32UTF8 database will be internally defined as having the width of 4000 bytes. It will hold up to 3000 characters from the ASCII range (the character limit), but only 1333 Chinese characters (the byte limit, 1333 * 3 bytes = 3999 bytes). A VARCHAR2(100 CHAR) column in an AL32UTF8 database will be internally defined as having the width of 400 bytes. It will hold up to any 100 Unicode characters.
The above implies that the CHAR limit works optimally if it is lower than 4000/mchw. With such restriction, the CHAR limit guarantees that the defined number of characters will fit into the column. Because the widest character in any Oracle character set has 4 bytes, if x <= 1000, VARCHAR2(x CHAR) is guaranteed to hold up to x characters in any database character set.
The declaration VARCHAR2(x):
- for objects defined in SYS schema means VARCHAR2(x BYTE),
- for objects defined in other schemas it means VARCHAR2(x BYTE) or VARCHAR2(x CHAR), depending on the value of the NLS_LENGTH_SEMANTICS parameter of the session using the declaration (see the NLS_SESSION_PARAMETERS view).
After an object is defined, its BYTE vs CHAR semantics is stored in the data dictionary and it does not depend on the NLS_LENGTH_SEMANTICS any longer. Even Export/Import will not change this.
Character length semantics rules are valid for table columns and for PL/SQL variables.
-- Sergiusz