NAME
DBD::Oracle - Oracle database driver for the DBI module
SYNOPSIS
use DBI;
$dbh = DBI->connect("dbi:Oracle:$dbname", $user, $passwd);
$dbh = DBI->connect("dbi:Oracle:host=$host;sid=$sid", $user, $passwd);
# See the DBI module documentation for full details
# for some advanced uses you may need Oracle type values:
use DBD::Oracle qw(:ora_types);
DESCRIPTION
DBD::Oracle is a Perl module which works with the DBI module to provide access to Oracle databases.
CONNECTING TO ORACLE
This is a topic which often causes problems. Mainly due to Oracle's many and sometimes complex ways of specifying and connecting to databases. (James Taylor and Lane Sharman have contributed much of the text in this section.)
Connecting without environment variables or tnsname.ora file
If you use the host=$host;sid=$sid
style syntax, for example:
$dbh = DBI->connect("dbi:Oracle:host=myhost.com;sid=ORCL", $user, $passwd);
then DBD::Oracle will construct a full connection descriptor string for you and Oracle will not need to consult the tnsname.ora file.
If a port
number is not specified then the descriptor will try both 1526 and 1521 in that order (e.g., new then old). You can check which port(s) are in use by typing "$ORACLE_HOME/bin/lsnrctl stat" on the server.
Oracle Environment Variables
Oracle typically uses two environment variables to specify default connections: ORACLE_SID and TWO_TASK.
ORACLE_SID is really unnecessary to set since TWO_TASK provides the same functionality in addition to allowing remote connections.
% setenv TWO_TASK T:hostname:ORACLE_SID # for csh shell
$ TWO_TASK=T:hostname:ORACLE_SID export TWO_TASK # for sh shell
% sqlplus username/password
Note that if you have *both* local and remote databases, and you have ORACLE_SID *and* TWO_TASK set, and you don't specify a fully qualified connect string on the command line, TWO_TASK takes precedence over ORACLE_SID (i.e. you get connected to remote system).
TWO_TASK=P:sid
will use the pipe driver for local connections using SQL*Net v1.
TWO_TASK=T:machine:sid
will use TCP/IP (or D for DECNET, etc.) for remote SQL*Net v1 connection.
TWO_TASK=dbname
will use the info stored in the SQL*Net v2 tnsnames.ora configuration file for local or remote connections.
The ORACLE_HOME environment variable should be set correctly. In general, the value used should match the version of Oracle that was used to build DBD::Oracle. If using dynamic linking then ORACLE_HOME should match the version of Oracle that will be used to load in the Oracle client libraries (via LD_LIBRARY_PATH, ldconfig, or similar on Unix).
ORACLE_HOME can be left unset if you aren't using any of Oracle's executables, but it is not recommended and error messages may not display. It should be set to the ORACLE_HOME directory of the version of Oracle that DBD::Oracle was compiled with.
Discouraging the use of ORACLE_SID makes it easier on the users to see what is going on. (It's unfortunate that TWO_TASK couldn't be renamed, since it makes no sense to the end user, and doesn't have the ORACLE prefix).
Connection Examples Using DBD::Oracle
First, how to connect to a local database without using a Listener:
$dbh = DBI->connect('dbi:Oracle:SID','scott', 'tiger');
you can also leave the SID empty:
$dbh = DBI->connect('dbi:Oracle:','scott', 'tiger');
in which case Oracle client code will use the ORACLE_SID environment variable (if TWO_TASK env var isn't defined).
Below are various ways of connecting to an oracle database using SQL*Net 1.x and SQL*Net 2.x. "Machine" is the computer the database is running on, "SID" is the SID of the database, "DB" is the SQL*Net 2.x connection descriptor for the database.
Note: Some of these formats may not work with Oracle 8+.
BEGIN {
$ENV{ORACLE_HOME} = '/home/oracle/product/7.x.x';
$ENV{TWO_TASK} = 'DB';
}
$dbh = DBI->connect('dbi:Oracle:','scott', 'tiger');
# - or -
$dbh = DBI->connect('dbi:Oracle:','scott/tiger');
Refer to your Oracle documentatiion for valid values of TWO_TASK.
Here are some variations (not setting TWO_TASK) in order of preference:
$dbh = DBI->connect('dbi:Oracle:DB','username','password')
$dbh = DBI->connect('dbi:Oracle:DB','username/password','')
$dbh = DBI->connect('dbi:Oracle:','username@DB','password')
$dbh = DBI->connect('dbi:Oracle:host=foobar;sid=ORCL;port=1521', 'scott/tiger', '')
$dbh = DBI->connect('dbi:Oracle:', q{scott/tiger@(DESCRIPTION=
(ADDRESS=(PROTOCOL=TCP)(HOST= foobar)(PORT=1521))
(CONNECT_DATA=(SID=ORCL)))}, "")
If you are having problems with login taking a long time (>10 secs say) then you might have tripped up on an Oracle bug. You can try using one of the ...@DB variants as a workaround. E.g.,
$dbh = DBI->connect('','username/password@DB','');
On the other hand, that may cause you to trip up on another Oracle bug that causes alternating connection attempts to fail! (In reality only a small proportion of people experience these problems.)
To connect to a local database with a user which has been set-up to authenticate via the OS ("ALTER USER username IDENTIFIED EXTERNALLY"):
$dbh = DBI->connect('dbi:Oracle:','/','');
Note the lack of a connection name (use the ORACLE_SID environment variable). If an explicit SID is used you'll probably get an ORA-01004 error.
That only works for local databases. (Authentication to remote Oracle databases using your unix login name without a password and is possible but it's not secure and not recommended so not documented here. If you can't find the information elsewhere then you probably shouldn't be trying to do it.)
Optimizing Oracle's listener
[By Lane Sharman <lane@bienlogic.com>] I spent a LOT of time optimizing listener.ora and I am including it here for anyone to benefit from. My connections over tnslistener on the same humble Netra 1 take an average of 10-20 milli seconds according to tnsping. If anyone knows how to make it better, please let me know!
LISTENER =
(ADDRESS_LIST =
(ADDRESS =
(PROTOCOL = TCP)
(Host = aa.bbb.cc.d)
(Port = 1521)
(QUEUESIZE=10)
)
)
STARTUP_WAIT_TIME_LISTENER = 0
CONNECT_TIMEOUT_LISTENER = 10
TRACE_LEVEL_LISTENER = OFF
SID_LIST_LISTENER =
(SID_LIST =
(SID_DESC =
(SID_NAME = xxxx)
(ORACLE_HOME = /xxx/local/oracle7-3)
(PRESPAWN_MAX = 40)
(PRESPAWN_LIST=
(PRESPAWN_DESC=(PROTOCOL=tcp) (POOL_SIZE=40) (TIMEOUT=120))
)
)
)
1) When the application is co-located on the host AND there is no need for outside SQLNet connectivity, stop the listener. You do not need it. Get your application/cgi/whatever working using pipes and shared memory. I am convinced that this is one of the connection bugs (sockets over the same machine). Note the $ENV{ORAPIPES} env var. The essential code to do this at the end of this section.
2) Be careful in how you implement the multi-threaded server. Currently I am not using it in the initxxxx.ora file but will be doing some more testing.
3) Be sure to create user rollback segments and use them; do not use the system rollback segments; however, you must also create a small rollback space for the system as well.
5) Use large tuning settings and get lots of RAM. Check out all the parameters you can set in v$parameters because there are quite a few not documented you may to set in your initxxx.ora file.
6) Use svrmgrl to control oracle from the command line. Write lots of small SQL scripts to get at V$ info.
use DBI;
# Environmental variables used by Oracle
$ENV{ORACLE_SID} = "xxx";
$ENV{ORACLE_HOME} = "/opt/oracle7";
$ENV{EPC_DISABLED} = "TRUE";
$ENV{ORAPIPES} = "V2";
my $dbname = "xxx";
my $dbuser = "xxx";
my $dbpass = "xxx";
my $dbh = DBI->connect("dbi:Oracle:$dbname", $dbuser, $dbpass)
|| die "Unale to connect to $dbname: $DBI::errstr\n";
Oracle utilities
If you are still having problems connecting then the Oracle adapters utility may offer some help. Run these two commands:
$ORACLE_HOME/bin/adapters
$ORACLE_HOME/bin/adapters $ORACLE_HOME/bin/sqlplus
and check the output. The "Protocol Adapters" section should be the same. It should include at least "IPC Protocol Adapter" and "TCP/IP Protocol Adapter".
If it generates any errors which look relevant then please talk to yor Oracle technical support (and not the dbi-users mailing list). Thanks. Thanks to Mark Dedlow for this information.
Constants
- :ora_session_modes
-
ORA_SYSDBA ORA_SYSOPER
- :ora_types
-
ORA_VARCHAR2 ORA_STRING ORA_NUMBER ORA_LONG ORA_ROWID ORA_DATE ORA_RAW ORA_LONGRAW ORA_CHAR ORA_CHARZ ORA_MLSLABEL ORA_NTY ORA_CLOB ORA_BLOB ORA_RSET
- SQLCS_IMPLICIT
- SQLCS_NCHAR
-
SQLCS_IMPLICIT and SQLCS_NCHAR are character set form values. See notes about Unicode elsewhere in this document.
- ORA_OCI
-
Oracle doesn't provide a formal API for determining the exact version number of the OCI client library used, so DBD::Oracle has to go digging (and sometimes has to more or less guess). The ORA_OCI constant holds the result of that process.
In string context ORA_OCI returns the full "A.B.C.D" version string.
In numeric context ORA_OCI returns the major.minor version number (8.1, 9.2, 10.0 etc). But note that version numbers are not actually floating point and so if Oracle ever makes a release that has a two digit minor version, such as
9.10
it will have a lower numeric value than the preceeding9.9
release. So use with care.The contents and format of ORA_OCI are subject to change (it may, for example, become a version object in later releases). I recommend that you avoid checking for exact values.
Connect Attributes
- ora_session_mode
-
The ora_session_mode attribute can be used to connect with SYSDBA authorization and SYSOPER authorization. The ORA_SYSDBA and ORA_SYSOPER constants can be imported using
use DBD::Oracle qw(:ora_session_modes);
This is one case where setting ORACLE_SID may be useful since connecting as SYSDBA or SYSOPER via SQL*Net is frequently disabled for security reasons.
Example:
$dsn = "dbi:Oracle:"; # no dbname here $ENV{ORACLE_SID} = "orcl"; # set ORACLE_SID as needed delete $ENV{TWO_TASK}; # make sure TWO_TASK isn't set $dbh = DBI->connect($dsn, "", "", { ora_session_mode => ORA_SYSDBA });
It has been reported that this only works if $dsn does not contain a SID so that Oracle then uses the value of the ORACLE_SID (not TWO_TASK) environment variable to connect to a local instance. Also the username and password should be empty, and the user executing the script needs to be part of the dba group or osdba group.
- ora_oratab_orahome
-
Passing a true value for the ora_oratab_orahome attribute will make DBD::Oracle change $ENV{ORACLE_HOME} to make the Oracle home directory specified in the
/etc/oratab
file if the database to connect to is specified as a SID that exists in the oratab file, and DBD::Oracle was built to use the Oracle 7 OCI API (not Oracle 8+). - ora_module_name
-
After connecting to the database the value of this attribute is passed to the SET_MODULE() function in the
DBMS_APPLICATION_INFO
PL/SQL package. This can be used to identify the application to the DBA for monitoring and performance tuning purposes. For example:DBI->connect($dsn, $user, $passwd, { ora_module_name => $0 });
-
Needs at least Perl 5.8.0 compiled with ithreads. Allows to share database connections between threads. The first connect will make the connection, all following calls to connect with the same ora_dbh_share attribute will use the same database connection. The value must be a reference to a already shared scalar which is initialized to an empty string.
our $orashr : shared = '' ; $dbh = DBI->connect ($dsn, $user, $passwd, {ora_dbh_share => \$orashr}) ;
- ora_use_proc_connection
-
This attribute allows to create a DBI handle for an existing SQLLIB database connection. This can be used to share database connections between Oracle ProC code and DBI running in an embedded Perl interpreter. The SQLLIB connection id is appended after the "dbi:Oracle:" initial argument to DBI::connect.
For example, if in ProC a connection is made like
EXEC SQL CONNECT 'user/pass@db' AT 'CONID';
the connection may be used from DBI after running something like
my $dbh = DBI->connect("dbi:Oracle:CONID", "", "", { ora_use_proc_connection => 1 });
To disconnect, first call $dbh->disconnect(), then disconnect in ProC.
This attribute requires DBD::Oracle to be built with the -ProC option to Makefile.PL. It is not available with OCI_V7. Not tested with Perl ithreads or with the ora_dbh_share connect attribute.
- ora_envhp
-
The first time a connection is made a new OCI 'environment' is created by DBD::Oracle and stored in the driver handle. Subsequent connects reuse (share) that same OCI environment by default.
The ora_envhp attribute can be used to disable the reuse of the OCI environment from a previous connect. If the value is
0
then a new OCI environment is allocated and used for this connection.The OCI environment is what holds information about the client side context, such as the local NLS environment. So by altering %ENV and setting ora_envhp to 0 you can create connections with different NLS settings. This is most useful for testing.
Database Handle Attributes
ora_ph_type
-
The default placeholder data type for the database session. The
TYPE
or "ora_type" attributes to "bind_param" in DBI and "bind_param_inout" in DBI override the data type for individual placeholders. The most frequent reason for using this attribute is to permit trailing spaces in values passed by placeholders.Constants for the values allowed for this attribute can be imported using
use DBD::Oracle qw(:ora_types);
Only the following values are permitted for this attribute.
- ORA_VARCHAR2
-
Strip trailing spaces and allow embedded \0 bytes. This is the normal default placeholder type.
- ORA_STRING
-
Don't strip trailing spaces and end the string at the first \0.
- ORA_CHAR
-
Don't strip trailing spaces and allow embedded \0. Force 'blank-padded comparison semantics'.
- ora_parse_error_offset
-
If the previous error was from a failed
prepare
due to a syntax error, this attribute gives the offset into theStatement
attribute where the error was found.
Prepare Attributes
These attributes may be used in the \%attr
parameter of the "prepare" in DBI database handle method.
- ora_placeholders
-
Set to false to disable processing of placeholders. Used mainly for loading a PL/SQL package that has been wrapped with Oracle's
wrap
utility. - ora_parse_lang
-
Tells the connected database how to interpret the SQL statement. If 1 (default), the native SQL version for the database is used. Other recognized values are 0 (old V6, treated as V7 in OCI8), 2 (old V7), 7 (V7), and 8 (V8). All other values have the same effect as 1.
- ora_auto_lob
-
If true (the default), fetching retreives the contents of the CLOB or BLOB column in most circumstances. If false, fetching retreives the Oracle "LOB Locator" of the CLOB or BLOB value.
See "Handling LOBs" for more details. See also the LOB tests in 05dbi.t of Oracle::OCI for examples of how to use LOB Locators.
- ora_check_sql
-
If 1 (default), force SELECT statements to be described in prepare(). If 0, allow SELECT statements to defer describe until execute().
See "Prepare postponed till execute" for more information.
Placeholder Binding Attributes
These attributes may be used in the \%attr
parameter of the "bind_param" in DBI or "bind_param_inout" in DBI statement handle methods.
- ora_type
-
Specify the placeholder's data type using an Oracle data type. A fatal error is raised if
ora_type
and the DBITYPE
attribute are used for the same placeholder. Some of these types are not supported by the current version of DBD::Oracle and will cause a fatal error if used. Constants for the Oracle datatypes may be imported usinguse DBD::Oracle qw(:ora_types);
Potentially useful values when DBD::Oracle was built using OCI 7 and later:
ORA_VARCHAR2, ORA_STRING, ORA_LONG, ORA_RAW, ORA_LONGRAW, ORA_CHAR, ORA_MLSLABEL, ORA_RSET
Additional values when DBD::Oracle was built using OCI 8 and later:
ORA_CLOB, ORA_BLOB, ORA_NTY
See "Binding Cursors" for the correct way to use ORA_RSET.
See "Handling LOBs" for how to use ORA_CLOB and ORA_BLOB.
See "Other Data Types" for more information.
See also "Placeholders and Bind Values" in DBI.
- ora_csform
-
Specify the OCI_ATTR_CHARSET_FORM for the bind value. Valid values are SQLCS_IMPLICIT (1) and SQLCS_NCHAR (2). Both those constants can be imported from the DBD::Oracle module. Rarely needed.
- ora_csid
-
Specify the integer OCI_ATTR_CHARSET_ID for the bind value. Character set names can't be used currently.
- ora_maxdata_size
-
Specify the integer OCI_ATTR_MAXDATA_SIZE for the bind value. May be needed if a character set conversion from client to server causes the data to use more space and so fail with a truncation error.
Trailing Spaces
The Oracle strips trailing spaces from VARCHAR placeholder values and uses Nonpadded Comparison Semantics with the result. This causes trouble if the spaces are needed for comparison with a CHAR value or to prevent the value from becoming '' which Oracle treats as NULL. Look for Blank-padded Comparison Semantics and Nonpadded Comparison Semantics in Oracle's SQL Reference or Server SQL Reference for more details.
Please remember that using spaces as a value or at the end of a value makes visually distinguishing values with different numbers of spaces difficult and should be avoided.
To preserve trailing spaces in placeholder values, either change the default placeholder type with "ora_ph_type" or the placeholder type for a particular call to "bind" in DBI or "bind_param_inout" in DBI with "ora_type" or TYPE
. Using ORA_CHAR with ora_type or SQL_CHAR
with TYPE
allows the placeholder to be used with Padded Comparison Semantics if the value it is being compared to is a CHAR, NCHAR, or literal.
Metadata
get_info()
DBD::Oracle supports get_info()
, but (currently) only a few info types.
table_info()
DBD::Oracle supports attributes for table_info()
.
In Oracle, the concept of user and schema is (currently) the same. Because database objects are owned by an user, the owner names in the data dictionary views correspond to schema names. Oracle does not support catalogs so TABLE_CAT is ignored as selection criterion.
Search patterns are supported for TABLE_SCHEM and TABLE_NAME.
TABLE_TYPE may contain a comma-separated list of table types. The following table types are supported:
TABLE
VIEW
SYNONYM
SEQUENCE
The result set is ordered by TABLE_TYPE, TABLE_SCHEM, TABLE_NAME.
The special enumerations of catalogs, schemas and table types are supported. However, TABLE_CAT is always NULL.
An identifier is passed as is, i.e. as the user provides or Oracle returns it. table_info()
performs a case-sensitive search. So, a selection criterion should respect upper and lower case. Normally, an identifier is case-insensitive. Oracle stores and returns it in upper case. Sometimes, database objects are created with quoted identifiers (for reserved words, mixed case, special characters, ...). Such an identifier is case-sensitive (if not all upper case). Oracle stores and returns it as given. table_info()
has no special quote handling, neither adds nor removes quotes.
primary_key_info()
Oracle does not support catalogs so TABLE_CAT is ignored as selection criterion. The TABLE_CAT field of a fetched row is always NULL (undef). See "table_info()" for more detailed information.
If the primary key constraint was created without an identifier, PK_NAME contains a system generated name with the form SYS_Cn.
The result set is ordered by TABLE_SCHEM, TABLE_NAME, KEY_SEQ.
An identifier is passed as is, i.e. as the user provides or Oracle returns it. See "table_info()" for more detailed information.
foreign_key_info()
This method (currently) supports the extended behavior of SQL/CLI, i.e. the result set contains foreign keys that refer to primary and alternate keys. The field UNIQUE_OR_PRIMARY distinguishes these keys.
Oracle does not support catalogs, so $pk_catalog
and $fk_catalog
are ignored as selection criteria (in the new style interface). The UK_TABLE_CAT and FK_TABLE_CAT fields of a fetched row are always NULL (undef). See "table_info()" for more detailed information.
If the primary or foreign key constraints were created without an identifier, UK_NAME or FK_NAME contains a system generated name with the form SYS_Cn.
The UPDATE_RULE field is always 3 ('NO ACTION'), because Oracle (currently) does not support other actions.
The DELETE_RULE field may contain wrong values. This is a known Bug (#1271663) in Oracle's data dictionary views. Currently (as of 8.1.7), 'RESTRICT' and 'SET DEFAULT' are not supported, 'CASCADE' is mapped correctly and all other actions (incl. 'SET NULL') appear as 'NO ACTION'.
The DEFERABILITY field is always NULL, because this columns is not present in the ALL_CONSTRAINTS view of older Oracle releases.
The result set is ordered by UK_TABLE_SCHEM, UK_TABLE_NAME, FK_TABLE_SCHEM, FK_TABLE_NAME, ORDINAL_POSITION.
An identifier is passed as is, i.e. as the user provides or Oracle returns it. See "table_info()" for more detailed information.
column_info()
Oracle does not support catalogs so TABLE_CAT is ignored as selection criterion. The TABLE_CAT field of a fetched row is always NULL (undef). See "table_info()" for more detailed information.
The CHAR_OCTET_LENGTH field is (currently) always NULL (undef).
Don't rely on the values of the BUFFER_LENGTH field! Especially the length of FLOATs may be wrong.
Datatype codes for non-standard types are subject to change.
Attention! The DATA_DEFAULT (COLUMN_DEF) column is of type LONG.
The result set is ordered by TABLE_SCHEM, TABLE_NAME, ORDINAL_POSITION.
An identifier is passed as is, i.e. as the user provides or Oracle returns it. See "table_info()" for more detailed information.
Unicode
DBD::Oracle now supports Unicode UTF-8. There are, however, a number of issues you should be aware of, so please read all this section carefully.
In this section we'll discuss "Perl and Unicode", then "Oracle and Unicode", and finally "DBD::Oracle and Unicode".
Information about Unicode in general can be found at: http://www.unicode.org/. It is well worth reading because there are many misconceptions about Unicode and you may be holding some of them.
Perl and Unicode
Perl began implementing Unicode with version 5.6, but the implementaion did not mature until version 5.8 and later. If you plan to use Unicode you are strongly urged to use perl 5.8.2 or later and to carefully read the perl documention on Unicode:
perldoc perluniintro # in perl 5.8 or later
perldoc perlunicode
And then read it again.
Perl's internal unicode format is UTF-8 which corresponds to the Oracle character set called AL32UTF8.
Oracle and Unicode
Oracle supports many characters sets, including several different forms of Unicode. These include:
AL16UTF16 => valid for NCHAR columns (CSID=2000)
UTF8 => valid for NCHAR columns (CSID=871), deprecated
AL32UTF8 => valid for NCHAR and CHAR columns (CSID=873)
When you create an Oracle database, you must specify the DATABASE character set (used for DDL, DML and CHAR datatypes) and the NATIONAL character set (used for NCHAR and NCLOB types). The character sets used in your database can be found using:
$hash_ref = $dbh->ora_nls_parameters()
$database_charset = $hash_ref->{NLS_CHARACTERSET};
$national_charset = $hash_ref->{NLS_NCHAR_CHARACTERSET};
The Oracle 9.2 and later default for the national character set is AL16UTF16. The default for the database character set is often US7ASCII. Although many experienced DBAs will consider an 8bit character set like WE8ISO8859P1 or WE8MSWIN1252. To use any character set with Oracle other than US7ASCII, requires that the NLS_LANG environment variable be set. See the " 8-bit text issues"" in "International NLS section below.
You are strongly urged to read the Oracle Internationalization documentation specifically with respect the choices and trade offs for creating a databases for use with international character sets.
Oracle uses the NLS_LANG environment variable to indicate what character set is being used on the client. When fetching data Oracle will convert from whatever the database character set is to the client character set specified by NLS_LANG. Similarly, when sending data to the database Oracle will convert from the character set specified by NLS_LANG to the database character set.
The NLS_NCHAR environment variable can be used to define a different character set for 'national' (NCHAR) character types.
Both UTF8 and AL32UTF32 can be used in NLS_LANG and NLS_NCHAR. For example:
NLS_LANG=AMERICAN_AMERICA.UTF8
NLS_LANG=AMERICAN_AMERICA.AL32UTF8
NLS_NCHAR=UTF8
NLS_NCHAR=AL32UTF8
Oracle 8 client libraries have a number of bugs related to character set handling, especially when connected to an Oracle 9+ server. For this reason a number of DBD::Oracle tests are disabled when using an Oracle 8 client. If you wish to use Unicode, I recommend upgrading client and server to Oracle 9 or later.
Oracle UTF8 is not UTF-8
AL32UTF8 should be used in preference to UTF8 if it works for you, which it should for Oracle 9.2 or later. If you're using an old version of Oracle that doesn't support AL32UTF8 then you should avoid using any Unicode characters that require surrogates, in other words characters beyond the Unicode BMP (Basic Multilingual Plane).
That's because the character set that Oracle calls "UTF8" doesn't conform to the UTF-8 standard in its handling of surrogate characters. Technically the encoding that Oracle calls "UTF8" is known as "CESU-8". Here are a couple of extracts from http://www.unicode.org/reports/tr26/:
CESU-8 is useful in 8-bit processing environments where binary
collation with UTF-16 is required. It is designed and recommended
for use only within products requiring this UTF-16 binary collation
equivalence. It is not intended nor recommended for open interchange.
As a very small percentage of characters in a typical data stream
are expected to be supplementary characters, there is a strong
possibility that CESU-8 data may be misinterpreted as UTF-8.
Therefore, all use of CESU-8 outside closed implementations is
strongly discouraged, such as the emittance of CESU-8 in output
files, markup language or other open transmission forms.
Oracle uses this internally because it collates (sorts) in the same order as UTF16, which is the basis of Oracle's internal collation definitions.
Rather than change UTF8 for clients Oracle chose to define a new character set called "AL32UTF8" which does conform to the UTF-8 standard. (The AL32UTF8 character set can't be used on the server because it would break collation.)
Because of that, for the rest of this document we'll use "AL32UTF8". If you're using an Oracle version below 9.2 you'll need to use "UTF8" until you upgrade.
DBD::Oracle and Unicode
DBD::Oracle Unicode support has been implemented for Oracle versions 9 or greater, and perl version 5.6 or greater (though we strongly suggest that you use perl 5.8.2 or later).
You can check which Oracle version your DBD::Oracle was built with by importing the ORA_OCI
constant from DBD::Oracle.
Fetching Data
Any data returned from Oracle to DBD::Oracle in the AL32UTF8 character set will be marked as UTF-8 to ensure correct handling by perl.
For Oracle to return data in the AL32UTF8 character set the NLS_LANG or NLS_NCHAR environment variable must be set as described in the previous section.
When fetching NCHAR, NVARCHAR, or NCLOB data from Oracle, DBD::Oracle will set the perl UTF-8 flag on the returned data if either NLS_NCHAR is AL32UTF8, or NLS_NCHAR is not set and NLS_LANG is AL32UTF8.
When fetching other character data from Oracle, DBD::Oracle will set the perl UTF-8 flag on the returned data if NLS_LANG is AL32UTF8.
Sending Data using Placeholders
Data bound to a placeholder is assumed to be in the default client character set (specified by NLS_LANG) except for a few special cases. These are listed here with the highest precedence first:
If the ora_csid
attribute is given to bind_param() then that is passed to Oracle and takes precedence.
If the value is a Perl Unicode string (UTF-8) then DBD::Oracle ensures that Oracle uses the Unicode character set, regardless of the NLS_LANG and NLS_NCHAR settings.
If the placeholder is for inserting an NCLOB then the client NLS_NCHAR character set is used. (That's useful but inconsistent with the other behaviour so may change. Best to be explicit by using the ora_csform
attribute.)
If the ora_csform
attribute is given to bind_param() then that determines if the value should be assumed to be in the default (NLS_LANG) or NCHAR (NLS_NCHAR) client character set.
use DBD::Oracle qw( SQLCS_IMPLICIT SQLCS_NCHAR );
...
$sth->bind_param(1, $value, { ora_csform => SQLCS_NCHAR });
or
$dbh->{ora_ph_csform} = SQLCS_NCHAR; # default for all future placeholders
Sending Data using SQL
Oracle assumes the SQL statement is in the default client character set (as specified by NLS_LANG). So Unicode strings containing non-ASCII characters should not be used unless the default client character set is AL32UTF8.
DBD::Oracle and Other Character Sets and Encodings
The only multi-byte Oracle character set supported by DBD::Oracle is "AL32UTF8" (and "UTF8"). Single-byte character sets should work well.
Other Data Types
DBD::Oracle does not explicitly support most Oracle data types. It simply asks Oracle to return them as strings and Oracle does so. Mostly. Similarly when binding placeholder values DBD::Oracle binds them as strings and Oracle converts them to the appropriate type, such as DATE, when used.
Some of these automatic conversions to and from strings use NLS setings to control the formating for output and the parsing for input. The most common example is the DATE type. The default NLS format for DATE might be DD-MON-YYYY and so when a DATE type is fetched that's how Oracle will format the date. NLS settings also control the default parsing of strings into DATE values. An error will be generated if the contents of the string don't match the NLS format. If you're dealing in dates which don't match the default NLS format then you can either change the default NLS format or, more commonly, use TO_CHAR(field, "format") and TO_DATE(?, "format") to explicitly specify formats for converting to and from strings.
A slightly more subtle problem can occur with NUMBER types. The default NLS settings might format numbers with a fullstop (".
") to separate thousands and a comma (",
") as the decimal point. Perl will generate warnings and use incorrect values when numbers, returned and formatted as strings in this way by Oracle, are used in a numeric context. You could explicitly convert each numeric value using the TO_CHAR(...) function but that gets tedious very quickly. The best fix is to change the NLS settings. That can be done for an individual connection by doing:
$dbh->do("ALTER SESSION SET NLS_NUMERIC_CHARACTERS = '.,'");
There are some types, like BOOLEAN, that Oracle does not automatically convert to or from strings (pity). These need to be converted explicitly using SQL or PL/SQL functions.
Examples:
# DATE values
my $sth0 = $dbh->prepare( <<SQL_END );
SELECT username, TO_CHAR( created, ? )
FROM all_users
WHERE created >= TO_DATE( ?, ? )
SQL_END
$sth0->execute( 'YYYY-MM-DD HH24:MI:SS', "2003", 'YYYY' );
# BOOLEAN values
my $sth2 = $dbh->prepare( <<PLSQL_END );
DECLARE
b0 BOOLEAN;
b1 BOOLEAN;
o0 VARCHAR2(32);
o1 VARCHAR2(32);
FUNCTION to_bool( i VARCHAR2 ) RETURN BOOLEAN IS
BEGIN
IF i IS NULL THEN RETURN NULL;
ELSIF i = 'F' OR i = '0' THEN RETURN FALSE;
ELSE RETURN TRUE;
END IF;
END;
FUNCTION from_bool( i BOOLEAN ) RETURN NUMBER IS
BEGIN
IF i IS NULL THEN RETURN NULL;
ELSIF i THEN RETURN 1;
ELSE RETURN 0;
END IF;
END;
BEGIN
-- Converting values to BOOLEAN
b0 := to_bool( :i0 );
b1 := to_bool( :i1 );
-- Converting values from BOOLEAN
:o0 := from_bool( b0 );
:o1 := from_bool( b1 );
END;
PLSQL_END
my ( $i0, $i1, $o0, $o1 ) = ( "", "Something else" );
$sth2->bind_param( ":i0", $i0 );
$sth2->bind_param( ":i1", $i1 );
$sth2->bind_param_inout( ":o0", \$o0, 32 );
$sth2->bind_param_inout( ":o1", \$o1, 32 );
$sth2->execute();
foreach ( $i0, $b0, $o0, $i1, $b1, $o1 ) {
$_ = "(undef)" if ! defined $_;
}
print "$i0 to $o0, $i1 to $o1\n";
# Result is : "'' to '(undef)', 'Something else' to '1'"
PL/SQL Examples
Most of these PL/SQL examples come from: Eric Bartley <bartley@cc.purdue.edu>.
/*
* PL/SQL to create package with stored procedures invoked by
* Perl examples. Execute using sqlplus.
*
* Use of "... OR REPLACE" prevents failure in the event that the
* package already exists.
*/
CREATE OR REPLACE PACKAGE plsql_example
IS
PROCEDURE proc_np;
PROCEDURE proc_in (
err_code IN NUMBER
);
PROCEDURE proc_in_inout (
test_num IN NUMBER,
is_odd IN OUT NUMBER
);
FUNCTION func_np
RETURN VARCHAR2;
END plsql_example;
/
CREATE OR REPLACE PACKAGE BODY plsql_example
IS
PROCEDURE proc_np
IS
whoami VARCHAR2(20) := NULL;
BEGIN
SELECT USER INTO whoami FROM DUAL;
END;
PROCEDURE proc_in (
err_code IN NUMBER
)
IS
BEGIN
RAISE_APPLICATION_ERROR(err_code, 'This is a test.');
END;
PROCEDURE proc_in_inout (
test_num IN NUMBER,
is_odd IN OUT NUMBER
)
IS
BEGIN
is_odd := MOD(test_num, 2);
END;
FUNCTION func_np
RETURN VARCHAR2
IS
ret_val VARCHAR2(20);
BEGIN
SELECT USER INTO ret_val FROM DUAL;
RETURN ret_val;
END;
END plsql_example;
/
/* End PL/SQL for example package creation. */
use DBI;
my($db, $csr, $ret_val);
$db = DBI->connect('dbi:Oracle:database','user','password')
or die "Unable to connect: $DBI::errstr";
# So we don't have to check every DBI call we set RaiseError.
# See the DBI docs now if you're not familiar with RaiseError.
$db->{RaiseError} = 1;
# Example 1 Eric Bartley <bartley@cc.purdue.edu>
#
# Calling a PLSQL procedure that takes no parameters. This shows you the
# basic's of what you need to execute a PLSQL procedure. Just wrap your
# procedure call in a BEGIN END; block just like you'd do in SQL*Plus.
#
# p.s. If you've used SQL*Plus's exec command all it does is wrap the
# command in a BEGIN END; block for you.
$csr = $db->prepare(q{
BEGIN
PLSQL_EXAMPLE.PROC_NP;
END;
});
$csr->execute;
# Example 2 Eric Bartley <bartley@cc.purdue.edu>
#
# Now we call a procedure that has 1 IN parameter. Here we use bind_param
# to bind out parameter to the prepared statement just like you might
# do for an INSERT, UPDATE, DELETE, or SELECT statement.
#
# I could have used positional placeholders (e.g. :1, :2, etc.) or
# ODBC style placeholders (e.g. ?), but I prefer Oracle's named
# placeholders (but few DBI drivers support them so they're not portable).
my $err_code = -20001;
$csr = $db->prepare(q{
BEGIN
PLSQL_EXAMPLE.PROC_IN(:err_code);
END;
});
$csr->bind_param(":err_code", $err_code);
# PROC_IN will RAISE_APPLICATION_ERROR which will cause the execute to 'fail'.
# Because we set RaiseError, the DBI will croak (die) so we catch that with eval.
eval {
$csr->execute;
};
print 'After proc_in: $@=',"'$@', errstr=$DBI::errstr, ret_val=$ret_val\n";
# Example 3 Eric Bartley <bartley@cc.purdue.edu>
#
# Building on the last example, I've added 1 IN OUT parameter. We still
# use a placeholders in the call to prepare, the difference is that
# we now call bind_param_inout to bind the value to the place holder.
#
# Note that the third parameter to bind_param_inout is the maximum size
# of the variable. You normally make this slightly larger than necessary.
# But note that the perl variable will have that much memory assigned to
# it even if the actual value returned is shorter.
my $test_num = 5;
my $is_odd;
$csr = $db->prepare(q{
BEGIN
PLSQL_EXAMPLE.PROC_IN_INOUT(:test_num, :is_odd);
END;
});
# The value of $test_num is _copied_ here
$csr->bind_param(":test_num", $test_num);
$csr->bind_param_inout(":is_odd", \$is_odd, 1);
# The execute will automagically update the value of $is_odd
$csr->execute;
print "$test_num is ", ($is_odd) ? "odd - ok" : "even - error!", "\n";
# Example 4 Eric Bartley <bartley@cc.purdue.edu>
#
# What about the return value of a PLSQL function? Well treat it the same
# as you would a call to a function from SQL*Plus. We add a placeholder
# for the return value and bind it with a call to bind_param_inout so
# we can access it's value after execute.
my $whoami = "";
$csr = $db->prepare(q{
BEGIN
:whoami := PLSQL_EXAMPLE.FUNC_NP;
END;
});
$csr->bind_param_inout(":whoami", \$whoami, 20);
$csr->execute;
print "Your database user name is $whoami\n";
$db->disconnect;
You can find more examples in the t/plsql.t file in the DBD::Oracle source directory.
Oracle 9.2 appears to have a bug where a variable bound with bind_param_inout() that isn't assigned to by the executed PL/SQL block may contain garbage. See http://www.mail-archive.com/dbi-users@perl.org/msg18835.html
Private database handle functions
These functions are called through the method func() which is described in the DBI documentation.
plsql_errstr
This function returns a string which describes the errors from the most recent PL/SQL function, procedure, package, or package body compile in a format similar to the output of the SQL*Plus command 'show errors'.
The function returns undef if the error string could not be retrieved due to a database error. Look in $dbh->errstr for the cause of the failure.
If there are no compile errors, an empty string is returned.
Example:
# Show the errors if CREATE PROCEDURE fails
$dbh->{RaiseError} = 0;
if ( $dbh->do( q{
CREATE OR REPLACE PROCEDURE perl_dbd_oracle_test as
BEGIN
PROCEDURE filltab( stuff OUT TAB ); asdf
END; } ) ) {} # Statement succeeded
}
elsif ( 6550 != $dbh->err ) { die $dbh->errstr; } # Utter failure
else {
my $msg = $dbh->func( 'plsql_errstr' );
die $dbh->errstr if ! defined $msg;
die $msg if $msg;
}
dbms_output_enable / dbms_output_put / dbms_output_get
These functions use the PL/SQL DBMS_OUTPUT package to store and retrieve text using the DBMS_OUTPUT buffer. Text stored in this buffer by dbms_output_put or any PL/SQL block can be retrieved by dbms_output_get or any PL/SQL block connected to the same database session.
Stored text is not available until after dbms_output_put or the PL/SQL block that saved it completes its execution. This means you CAN NOT use these functions to monitor long running PL/SQL procedures.
Example 1:
# Enable DBMS_OUTPUT and set the buffer size
$dbh->{RaiseError} = 1;
$dbh->func( 1000000, 'dbms_output_enable' );
# Put text in the buffer . . .
$dbh->func( @text, 'dbms_output_put' );
# . . . and retreive it later
@text = $dbh->func( 'dbms_output_get' );
Example 2:
$dbh->{RaiseError} = 1;
$sth = $dbh->prepare(q{
DECLARE tmp VARCHAR2(50);
BEGIN
SELECT SYSDATE INTO tmp FROM DUAL;
dbms_output.put_line('The date is '||tmp);
END;
});
$sth->execute;
# retreive the string
$date_string = $dbh->func( 'dbms_output_get' );
- dbms_output_enable ( [ buffer_size ] )
-
This function calls DBMS_OUTPUT.ENABLE to enable calls to package DBMS_OUTPUT procedures GET, GET_LINE, PUT, and PUT_LINE. Calls to these procedures are ignored unless DBMS_OUTPUT.ENABLE is called first.
The buffer_size is the maximum amount of text that can be saved in the buffer and must be between 2000 and 1,000,000. If buffer_size is not given, the default is 20,000 bytes.
- dbms_output_put ( [ @lines ] )
-
This function calls DBMS_OUTPUT.PUT_LINE to add lines to the buffer.
If all lines were saved successfully the function returns 1. Depending on the context, an empty list or undef is returned for failure.
If any line causes buffer_size to be exceeded, a buffer overflow error is raised and the function call fails. Some of the text might be in the buffer.
- dbms_output_get
-
This function calls DBMS_OUTPUT.GET_LINE to retrieve lines of text from the buffer.
In an array context, all complete lines are removed from the buffer and returned as a list. If there are no complete lines, an empty list is returned.
In a scalar context, the first complete line is removed from the buffer and returned. If there are no complete lines, undef is returned.
Any text in the buffer after a call to DBMS_OUTPUT.GET_LINE or DBMS_OUTPUT.GET is discarded by the next call to DBMS_OUTPUT.PUT_LINE, DBMS_OUTPUT.PUT, or DBMS_OUTPUT.NEW_LINE.
- reauthenticate ( $username, $password )
-
Starts a new session against the current database using the credentials supplied.
- ora_nls_parameters ( [ $refresh ] )
-
Returns a hash reference containing the current NLS parameters, as given by the v$nls_parameters view. The values fetched are cached between calls. To cause the latest values to be fetched, pass a true value to the function.
- ora_can_unicode ( [ $refresh ] )
-
Returns a number indicating whether either of the database character sets is a Unicode encoding. Calls ora_nls_parameters() and passes the optional $refresh parameter to it.
0 = Neither character set is a Unicode encoding.
1 = National character set is a Unicode encoding.
2 = Database character set is a Unicode encoding.
3 = Both character sets are Unicode encodings.
Prepare postponed till execute
The DBD::Oracle module can avoid an explicit 'describe' operation prior to the execution of the statement unless the application requests information about the results (such as $sth->{NAME}). This reduces communication with the server and increases performance (reducing the number of PARSE_CALLS inside the server).
However, it also means that SQL errors are not detected until execute()
(or $sth->{NAME} etc) is called instead of when prepare()
is called. Note that if the describe is triggered by the use of $sth->{NAME} or a similar attribute and the describe fails then an exception is thrown even if RaiseError
is false!
Set "ora_check_sql" to 0 in prepare() to enable this behaviour.
Handling LOBs
Simple Usage
The value of an Oracle LOB column is not the content of the LOB. It's a 'LOB Locator' which, after being selected or inserted needs extra processing to read or write the content of the LOB.
When fetching LOBs they are, by default, made to look just like LONGs and are subject to the LongReadLen and LongTruncOk attributes. Note that with OCI 7 DBD::Oracle pre-allocates the whole buffer (LongReadLen) at the time the statement is prepared. With OCI 8+ it grows the buffer to the amount needed for the largest LOB to be fetched so far.
When inserting or updating LOBs some major magic has to be performed behind the scenes to make it transparent. Basically the driver has to insert a 'LOB Locator' and then refetch the newly inserted LOB Locator before being able to write the data into it. However, it works well most of the time, and I've made it as fast as possible, just one extra server-round-trip per insert or update after the first. For the time being, only single-row LOB updates are supported.
To insert or update a large LOB using a placeholder, DBD::Oracle has to know in advance that it is a LOB type. So you need to say:
$sth->bind_param($field_num, $lob_value, { ora_type => ORA_CLOB });
The ORA_CLOB and ORA_BLOB constants can be imported using
use DBD::Oracle qw(:ora_types);
or use the corresponding integer values (112 and 113).
One further wrinkle: for inserts and updates of LOBs, DBD::Oracle has to be able to tell which parameters relate to which table fields. In all cases where it can possibly work it out for itself, it does, however, if there are multiple LOB fields of the same type in the table then you need to tell it which field each LOB param relates to:
$sth->bind_param($idx, $value, { ora_type=>ORA_CLOB, ora_field=>'foo' });
There are some limitations inherent in the way DBD::Oracle makes typical LOB operations simple by hiding the LOB Locator processing:
- Can't pass LOBs to/from PL/SQL (except as strings less than 32KB)
- Can't read/write LOBs in chunks (except via DBMS_LOB.WRITEAPPEND in PL/SQL)
- To INSERT a LOB, you need UPDATE privilege.
The alternative is to disable the automatic LOB Locator processing. If "ora_auto_lob" is 0 in prepare(), you can fetch the LOB Locators and do all the work yourself using the ora_lob_*() methods and/or Oracle::OCI. See the "LOB Methods" section below.
LOB Locator Methods
The following driver-specific methods let you manipulate "LOB Locators". LOB locators can be selected from tables directly, if the ora_auto_lob
attribute is false, or returned via PL/SQL procedure calls.
(If using a DBI version earlier than 1.36 they must be called via the func() method. Note that methods called via func() don't honour RaiseError etc, and so it's important to check $dbh->err after each call. It's recommended that you upgrade to DBI 1.38 or later.)
Warning: Currently multi-byte character set issues have not been fully worked out. So these methods may not do what you expect if either the perl data is utf8 or the CLOB is a multi-byte character set (including uft8). The current behaviour in these situations may not be correct and is subject to change. Testing and patches are most welcome.
- ora_lob_read
-
$data = $dbh->ora_lob_read($lob_locator, $offset, $length);
Read a portion of the LOB. $offset starts at 1. Uses the Oracle OCILobRead function.
- ora_lob_write
-
$rc = $dbh->ora_lob_write($lob_locator, $offset, $data);
Write/overwrite a portion of the LOB. $offset starts at 1. Uses the Oracle OCILobWrite function.
- ora_lob_append
-
$rc = $dbh->ora_lob_append($lob_locator, $data);
Append $data to the LOB. Uses the Oracle OCILobWriteAppend function.
NOTE: This method should not be used if either the client or the server are Oracle version 8 due to Oracle bug #886191.
- ora_lob_trim
-
$rc = $dbh->ora_lob_trim($lob_locator, $length);
Trims the length of the LOB to $length. Uses the Oracle OCILobTrim function.
- ora_lob_length
-
$length = $dbh->ora_lob_length($lob_locator);
Returns the length of the LOB. Uses the Oracle OCILobGetLength function.
LOB Locator Method Examples
Note: Make sure you first read the note in the section above about multi-byte character set issues with these methods.
The following examples demonstrate the usage of LOB Locators to read, write, and append data, and to query the size of large data.
The following examples assume a table containing two large object columns, one binary and one character, with a primary key column, defined as follows:
CREATE TABLE lob_example (
lob_id INTEGER PRIMARY KEY,
bindata BLOB,
chardata CLOB
)
It also assumes a sequence for use in generating unique lob_id field values, defined as follows:
CREATE SEQUENCE lob_example_seq
Example: Inserting a new row with large data
Unless enough memory is available to store and bind the entire lob data for insert all at once, the lob columns must be written iteratively, piece by piece. In the case of a new row, this is performed by first inserting a row, with empty values in the lob columns, then modifying the row by writing the large data iteratively to the lob columns using their LOB locators as handles.
The insert statement must create token values in the lob columns. Here, we use the empty string for both the binary and character large object columns 'bindata' and 'chardata'.
After the INSERT statement, a SELECT statement is used to acquire lob locators to the 'bindata' and 'chardata' fields of the newly inserted row. Because these lob locators are subsequently written, they must be acquired from a select statement containing the clause 'FOR UPDATE' (lob locators are only valid within the transaction that fetched them, so can't be used effectively if AutoCommit is enabled).
my $lob_id = $dbh->selectrow_array( <<" SQL" );
SELECT lob_example_seq.nextval FROM DUAL
SQL
my $sth = $dbh->prepare( <<" SQL" );
INSERT INTO lob_example
( lob_id, bindata, chardata )
VALUES ( ?, EMPTY_BLOB(),EMPTY_CLOB() )
SQL
$sth->execute( $lob_id );
$sth = $dbh->prepare( <<" SQL", { ora_auto_lob => 0 } );
SELECT bindata, chardata
FROM lob_example
WHERE lob_id = ?
FOR UPDATE
SQL
$sth->execute( $lob_id );
my ( $bin_locator, $char_locator ) = $sth->fetchrow_array();
$sth->finish();
open BIN_FH, "/binary/data/source" or die;
open CHAR_FH, "/character/data/source" or die;
my $chunk_size = 4096; # Arbitrary chunk size
# BEGIN WRITING BIN_DATA COLUMN
my $offset = 1; # Offsets start at 1, not 0
my $length = 0;
my $buffer = '';
while( $length = read( BIN_FH, $buffer, $chunk_size ) ) {
$dbh->ora_lob_write( $bin_locator, $offset, $buffer );
$offset += $length;
}
# BEGIN WRITING CHAR_DATA COLUMN
$offset = 1; # Offsets start at 1, not 0
$length = 0;
$buffer = '';
while( $length = read( CHAR_FH, $buffer, $chunk_size ) ) {
$dbh->ora_lob_write( $char_locator, $offset, $buffer );
$offset += $length;
}
In this example we demonstrate the use of ora_lob_write() iteratively to append data to the columns 'bin_data' and 'char_data'. Had we used ora_lob_append(), we could have saved ourselves the trouble of keeping track of the offset into the lobs. The snippet of code beneath the comment 'BEGIN WRITING BIN_DATA COLUMN' could look as follows:
my $buffer = '';
while ( read( BIN_FH, $buffer, $chunk_size ) ) {
$dbh->ora_lob_append( $bin_locator, $buffer );
}
The scalar variables $offset and $length are no longer needed, because ora_lob_append() keeps track of the offset for us.
Example: Updating an existing row with large data
In this example, we demonstrate a technique for overwriting a portion of a blob field with new binary data. The blob data before and after the section overwritten remains unchanged. Hence, this technique could be used for updating fixed length subfields embedded in a binary field.
my $lob_id = 5; # Arbitrary row identifier, for example
$sth = $dbh->prepare( <<" SQL", { ora_auto_lob => 0 } );
SELECT bindata
FROM lob_example
WHERE lob_id = ?
FOR UPDATE
SQL
$sth->execute( $lob_id );
my ( $bin_locator ) = $sth->fetchrow_array();
my $offset = 100234;
my $data = "This string will overwrite a portion of the blob";
$dbh->ora_lob_write( $bin_locator, $offset, $data );
After running this code, the row where lob_id = 5 will contain, starting at position 100234 in the bin_data column, the string "This string will overwrite a portion of the blob".
Example: Streaming character data from the database
In this example, we demonstrate a technique for streaming data from the database to a file handle, in this case STDOUT. This allows more data to be read in and written out than could be stored in memory at a given time.
my $lob_id = 17; # Arbitrary row identifier, for example
$sth = $dbh->prepare( <<" SQL", { ora_auto_lob => 0 } );
SELECT chardata
FROM lob_example
WHERE lob_id = ?
SQL
$sth->execute( $lob_id );
my ( $char_locator ) = $sth->fetchrow_array();
my $chunk_size = 1034; # Arbitrary chunk size, for example
my $offset = 1; # Offsets start at 1, not 0
while( my $data = $dbh->ora_lob_read( $char_locator, $offset, $chunk_size ) ) {
print STDOUT $data;
$offset += $chunk_size;
}
Notice that the select statement does not contain the phrase "FOR UPDATE". Because we are only reading from the lob locator returned, and not modifying the lob it refers to, the select statement does not require the "FOR UPDATE" clause.
Example: Truncating existing large data
In this example, we truncate the data already present in a large object column in the database. Specifically, for each row in the table, we truncate the 'bindata' value to half its previous length.
After acquiring a lob locator for the column, we query its length, then we trim the length by half. Because we modify the large objects with the call to ora_lob_trim(), we must select the lob locators 'FOR UPDATE'.
my $sth = $dbh->prepare( <<" SQL", { ora_auto_lob => 0 } );
SELECT bindata
FROM lob_example
FOR UPATE
SQL
$sth->execute();
while( my ( $bin_locator ) = $sth->fetchrow_array() ) {
my $binlength = $dbh->ora_lob_length( $bin_locator );
if( $binlength > 0 ) {
$dbh->ora_lob_trim( $bin_locator, $binlength/2 );
}
}
Binding Cursors
Cursors can be returned from PL/SQL blocks, either from stored functions (or procedures with OUT parameters) or from direct OPEN
statements, as shown below:
use DBI;
use DBD::Oracle qw(:ora_types);
my $dbh = DBI->connect(...);
my $sth1 = $dbh->prepare(q{
BEGIN OPEN :cursor FOR
SELECT table_name, tablespace_name
FROM user_tables WHERE tablespace_name = :space;
END;
});
$sth1->bind_param(":space", "USERS");
my $sth2;
$sth1->bind_param_inout(":cursor", \$sth2, 0, { ora_type => ORA_RSET } );
$sth1->execute;
# $sth2 is now a valid DBI statement handle for the cursor
while ( my @row = $sth2->fetchrow_array ) { ... }
The only special requirement is the use of bind_param_inout()
with an attribute hash parameter that specifies ora_type
as ORA_RSET
. If you don't do that you'll get an error from the execute()
like: "ORA-06550: line X, column Y: PLS-00306: wrong number or types of arguments in call to ...".
Here's an alternative form using a function that returns a cursor. This example uses the pre-defined weak (or generic) REF CURSOR type SYS_REFCURSOR. This is an Oracle 9 feature. For Oracle 8, you must create your own REF CURSOR type in a package (see the curref.pl
script mentioned at the end of this section).
# Create the function that returns a cursor
$dbh->do(q{
CREATE OR REPLACE FUNCTION sp_ListEmp RETURN SYS_REFCURSOR
AS l_cursor SYS_REFCURSOR;
BEGIN
OPEN l_cursor FOR select ename, empno from emp
ORDER BY ename;
RETURN l_cursor;
END;
});
# Use the function that returns a cursor
my $sth1 = $dbh->prepare(q{BEGIN :cursor := sp_ListEmp; END;});
my $sth2;
$sth1->bind_param_inout(":cursor", \$sth2, 0, { ora_type => ORA_RSET } );
$sth1->execute;
# $sth2 is now a valid DBI statement handle for the cursor
while ( my @row = $sth2->fetchrow_array ) { ... }
A cursor obtained from PL/SQL as above may be passed back to PL/SQL by binding for input, as shown in this example, which explicitly closes a cursor:
my $sth3 = $dbh->prepare("BEGIN CLOSE :cursor; END;");
$sth3->bind_param(":cursor", $sth2, { ora_type => ORA_RSET } );
$sth3->execute;
It is not normally necessary to close a cursor explicitly in this way. Oracle will close the cursor automatically at the first client-server interaction after the cursor statement handle is destroyed. An explicit close may be desirable if the reference to the cursor handle from the PL/SQL statement handle delays the destruction of the cursor handle for too long. This reference remains until the PL/SQL handle is re-bound, re-executed or destroyed.
See the curref.pl
script in the Oracle.ex directory in the DBD::Oracle source distribution for a complete working example.
Fetching Nested Cursors
Oracle supports the use of select list expressions of type REF CURSOR. These may be explicit cursor expressions - CURSOR(SELECT ...)
, or calls to PL/SQL functions which return REF CURSOR values. The values of these expressions are known as nested cursors.
The value returned to a Perl program when a nested cursor is fetched is a statement handle. This statement handle is ready to be fetched from. It should not (indeed, must not) be executed.
Oracle imposes a restriction on the order of fetching when nested cursors are used. Suppose $sth1
is a handle for a select statement involving nested cursors, and $sth2
is a nested cursor handle fetched from $sth1
. $sth2
can only be fetched from while $sth1
is still active, and the row containing $sth2
is still current in $sth1
. Any attempt to fetch another row from $sth1
renders all nested cursor handles previously fetched from $sth1
defunct.
Fetching from such a defunct handle results in an error with the message ERROR nested cursor is defunct (parent row is no longer current)
.
This means that the fetchall...
or selectall...
methods are not useful for queries returning nested cursors. By the time such a method returns, all the nested cursor handles it has fetched will be defunct.
It is necessary to use an explicit fetch loop, and to do all the fetching of nested cursors within the loop, as the following example shows:
use DBI;
my $dbh = DBI->connect(...);
my $sth = $dbh->prepare(q{
SELECT dname, CURSOR(
SELECT ename FROM emp
WHERE emp.deptno = dept.deptno
ORDER BY ename
) FROM dept ORDER BY dname
});
$sth->execute;
while ( my ($dname, $nested) = $sth->fetchrow_array ) {
print "$dname\n";
while ( my ($ename) = $nested->fetchrow_array ) {
print " $ename\n";
}
}
The cursor returned by the function sp_ListEmp
defined in the previous section can be fetched as a nested cursor as follows:
my $sth = $dbh->prepare(q{SELECT sp_ListEmp FROM dual});
$sth->execute;
my ($nested) = $sth->fetchrow_array;
while ( my @row = $nested->fetchrow_array ) { ... }
Pre-fetching Nested Cursors
By default, DBD::Oracle pre-fetches rows in order to reduce the number of round trips to the server. For queries which do not involve nested cursors, the number of pre-fetched rows is controlled by the DBI database handle attribute RowCacheSize
(q.v.).
In Oracle, server side open cursors are a controlled resource, limited in number, on a per session basis, to the value of the initialization parameter OPEN_CURSORS
. Nested cursors count towards ths limit. Each nested cursor in the current row counts 1, as does each nested cursor in a pre-fetched row. Defunct nested cursors do not count.
An Oracle specific database handle attribute, ora_max_nested_cursors
, further controls pre-fetching for queries involving nested cursors. For each statement handle, the total number of nested cursors in pre-fetched rows is limited to the value of this parameter. The default value is 0, which disables pre-fetching for queries involving nested cursors.
Returning A Value from an INSERT
Oracle supports an extended SQL insert syntax which will return one or more of the values inserted. This can be particularly useful for single-pass insertion of values with re-used sequence values (avoiding a separate "select seq.nextval from dual" step).
$sth = $dbh->prepare(qq{
INSERT INTO foo (id, bar)
VALUES (foo_id_seq.nextval, :bar)
RETURNING id INTO :id
});
$sth->bind_param(":bar", 42);
$sth->bind_param_inout(":id", \my $new_id, 99);
$sth->execute;
print "The id of the new record is $new_id\n";
If you have many columns to bind you can use code like this:
@params = (... column values for record to be inserted ...);
$sth->bind_param($_, $params[$_-1]) for (1..@params);
$sth->bind_param_inout(@params+1, \my $new_id, 99);
$sth->execute;
Returning A Recordset
DBD::Oracle does not currently support binding a PL/SQL table (aka array) as an IN OUT parameter to any Perl data structure. You cannot therefore call a PL/SQL function or procedure from DBI that uses a non-atomic data type as either a parameter, or a return value. However, if you are using Oracle 9.0.1 or later, you can make use of table (or pipelined) functions.
For example, assume you have the existing PL/SQL Package :
CREATE OR REPLACE PACKAGE Array_Example AS -- TYPE tRec IS RECORD ( Col1 NUMBER, Col2 VARCHAR2 (10), Col3 DATE) ; -- TYPE taRec IS TABLE OF tRec INDEX BY BINARY_INTEGER ; -- FUNCTION Array_Func RETURN taRec ; -- END Array_Example ;
CREATE OR REPLACE PACKAGE BODY Array_Example AS -- FUNCTION Array_Func RETURN taRec AS -- l_Ret taRec ; -- BEGIN FOR i IN 1 .. 5 LOOP l_Ret (i).Col1 := i ; l_Ret (i).Col2 := 'Row : ' || i ; l_Ret (i).Col3 := TRUNC (SYSDATE) + i ; END LOOP ; RETURN l_Ret ; END ; -- END Array_Example ; /
Currently, there is no way to directly call the function Array_Example.Array_Func from DBI. However, by making the following relatively painless additions, its not only possible, but extremely efficient.
First, you need to create database object types that correspond to the record and table types in the package. From the above example, these would be :
CREATE OR REPLACE TYPE tArray_Example__taRec
AS OBJECT (
Col1 NUMBER,
Col2 VARCHAR2 (10),
Col3 DATE
) ;
CREATE OR REPLACE TYPE taArray_Example__taRec
AS TABLE OF tArray_Example__taRec ;
Now, assuming the existing function needs to remain unchanged (it is probably being called from other PL/SQL code), we need to add a new function to the package. Here's the new package specification and body :
CREATE OR REPLACE PACKAGE Array_Example AS
--
TYPE tRec IS RECORD (
Col1 NUMBER,
Col2 VARCHAR2 (10),
Col3 DATE) ;
--
TYPE taRec IS TABLE OF tRec INDEX BY BINARY_INTEGER ;
--
FUNCTION Array_Func RETURN taRec ;
FUNCTION Array_Func_DBI RETURN taArray_Example__taRec PIPELINED ;
--
END Array_Example ;
CREATE OR REPLACE PACKAGE BODY Array_Example AS
--
FUNCTION Array_Func RETURN taRec AS
l_Ret taRec ;
BEGIN
FOR i IN 1 .. 5 LOOP
l_Ret (i).Col1 := i ;
l_Ret (i).Col2 := 'Row : ' || i ;
l_Ret (i).Col3 := TRUNC (SYSDATE) + i ;
END LOOP ;
RETURN l_Ret ;
END ;
FUNCTION Array_Func_DBI RETURN taArray_Example__taRec PIPELINED AS
l_Set taRec ;
BEGIN
l_Set := Array_Func ;
FOR i IN l_Set.FIRST .. l_Set.LAST LOOP
PIPE ROW (
tArray_Example__taRec (
l_Set (i).Col1,
l_Set (i).Col2,
l_Set (i).Col3
)
) ;
END LOOP ;
RETURN ;
END ;
--
END Array_Example ;
As you can see, the new function is very simple. Now, it is a simple matter of calling the function as a straight-forward SELECT from your DBI code. From the above example, the code would look something like this :
my $sth = $dbh->prepare('SELECT * FROM TABLE(Array_Example.Array_Func_DBI)');
$sth->execute;
while ( my ($col1, $col2, $col3) = $sth->fetchrow_array {
...
}
Timezones
If TWO_TASK isn't set, Oracle uses the TZ variable from the local environment.
If TWO_TASK IS set, Oracle uses the TZ variable of the listener process running on the server.
You could have multiple listeners, each with their own TZ, and assign users to the appropriate listener by setting TNS_ADMIN to a directory that contains a tnsnames.ora file that points to the port that their listener is on.
[Brad Howerter, who supplied this info said: "I've done this to simulate running a perl script at the end of the previous month even though it was the 6th of the new month. I had the dba start up a listener with TZ=X+144. (144 hours = 6 days)"]
Oracle Related Links
Oracle on Linux
http://www.datamgmt.com/maillist.html
http://www.eGroups.com/list/oracle-on-linux
http://www.wmd.de/wmd/staff/pauck/misc/oracle_on_linux.html
ftp://oracle-ftp.oracle.com/server/patch_sets/LINUX
http://www.ixora.com.au/
Free Oracle Tools and Links
ora_explain supplied and installed with DBD::Oracle.
http://www.orafaq.com/
http://vonnieda.org/oracletool/
Commercial Oracle Tools and Links
Assorted tools and references for general information. No recommendation implied.
http://www.platinum.com/products/oracle.htm
http://www.SoftTreeTech.com
http://www.databasegroup.com
Also PL/Vision from RevealNet and Steven Feuerstein, and "Q" from Savant Corporation.
SEE ALSO
http://search.cpan.org/~timb/DBD-Oracle/MANIFEST for all files in the DBD::Oracle source distribution including the examples in the Oracle.ex directory
http://search.cpan.org/search?query=Oracle&mode=dist
AUTHOR
DBD::Oracle by Tim Bunce. DBI by Tim Bunce.
ACKNOWLEDGEMENTS
A great many people have helped me with DBD::Oracle over the 12 years between 1994 and 2006. Far too many to name, but I thank them all. Many are named in the Changes file.
See also "ACKNOWLEDGEMENTS" in DBI.
MAINTAINER
As of release 1.17 in February 2006 The Pythian Group, Inc. (http://www.pythian.com) are taking the lead in maintaining DBD::Oracle with my assistance and gratitude. That frees more of my time to work on DBI for perl 5 and perl 6.
COPYRIGHT
The DBD::Oracle module is Copyright (c) 1994-2006 Tim Bunce. Ireland.
The DBD::Oracle module is free open source software; you can redistribute it and/or modify it under the same terms as Perl 5.
CONTRIBUTING
If you'd like DBD::Oracle to do something new or different the best way to make that happen is to do it yourself and email to dbi-dev@perl.org a patch of the source code (using 'diff' - see below) that shows the changes.
How to create a patch using Subversion
The DBD::Oracle source code is maintained using Subversion (a replacement for CVS, see http://subversion.tigris.org/). To access the source you'll need to install a Subversion client. Then, to get the source code, do:
svn checkout http://svn.perl.org/modules/dbd-oracle/trunk
If it prompts for a username and password use your perl.org account if you have one, else just 'guest' and 'guest'. The source code will be in a new subdirectory called trunk
.
To keep informed about changes to the source you can send an empty email to dbd-oracle-changes-subscribe@perl.org after which you'll get an email with the change log message and diff of each change checked-in to the source.
After making your changes you can generate a patch file, but before you do, make sure your source is still upto date using:
svn update
If you get any conflicts reported you'll need to fix them first. Then generate the patch file from within the trunk
directory using:
svn diff > foo.patch
Read the patch file, as a sanity check, and then email it to dbi-dev@perl.org.
How to create a patch without Subversion
Unpack a fresh copy of the distribution:
tar xfz DBD-Oracle-1.40.tar.gz
Rename the newly created top level directory:
mv DBD-Oracle-1.40 DBD-Oracle-1.40.your_foo
Edit the contents of DBD-Oracle-1.40.your_foo/* till it does what you want.
Test your changes and then remove all temporary files:
make test && make distclean
Go back to the directory you originally unpacked the distribution:
cd ..
Unpack another copy of the original distribution you started with:
tar xfz DBD-Oracle-1.40.tar.gz
Then create a patch file by performing a recursive diff
on the two top level directories:
diff -r -u DBD-Oracle-1.40 DBD-Oracle-1.40.your_foo > DBD-Oracle-1.40.your_foo.patch
Speak before you patch
For anything non-trivial or possibly controversial it's a good idea to discuss (on dbi-dev@perl.org) the changes you propose before actually spending time working on them. Otherwise you run the risk of them being rejected because they don't fit into some larger plans you may not be aware of.
2 POD Errors
The following errors were encountered while parsing the POD:
- Around line 1088:
'=item' outside of any '=over'
- Around line 1125:
You forgot a '=back' before '=head2'