ADOdb Library for PHP

V4.21 20 Mar 2004 (c) 2000-2004 John Lim (

This software is dual licensed using BSD-Style and LGPL. This means you can use it in compiled proprietary and commercial products.

Useful ADOdb links: Download   Other Docs

Unique Features
How People are using ADOdb
Feature Requests and Bug Reports
Minimum Install
Initializing Code and Connection Examples
ADONewConnection NewADOConnection
High Speed ADOdb - tuning tips

Hacking and Modifying ADOdb Safely
PHP5 Features

foreach iterators exceptions
Supported Databases
Example 1: Select
Example 2: Advanced Select
Example 3: Insert
Example 4: Debugging  rs2html example
Example 5: MySQL and Menus
Example 6: Connecting to Multiple Databases at once
Example 7: Generating Update and Insert SQL
Example 8: Implementing Scrolling with Next and Previous
Example 9: Exporting in CSV or Tab-Delimited Format
Example 10: Custom filters
Example 11: Smart Transactions

Using Custom Error Handlers and PEAR_Error
Data Source Names
Pivot Tables


Connections: Connect PConnect NConnect
Executing SQL: Execute CacheExecute SelectLimit CacheSelectLimit Param Prepare PrepareSP InParameter OutParameter
              GetOne CacheGetOne GetRow CacheGetRow GetAll CacheGetAll GetCol CacheGetCol GetAssoc CacheGetAssoc Replace
               ExecuteCursor (oci8 only)
Generates SQL strings: GetUpdateSQL GetInsertSQL Concat IfNull length random substr qstr Param OffsetDate SQLDate DBDate DBTimeStamp
Blobs: UpdateBlob UpdateClob UpdateBlobFile BlobEncode BlobDecode
Paging/Scrolling: PageExecute CachePageExecute
Cleanup: CacheFlush Close
Transactions: StartTrans CompleteTrans FailTrans HasFailedTrans BeginTrans CommitTrans RollbackTrans
Fetching Data:
Strings: concat length qstr quote substr
Dates: DBDate DBTimeStamp UnixDate UnixTimeStamp OffsetDate SQLDate
Row Management: Affected_Rows Insert_ID RowLock GenID CreateSequence DropSequence
Error Handling: ErrorMsg ErrorNo MetaError MetaErrorMsg
Data Dictionary (metadata): MetaDatabases MetaTables MetaColumns MetaColumnNames MetaPrimaryKeys MetaForeignKeys ServerInfo
Statistics and Query-Rewriting: LogSQL fnExecute and fnCacheExecute
Deprecated: Bind BlankRecordSet Parameter

Returns one field: Fields
Returns one row:FetchRow FetchInto FetchObject FetchNextObject FetchObj FetchNextObj GetRowAssoc
Returns all rows:GetArray GetRows GetAssoc
Scrolling:Move MoveNext MoveFirst MoveLast AbsolutePosition CurrentRow AtFirstPage AtLastPage AbsolutePage

Menu generation:GetMenu GetMenu2
Dates:UserDate UserTimeStamp UnixDate UnixTimeStamp
Recordset Info:RecordCount PO_RecordSet NextRecordSet
Field Info:FieldCount FetchField MetaType
Cleanup: Close

rs2html  example
Differences between ADOdb and ADO
Database Driver Guide
Change Log


PHP's database access functions are not standardised. This creates a need for a database class library to hide the differences between the different database API's (encapsulate the differences) so we can easily switch databases. PHP 4.0.5 or later is now required (because we use array-based str_replace).

We currently support MySQL, Oracle, Microsoft SQL Server, Sybase, Sybase SQL Anywhere, Informix, PostgreSQL, FrontBase, SQLite, Interbase (Firebird and Borland variants), Foxpro, Access, ADO, DB2, SAP DB and ODBC. We have had successful reports of connecting to Progress and CacheLite via ODBC. We hope more people will contribute drivers to support other databases.

PHP4 supports session variables. You can store your session information using ADOdb for true portability and scalability. See adodb-session.php for more information.

Also read (also available as tips_portable_sql.htm in the release) for tips on writing portable SQL.

Unique Features of ADOdb

How People are using ADOdb

Here are some examples of how people are using ADOdb (for a much longer list, visit

Feature Requests and Bug Reports

Feature requests and bug reports can be emailed to or posted to the ADOdb Help forums at

Installation Guide

Make sure you are running PHP 4.0.4 or later. Unpack all the files into a directory accessible by your webserver.

To test, try modifying some of the tutorial examples. Make sure you customize the connection settings correctly. You can debug using $db->debug = true as shown below:

	$db = ADONewConnection($dbdriver); # eg 'mysql' or 'postgres'
	$db->debug = true;
	$db->Connect($server, $user, $password, $database);
	$rs = $db->Execute('select * from some_small_table');
	print "<pre>";
	print "</pre>";

Minimum Install

For developers who want to release a minimal install of ADOdb, you will need:


Code Initialization Examples

When running ADOdb, at least two files are loaded. First is adodb/, which contains all functions used by all database classes. The code specific to a particular database is in the adodb/driver/adodb-????.inc.php file.

For example, to connect to a mysql database:

$conn = &ADONewConnection('mysql');

Whenever you need to connect to a database, you create a Connection object using the ADONewConnection($driver) function. NewADOConnection($driver) is an alternative name for the same function.

At this point, you are not connected to the database. You will first need to decide whether to use persistent or non-persistent connections. The advantage of persistent connections is that they are faster, as the database connection is never closed (even when you call Close()). Non-persistent connections take up much fewer resources though, reducing the risk of your database and your web-server becoming overloaded.

For persistent connections, use $conn->PConnect(), or $conn->Connect() for non-persistent connections. Some database drivers also support NConnect(), which forces the creation of a new connection.

Connection Gotcha: If you create two connections, but both use the same userid and password, PHP will share the same connection. This can cause problems if the connections are meant to different databases. The solution is to always use different userid's for different databases, or use NConnect().

Examples of Connecting to Databases

MySQL and Most Other Database Drivers

MySQL connections are very straightforward, and the parameters are identical to mysql_connect:

	$conn = &ADONewConnection('mysql'); 

Most other database drivers use a similar convention: Connect($server, $user, $password, $database). Exceptions are listed below.


PostgreSQL accepts connections using:

a. the standard connection string:

	$conn = &ADONewConnection('postgres7'); 
	$conn->PConnect('host=localhost port=5432 dbname=mary');

b. the classical 4 parameters:



You define the database in the $host parameter:
	$conn = &ADONewConnection('ibase'); 


Sqlite will create database if it does not exist.
	$conn = &ADONewConnection('sqlite'); 
	$conn->PConnect('c:\path\to\sqlite.db'); # sqlite will create if does not exist


With Oracle, you can connect in multiple ways.

a. PHP and Oracle reside on the same machine, use default SID.

	$conn->Connect(false, 'scott', 'tiger');

b. TNS Name defined, eg. 'myTNS'

	$conn->PConnect(false, 'scott', 'tiger', 'myTNS');


 	$conn->PConnect('myTNS', 'scott', 'tiger');

c. Host Address and SID

	$conn->Connect('', 'scott', 'tiger', 'SID');

d. Host Address and Service Name

	$conn->Connect('', 'scott', 'tiger', 'servicename');

DSN-less ODBC (access and mssql examples)

ODBC DSN's can be created in the ODBC control panel, or you can use a DSN-less connection.To use DSN-less connections with ODBC you need PHP 4.3 or later.

For Microsoft Access:

	$db =& ADONewConnection('access');
	$dsn = "Driver={Microsoft Access Driver (*.mdb)};Dbq=d:\\northwind.mdb;Uid=Admin;Pwd=;";
For Microsoft SQL Server:
	$db =& ADONewConnection('odbc_mssql');
	$dsn = "Driver={SQL Server};Server=localhost;Database=northwind;";
DSN-less Connections with ADO
If you are using versions of PHP earlier than PHP 4.3.0, DSN-less connections only work with Microsoft's ADO, which is Microsoft's COM based API. An example using the ADOdb library and Microsoft's ADO:
	$db = &ADONewConnection("ado_mssql");
	print "<h1>Connecting DSN-less $db->databaseType...</h1>";
		. "SERVER=flipper;DATABASE=ai;UID=sa;PWD=;"  ;
	$rs = $db->Execute("select * from table");
	$arr = $rs->GetArray();

High Speed ADOdb - tuning tips

ADOdb is a big class library, yet it consistently beats all other PHP class libraries in performance. This is because it is designed in a layered fashion, like an onion, with the fastest functions in the innermost layer. Stick to the following functions for best performance:

Innermost Layer

Connect, PConnect, NConnect
Execute, CacheExecute
SelectLimit, SelectLimit
MoveNext, Close
qstr, Affected_Rows, Insert_ID

The fastest way to access the fields is by accessing the array $recordset->fields directly. Also set the global variables $ADODB_FETCH_MODE = ADODB_FETCH_NUM, and (for oci8, ibase/firebird and odbc) $ADODB_COUNTRECS = false before you connect to your database. At the time of writing (Dec 2003).

Consider using bind parameters if your database supports it, as it improves query plan reuse. Use ADOdb's performance tuning system to identify bottlenecks quickly. At the time of writing (Dec 2003), this means oci8 and odbc drivers.

Lastly make sure you have a PHP accelerator cache installed such as APC, Turck MMCache, Zend Accelerator or ionCube.

Informix tips: Disable scrollable cursors with $db->cursorType = 0.

Hacking ADOdb Safely

You might want to modify ADOdb for your own purposes. Luckily you can still maintain backward compatibility by sub-classing ADOdb and using the $ADODB_NEWCONNECTION variable. $ADODB_NEWCONNECTION allows you to override the behaviour of ADONewConnection(). ADOConnection() checks for this variable and will call the function-name stored in this variable if it is defined.

In the following example, new functionality for the connection object is placed in the hack_mysql and hack_postgres7 classes. The recordset class naming convention can be controlled using $rsPrefix. Here we set it to 'hack_rs_', which will make ADOdb use hack_rs_mysql and hack_rs_postgres7 as the recordset classes. If you want to use the default ADOdb drivers return false.

class hack_mysql extends adodb_mysql {
var $rsPrefix = 'hack_rs_';
  /* Your mods here */

class hack_rs_mysql extends ADORecordSet_mysql {
 /* Your mods here */

class hack_postgres7 extends adodb_postgres7 {
var $rsPrefix = 'hack_rs_';
  /* Your mods here */

class hack_rs_postgres7 extends ADORecordSet_postgres7 {
 /* Your mods here */

$ADODB_NEWCONNECTION = 'hack_factory';

function& hack_factory($driver)
	if ($driver !== 'mysql' && $driver !== 'postgres7') return false;
	$driver = 'hack_'.$driver;
	$obj = new $driver();
	return $obj;


Don't forget to call the constructor of the parent class.

PHP5 Features

ADOdb 4.02 or later will transparently determine which version of PHP you are using. If PHP5 is detected, the following features become available:

Databases Supported

Name Tested Database RecordCount() usable Prerequisites Operating Systems
access B Microsoft Access/Jet. You need to create an ODBC DSN. Y/N ODBC Windows only
ado B

Generic ADO, not tuned for specific databases. Allows DSN-less connections. For best performance, use an OLEDB provider. This is the base class for all ado drivers.

You can set $db->codePage before connecting.

? depends on database ADO or OLEDB provider Windows only
ado_access B Microsoft Access/Jet using ADO. Allows DSN-less connections. For best performance, use an OLEDB provider. Y/N ADO or OLEDB provider Windows only
ado_mssql B Microsoft SQL Server using ADO. Allows DSN-less connections. For best performance, use an OLEDB provider. Y/N ADO or OLEDB provider Windows only
db2 A DB2. Should work reliably as based on ODBC driver. Y/N DB2 CLI/ODBC interface

Unix and Windows. Unix install hints. I have had reports that the $host and $database params have to be reversed in Connect() when using the CLI interface.

vfp A Microsoft Visual FoxPro. You need to create an ODBC DSN. Y/N ODBC Windows only
fbsql C FrontBase. Y ?

Unix and Windows

ibase B Interbase 6 or earlier. Some users report you might need to use this
$db->PConnect('localhost:c:/ibase/employee.gdb', "sysdba", "masterkey") to connect. Lacks Affected_Rows currently.

You can set $db->dialect, $db->buffers and $db->charSet before connecting.
Y/N Interbase client Unix and Windows
firebird C Firebird version of interbase. Y/N Interbase client Unix and Windows
borland_ibase C Borland version of Interbase 6.5 or later. Very sad that the forks differ. Y/N Interbase client Unix and Windows
informix72 C Informix databases before Informix 7.3 that do no support SELECT FIRST. Y/N Informix client Unix and Windows
informix C Generic informix driver. Y/N Informix client Unix and Windows
mssql A

Microsoft SQL Server 7 and later. Works with Microsoft SQL Server 2000 also. Note that date formating is problematic with this driver. For example, the PHP mssql extension does not return the seconds for datetime!

Y/N Mssql client

Unix and Windows.
Unix install howto and another one.

mssqlpo A

Portable mssql driver. Identical to above mssql driver, except that '||', the concatenation operator, is converted to '+'. Useful for porting scripts from most other sql variants that use ||.

Y/N Mssql client

Unix and Windows.
Unix install howto

mysql A MySQL without transaction support. You can also set $db->clientFlags before connecting. Y/N MySQL client Unix and Windows
mysqlt or maxsql A

MySQL with transaction support. We recommend using || as the concat operator for best portability. This can be done by running MySQL using:
mysqld --ansi or mysqld --sql-mode=PIPES_AS_CONCAT

Y/N MySQL client Unix and Windows
oci8 A Oracle 8/9. Has more functionality than oracle driver (eg. Affected_Rows). You might have to putenv('ORACLE_HOME=...') before Connect/PConnect.

There are 2 ways of connecting - with server IP and service name:
or using an entry in TNSNAMES.ORA or ONAMES or HOSTNAMES:
PConnect(false, 'scott', 'tiger', $oraname).

Since 2.31, we support Oracle REF cursor variables directly (see ExecuteCursor).

Y/N Oracle client Unix and Windows
oci805 C Supports reduced Oracle functionality for Oracle 8.0.5. SelectLimit is not as efficient as in the oci8 or oci8po drivers. Y/N Oracle client Unix and Windows
oci8po A Oracle 8/9 portable driver. This is nearly identical with the oci8 driver except (a) bind variables in Prepare() use the ? convention, instead of :bindvar, (b) field names use the more common PHP convention of lowercase names.

Use this driver if porting from other databases is important. Otherwise the oci8 driver offers better performance.

Y/N Oracle client Unix and Windows
odbc A Generic ODBC, not tuned for specific databases. To connect, use
PConnect('DSN','user','pwd'). This is the base class for all odbc derived drivers.
? depends on database ODBC Unix and Windows. Unix hints.
odbc_mssql C Uses ODBC to connect to MSSQL Y/N ODBC Unix and Windows.
odbc_oracle C Uses ODBC to connect to Oracle Y/N ODBC Unix and Windows.
odbtp C Generic odbtp driver. Odbtp is a software for accessing Windows ODBC data sources from other operating systems. Y/N odbtp Unix and Windows
oracle C Implements old Oracle 7 client API. Use oci8 driver if possible for better performance. Y/N Oracle client Unix and Windows
postgres A Generic PostgreSQL driver. Currently identical to postgres7 driver. Y PostgreSQL client Unix and Windows.
postgres64 A For PostgreSQL 6.4 and earlier which does not support LIMIT internally. Y PostgreSQL client Unix and Windows.
postgres7 A PostgreSQL which supports LIMIT and other version 7 functionality. Y PostgreSQL client Unix and Windows.
sapdb C SAP DB. Should work reliably as based on ODBC driver. Y/N SAP ODBC client


sqlanywhere C Sybase SQL Anywhere. Should work reliably as based on ODBC driver. Y/N SQL Anywhere ODBC client


sqlite B SQLite. Only tested on PHP5. Y -

Unix and Windows.

sybase C Sybase. Y/N Sybase client

Unix and Windows.

The "Tested" column indicates how extensively the code has been tested and used.
A = well tested and used by many people
B = tested and usable, but some features might not be implemented
C = user contributed or experimental driver. Might not fully support all of the latest features of ADOdb.

The column "RecordCount() usable" indicates whether RecordCount() return the number of rows, or returns -1 when a SELECT statement is executed. If this column displays Y/N then the RecordCount() is emulated when the global variable $ADODB_COUNTRECS=true (this is the default). Note that for large recordsets, it might be better to disable RecordCount() emulation because substantial amounts of memory are required to cache the recordset for counting. Also there is a speed penalty of 40-50% if emulation is required. This is emulated in most databases except for PostgreSQL and MySQL. This variable is checked every time a query is executed, so you can selectively choose which recordsets to count.


Example 1: Select Statement

Task: Connect to the Access Northwind DSN, display the first 2 columns of each row.

In this example, we create a ADOConnection object, which represents the connection to the database. The connection is initiated with PConnect, which is a persistent connection. Whenever we want to query the database, we call the ADOConnection.Execute() function. This returns an ADORecordSet object which is actually a cursor that holds the current row in the array fields[]. We use MoveNext() to move from row to row.

NB: A useful function that is not used in this example is SelectLimit, which allows us to limit the number of rows shown.

include('');	   # load code common to ADOdb
$conn = &ADONewConnection('access');	# create a connection
$conn->PConnect('northwind');   # connect to MS-Access, northwind DSN
$recordSet = &$conn->Execute('select * from products');
if (!$recordSet) 
	print $conn->ErrorMsg();
while (!$recordSet->EOF) {
	print $recordSet->fields[0].' '.$recordSet->fields[1].'<BR>';

$recordSet->Close(); # optional
$conn->Close(); # optional


The $recordSet returned stores the current row in the $recordSet->fields array, indexed by column number (starting from zero). We use the MoveNext() function to move to the next row. The EOF property is set to true when end-of-file is reached. If an error occurs in Execute(), we return false instead of a recordset.

The $recordSet->fields[] array is generated by the PHP database extension. Some database extensions only index by number and do not index the array by field name. To force indexing by name - that is associative arrays - use the SetFetchMode function. Each recordset saves and uses whatever fetch mode was set when the recordset was created in Execute() or SelectLimit().

	$rs1 = $db->Execute('select * from table');
	$rs2 = $db->Execute('select * from table');
	print_r($rs1->fields); # shows array([0]=>'v0',[1] =>'v1')
	print_r($rs2->fields); # shows array(['col1']=>'v0',['col2'] =>'v1')

To get the number of rows in the select statement, you can use $recordSet->RecordCount(). Note that it can return -1 if the number of rows returned cannot be determined.

Example 2: Advanced Select with Field Objects

Select a table, display the first two columns. If the second column is a date or timestamp, reformat the date to US format.

include('');	   # load code common to ADOdb
$conn = &ADONewConnection('access');	# create a connection
$conn->PConnect('northwind');   # connect to MS-Access, northwind dsn
$recordSet = &$conn->Execute('select CustomerID,OrderDate from Orders');
if (!$recordSet) 
	print $conn->ErrorMsg();
while (!$recordSet->EOF) {
	$fld = $recordSet->FetchField(1);
	$type = $recordSet->MetaType($fld->type);

	if ( $type == 'D' || $type == 'T') 
		print $recordSet->fields[0].' '.
		print $recordSet->fields[0].' '.$recordSet->fields[1].'<BR>';

$recordSet->Close(); # optional
$conn->Close(); # optional


In this example, we check the field type of the second column using FetchField(). This returns an object with at least 3 fields.

We then use MetaType() to translate the native type to a generic type. Currently the following generic types are defined:

If the metatype is of type date or timestamp, then we print it using the user defined date format with UserDate(), which converts the PHP SQL date string format to a user defined one. Another use for MetaType() is data validation before doing an SQL insert or update.

Example 3: Inserting

Insert a row to the Orders table containing dates and strings that need to be quoted before they can be accepted by the database, eg: the single-quote in the word John's.

include('');	   # load code common to ADOdb
$conn = &ADONewConnection('access');	# create a connection

$conn->PConnect('northwind');   # connect to MS-Access, northwind dsn
$shipto = $conn->qstr("John's Old Shoppe");

$sql = "insert into orders (customerID,EmployeeID,OrderDate,ShipName) ";
$sql .= "values ('ANATR',2,".$conn->DBDate(time()).",$shipto)";

if ($conn->Execute($sql) === false) {
	print 'error inserting: '.$conn->ErrorMsg().'<BR>';

In this example, we see the advanced date and quote handling facilities of ADOdb. The unix timestamp (which is a long integer) is appropriately formated for Access with DBDate(), and the right escape character is used for quoting the John's Old Shoppe, which is John''s Old Shoppe and not PHP's default John's Old Shoppe with qstr().

Observe the error-handling of the Execute statement. False is returned by Execute() if an error occured. The error message for the last error that occurred is displayed in ErrorMsg(). Note: php_track_errors might have to be enabled for error messages to be saved.

Example 4: Debugging

include('');	   # load code common to ADOdb
$conn = &ADONewConnection('access');	# create a connection
$conn->PConnect('northwind');   # connect to MS-Access, northwind dsn
$shipto = $conn->qstr("John's Old Shoppe");
$sql = "insert into orders (customerID,EmployeeID,OrderDate,ShipName) ";
$sql .= "values ('ANATR',2,".$conn->FormatDate(time()).",$shipto)";
$conn->debug = true;
if ($conn->Execute($sql) === false) print 'error inserting';

In the above example, we have turned on debugging by setting debug = true. This will display the SQL statement before execution, and also show any error messages. There is no need to call ErrorMsg() in this case. For displaying the recordset, see the rs2html() example.

Also see the section on Custom Error Handlers.

Example 5: MySQL and Menus

Connect to MySQL database agora, and generate a <select> menu from an SQL statement where the <option> captions are in the 1st column, and the value to send back to the server is in the 2nd column.

include(''); # load code common to ADOdb
$conn = &ADONewConnection('mysql');  # create a connection
$conn->PConnect('localhost','userid','','agora');# connect to MySQL, agora db
$sql = 'select CustomerName, CustomerID from customers';
$rs = $conn->Execute($sql);
print $rs->GetMenu('GetCust','Mary Rosli');

Here we define a menu named GetCust, with the menu option 'Mary Rosli' selected. See GetMenu(). We also have functions that return the recordset as an array: GetArray(), and as an associative array with the key being the first column: GetAssoc().

Example 6: Connecting to 2 Databases At Once

include('');	 # load code common to ADOdb
$conn1 = &ADONewConnection('mysql');  # create a mysql connection
$conn2 = &ADONewConnection('oracle');  # create a oracle connection

$conn1->PConnect($server, $userid, $password, $database);
$conn2->PConnect(false, $ora_userid, $ora_pwd, $oraname);

$conn1->Execute('insert ...');
$conn2->Execute('update ...');

Example 7: Generating Update and Insert SQL

ADOdb 1.31 and later supports two new recordset functions: GetUpdateSQL( ) and GetInsertSQL( ). This allow you to perform a "SELECT * FROM table query WHERE...", make a copy of the $rs->fields, modify the fields, and then generate the SQL to update or insert into the table automatically.

We show how the functions can be used when accessing a table with the following fields: (ID, FirstName, LastName, Created).

Before these functions can be called, you need to initialize the recordset by performing a select on the table. Idea and code by Jonathan Younger

# SAMPLE GetUpdateSQL() and GetInsertSQL() code

# This code tests an insert

$sql = "SELECT * FROM ADOXYZ WHERE id = -1"; 
# Select an empty record from the database

$conn = &ADONewConnection("mysql");  # create a connection
$conn->PConnect("localhost", "admin", "", "test"); # connect to MySQL, testdb
$rs = $conn->Execute($sql); # Execute the query and get the empty recordset

$record = array(); # Initialize an array to hold the record data to insert

# Set the values for the fields in the record
# Note that field names are case-insensitive
$record["firstname"] = "Bob";
$record["lastNamE"] = "Smith";
$record["creaTed"] = time();

# Pass the empty recordset and the array containing the data to insert
# into the GetInsertSQL function. The function will process the data and return
# a fully formatted insert sql statement.
$insertSQL = $conn->GetInsertSQL($rs, $record);

$conn->Execute($insertSQL); # Insert the record into the database

# This code tests an update

$sql = "SELECT * FROM ADOXYZ WHERE id = 1"; 
# Select a record to update

$rs = $conn->Execute($sql); # Execute the query and get the existing record to update

$record = array(); # Initialize an array to hold the record data to update

# Set the values for the fields in the record
# Note that field names are case-insensitive
$record["firstname"] = "Caroline";
$record["LasTnAme"] = "Smith"; # Update Caroline's lastname from Miranda to Smith

# Pass the single record recordset and the array containing the data to update
# into the GetUpdateSQL function. The function will process the data and return
# a fully formatted update sql statement with the correct WHERE clause.
# If the data has not changed, no recordset is returned
$updateSQL = $conn->GetUpdateSQL($rs, $record);

$conn->Execute($updateSQL); # Update the record in the database

Example 8: Implementing Scrolling with Next and Previous

The following code creates a very simple recordset pager, where you can scroll from page to page of a recordset.


$db = NewADOConnection('mysql');


$sql = "select * from adoxyz ";

$pager = new ADODB_Pager($db,$sql);

This will create a basic record pager that looks like this:

|<   <<   >>   >|  
ID First Name Last Name Date Created
36  Alan  Turing  Sat 06, Oct 2001 
37  Serena  Williams  Sat 06, Oct 2001 
38  Yat Sun  Sun  Sat 06, Oct 2001 
39  Wai Hun  See  Sat 06, Oct 2001 
40  Steven  Oey  Sat 06, Oct 2001 
Page 8/10

The number of rows to display at one time is controled by the Render($rows) method. If you do not pass any value to Render(), ADODB_Pager will default to 10 records per page.

You can control the column titles by modifying your SQL (supported by most databases):

$sql = 'select id as "ID", firstname as "First Name", 
		  lastname as "Last Name", created as "Date Created" 
from adoxyz';

The above code can be found in the adodb/tests/testpaging.php example included with this release, and the class ADODB_Pager in adodb/ The ADODB_Pager code can be adapted by a programmer so that the text links can be replaced by images, and the dull white background be replaced with more interesting colors.

You can also allow display of html by setting $pager->htmlSpecialChars = false.

Some of the code used here was contributed by Iván Oliva and Cornel G.

Example 9: Exporting in CSV or Tab-Delimited Format

We provide some helper functions to export in comma-separated-value (CSV) and tab-delimited formats:

$db = &NewADOConnection('mysql');
$db->Connect($server, $userid, $password, $database);

$rs = $db->Execute('select fname as "First Name", surname as "Surname" from table'); print "<pre>";
print rs2csv($rs); # return a string, CSV format

print '<hr>';
$rs->MoveFirst(); # note, some databases do not support MoveFirst
print rs2tab($rs,false); # return a string, tab-delimited # false == suppress field names in first line

print '<hr>';
rs2tabout($rs); # send to stdout directly (there is also an rs2csvout function) print "</pre>"; $rs->MoveFirst();
$fp = fopen($path, "w"); if ($fp) {
rs2csvfile($rs, $fp); # write to file (there is also an rs2tabfile function) fclose($fp);

Carriage-returns or newlines are converted to spaces. Field names are returned in the first line of text. Strings containing the delimiter character are quoted with double-quotes. Double-quotes are double-quoted again. This conforms to Excel import and export guide-lines.

All the above functions take as an optional last parameter, $addtitles which defaults to true. When set to false field names in the first line are suppressed.

Example 10: Recordset Filters

Sometimes we want to pre-process all rows in a recordset before we use it. For example, we want to ucwords all text in recordset.


// ucwords() every element in the recordset
function do_ucwords(&$arr,$rs)
	foreach($arr as $k => $v) {
		$arr[$k] = ucwords($v);

$db = NewADOConnection('mysql');

$rs = $db->Execute('select ... from table');
$rs = RSFilter($rs,'do_ucwords');

The RSFilter function takes 2 parameters, the recordset, and the name of the filter function. It returns the processed recordset scrolled to the first record. The filter function takes two parameters, the current row as an array, and the recordset object. For future compatibility, you should not use the original recordset object.

Example 11: Smart Transactions

The old way of doing transactions required you to use
$ok = $conn->Execute($sql);
if ($ok) $ok = $conn->Execute($sql2);
if (!$ok) $conn->RollbackTrans();
else $conn->CommitTrans();
This is very complicated for large projects because you have to track the error status. Smart Transactions is much simpler. You start a smart transaction by calling StartTrans():
CompleteTrans() detects when an SQL error occurs, and will Rollback/Commit as appropriate. To specificly force a rollback even if no error occured, use FailTrans(). Note that the rollback is done in CompleteTrans(), and not in FailTrans().
if (!CheckRecords()) $conn->FailTrans();

You can also check if a transaction has failed, using HasFailedTrans(), which returns true if FailTrans() was called, or there was an error in the SQL execution. Make sure you call HasFailedTrans() before you call CompleteTrans(), as it is only works between StartTrans/CompleteTrans.

Lastly, StartTrans/CompleteTrans is nestable, and only the outermost block is executed. In contrast, BeginTrans/CommitTrans/RollbackTrans is NOT nestable.

  $conn->StartTrans();    # ignored
  if (!CheckRecords()) $conn->FailTrans();
  $conn->CompleteTrans(); # ignored

Note: Savepoints are currently not supported.

Using Custom Error Handlers and PEAR_Error

Apart from the old $con->debug = true; way of debugging, ADOdb 1.50 onwards provides another way of handling errors using ADOdb's custom error handlers.

ADOdb provides two custom handlers which you can modify for your needs. The first one is in the file. This makes use of the standard PHP functions error_reporting to control what error messages types to display, and trigger_error which invokes the default PHP error handler.

Including the above file will cause trigger_error($errorstring,E_USER_ERROR) to be called when
(a) Connect() or PConnect() fails, or
(b) a function that executes SQL statements such as Execute() or SelectLimit() has an error.
(c) GenID() appears to go into an infinite loop.

The $errorstring is generated by ADOdb and will contain useful debugging information similar to the error.log data generated below. This file should be included before you create any ADOConnection objects.

If you define error_reporting(0), no errors will be passed to the error handler. If you set error_reporting(E_ALL), all errors will be passed to the error handler. You still need to use ini_set("display_errors", "0" or "1") to control the display of errors.

error_reporting(E_ALL); # pass any error messages triggered to error handler
$c = NewADOConnection('mysql');
$rs=$c->Execute('select * from productsz'); #invalid table productsz');
if ($rs) rs2html($rs);

If you want to log the error message, you can do so by defining the following optional constants ADODB_ERROR_LOG_TYPE and ADODB_ERROR_LOG_DEST. ADODB_ERROR_LOG_TYPE is the error log message type (see error_log in the PHP manual). In this case we set it to 3, which means log to the file defined by the constant ADODB_ERROR_LOG_DEST.

error_reporting(E_ALL); # report all errors
ini_set("display_errors", "0"); # but do not echo the errors

$c = NewADOConnection('mysql');
$rs=$c->Execute('select * from productsz'); ## invalid table productsz
if ($rs) rs2html($rs);
The following message will be logged in the error.log file:
(2001-10-28 14:20:38) mysql error: [1146: Table 'northwind.productsz' doesn't exist] in
 EXECUTE("select * from productsz")
The second error handler is This will create a PEAR_Error derived object whenever an error occurs. The last PEAR_Error object created can be retrieved using ADODB_Pear_Error().
$c = NewADOConnection('mysql');
$rs=$c->Execute('select * from productsz'); #invalid table productsz');
if ($rs) rs2html($rs);
else {
	$e = ADODB_Pear_Error();
	echo '<p>',$e->message,'</p>';

You can use a PEAR_Error derived class by defining the constant ADODB_PEAR_ERROR_CLASS before the file is included. For easy debugging, you can set the default error handler in the beginning of the PHP script to PEAR_ERROR_DIE, which will cause an error message to be printed, then halt script execution:


Note that we do not explicitly return a PEAR_Error object to you when an error occurs. We return false instead. You have to call ADODB_Pear_Error() to get the last error or use the PEAR_ERROR_DIE technique.

Error Messages

Error messages are outputted using the static method ADOConnnection::outp($msg,$newline=true). By default, it sends the messages to the client. You can override this to perform error-logging.

Data Source Names

We now support connecting using PEAR style DSN's. A DSN is a connection string of the form:

$dsn = "$driver://$username:$password@$hostname/$databasename";

You pass the DSN to the static class function DB::Connect. An example:

   $username = 'root';
   $password = '';
   $hostname = 'localhost';
   $databasename = 'xphplens';
   $driver = 'mysql';
   $dsn = "$driver://$username:$password@$hostname/$databasename";
   $db = DB::Connect($dsn);
$rs = $db->Execute('select firstname,lastname from adoxyz'); $cnt = 0; while ($arr = $rs->FetchRow()) { print_r($arr); print "<br>"; }

This requires PEAR to be installed and in the default include path in php.ini.

Caching of Recordsets

ADOdb now supports caching of recordsets using the CacheExecute( ), CachePageExecute( ) and CacheSelectLimit( ) functions. There are similar to the non-cache functions, except that they take a new first parameter, $secs2cache.

An example:

include(''); # load code common to ADOdb
$ADODB_CACHE_DIR = '/usr/ADODB_cache';
$conn = &ADONewConnection('mysql');  # create a connection
$conn->PConnect('localhost','userid','','agora');# connect to MySQL, agora db
$sql = 'select CustomerName, CustomerID from customers';
$rs = $conn->CacheExecute(15,$sql);

The first parameter is the number of seconds to cache the query. Subsequent calls to that query will used the cached version stored in $ADODB_CACHE_DIR. To force a query to execute and flush the cache, call CacheExecute() with the first parameter set to zero. Alternatively, use the CacheFlush($sql) call.

For the sake of security, we recommend you set register_globals=off in php.ini if you are using $ADODB_CACHE_DIR.

In ADOdb 1.80 onwards, the secs2cache parameter is optional in CacheSelectLimit() and CacheExecute(). If you leave it out, it will use the $connection->cacheSecs parameter, which defaults to 60 minutes.

	$conn->cacheSecs = 3600*24; # cache 24 hours
	$rs = $conn->CacheExecute('select * from table');

Please note that magic_quotes_runtime should be turned off. More info.

Pivot Tables

Since ADOdb 2.30, we support the generation of SQL to create pivot tables, also known as cross-tabulations. For further explanation read this DevShed Cross-Tabulation tutorial. We assume that your database supports the SQL case-when expression.

In this example, we will use the Northwind database from Microsoft. In the database, we have a products table, and we want to analyze this table by suppliers versus product categories. We will place the suppliers on each row, and pivot on categories. So from the table on the left, we generate the pivot-table on the right:

Supplier Category
supplier1 category1
supplier2 category1
supplier2 category2
  category1 category2 total
supplier1 1 0 1
supplier2 1 1 2

The following code will generate the SQL for a cross-tabulation:

# Query the main "product" table
# Set the rows to CompanyName
# and the columns to the values of Categories
# and define the joins to link to lookup tables 
# "categories" and "suppliers"
 include "adodb/pivottable.php";
 $sql = PivotTableSQL(
 	$gDB,                                      # adodb connection
 	'products p ,categories c ,suppliers s',   # tables
	'CompanyName',                             # rows (multiple fields allowed)
	'CategoryName',                            # column to pivot on 
	'p.CategoryID = c.CategoryID and s.SupplierID= p.SupplierID' # joins/where

This will generate the following SQL:

SELECT CompanyName,
SUM(CASE WHEN CategoryName='Beverages' THEN 1 ELSE 0 END) AS "Beverages",
SUM(CASE WHEN CategoryName='Condiments' THEN 1 ELSE 0 END) AS "Condiments",
SUM(CASE WHEN CategoryName='Confections' THEN 1 ELSE 0 END) AS "Confections",
SUM(CASE WHEN CategoryName='Dairy Products' THEN 1 ELSE 0 END) AS "Dairy Products",
SUM(CASE WHEN CategoryName='Grains/Cereals' THEN 1 ELSE 0 END) AS "Grains/Cereals",
SUM(CASE WHEN CategoryName='Meat/Poultry' THEN 1 ELSE 0 END) AS "Meat/Poultry",
SUM(CASE WHEN CategoryName='Produce' THEN 1 ELSE 0 END) AS "Produce",
SUM(CASE WHEN CategoryName='Seafood' THEN 1 ELSE 0 END) AS "Seafood",
SUM(1) as Total
FROM products p ,categories c ,suppliers s WHERE p.CategoryID = c.CategoryID and s.SupplierID= p.SupplierID
GROUP BY CompanyName

You can also pivot on numerical columns and generate totals by using ranges. This code was revised in ADODB 2.41 and is not backward compatible. The second example shows this:

 $sql = PivotTableSQL(
 	$gDB,                                       # adodb connection
 	'products p ,categories c ,suppliers s',    # tables
	'CompanyName',                              # rows (multiple fields allowed)
   array(                                       # column ranges
	' 0 '      => 'UnitsInStock <= 0',
	"1 to 5"   => '0 < UnitsInStock and UnitsInStock <= 5',
	"6 to 10"  => '5 < UnitsInStock and UnitsInStock <= 10',
	"11 to 15" => '10 < UnitsInStock and UnitsInStock <= 15',
	"16+"      => '15 < UnitsInStock'
	' p.CategoryID = c.CategoryID and s.SupplierID= p.SupplierID', # joins/where
	'UnitsInStock',                             # sum this field
	'Sum '                                      # sum label prefix

Which generates:

SELECT CompanyName,
SUM(CASE WHEN UnitsInStock <= 0 THEN UnitsInStock ELSE 0 END) AS "Sum 0 ",
SUM(CASE WHEN 0 < UnitsInStock and UnitsInStock <= 5 THEN UnitsInStock ELSE 0 END) AS "Sum 1 to 5",
SUM(CASE WHEN 5 < UnitsInStock and UnitsInStock <= 10 THEN UnitsInStock ELSE 0 END) AS "Sum 6 to 10",
SUM(CASE WHEN 10 < UnitsInStock and UnitsInStock <= 15 THEN UnitsInStock ELSE 0 END) AS "Sum 11 to 15",
SUM(CASE WHEN 15 < UnitsInStock THEN UnitsInStock ELSE 0 END) AS "Sum 16+",
SUM(UnitsInStock) AS "Sum UnitsInStock",
SUM(1) as Total,
FROM products p ,categories c ,suppliers s WHERE p.CategoryID = c.CategoryID and s.SupplierID= p.SupplierID
GROUP BY CompanyName

Class Reference

Function parameters with [ ] around them are optional.

Global Variables


If the database driver API does not support counting the number of records returned in a SELECT statement, the function RecordCount() is emulated when the global variable $ADODB_COUNTRECS is set to true, which is the default. We emulate this by buffering the records, which can take up large amounts of memory for big recordsets. Set this variable to false for the best performance. This variable is checked every time a query is executed, so you can selectively choose which recordsets to count.


If you are using recordset caching, this is the directory to save your recordsets in. Define this before you call any caching functions such as CacheExecute( ). We recommend setting register_globals=off in php.ini if you use this feature for security reasons.

If you are using Unix and apache, you might need to set your cache directory permissions to something similar to the following:

chown -R apache /path/to/adodb/cache
chgrp -R apache /path/to/adodb/cache


Determines whether to right trim CHAR fields (and also VARCHAR for ibase/firebird). Set to true to trim. Default is false. Currently works for oci8po, ibase and firebird drivers. Added in ADOdb 4.01.


Determines the language used in MetaErrorMsg(). The default is 'en', for English. To find out what languages are supported, see the files in adodb/lang/adodb-$, where $lang is the supported langauge.


This is a global variable that determines how arrays are retrieved by recordsets. The recordset saves this value on creation (eg. in Execute( ) or SelectLimit( )), and any subsequent changes to $ADODB_FETCH_MODE have no affect on existing recordsets, only on recordsets created in the future.

The following constants are defined:


An example:

	$rs1 = $db->Execute('select * from table');
	$rs2 = $db->Execute('select * from table');
	print_r($rs1->fields); # shows array([0]=>'v0',[1] =>'v1')
	print_r($rs2->fields); # shows array(['col1']=>'v0',['col2'] =>'v1')

As you can see in the above example, both recordsets store and use different fetch modes based on the $ADODB_FETCH_MODE setting when the recordset was created by Execute().

If no fetch mode is predefined, the fetch mode defaults to ADODB_FETCH_DEFAULT. The behaviour of this default mode varies from driver to driver, so do not rely on ADODB_FETCH_DEFAULT. For portability, we recommend sticking to ADODB_FETCH_NUM or ADODB_FETCH_ASSOC. Many drivers do not support ADODB_FETCH_BOTH.

SetFetchMode Function

Some programmers prefer to use a more object-oriented solution, where the fetch mode is set by a object function, SetFetchMode. Once this function is called for a connection object, that connection object will ignore the global variable $ADODB_FETCH_MODE and will use the internal fetchMode property exclusively.

	$rs1 = $db->Execute('select * from table');
	$rs2 = $db->Execute('select * from table');
	print_r($rs1->fields); # shows array([0]=>'v0',[1] =>'v1')
	print_r($rs2->fields); # shows array(['col1']=>'v0',['col2'] =>'v1')

To retrieve the previous fetch mode, you can use check the $db->fetchMode property, or use the return value of SetFetchMode( ).


You can control the associative fetch case for certain drivers which behave differently. For the sybase, oci8po, mssql, odbc and ibase drivers and all drivers derived from them, ADODB_ASSOC_CASE will by default generate recordsets where the field name keys are lower-cased. Use the constant ADODB_ASSOC_CASE to change the case of the keys. There are 3 possible values:

0 = assoc lowercase field names. $rs->fields['orderid']
1 = assoc uppercase field names. $rs->fields['ORDERID']
2 = use native-case field names. $rs->fields['OrderID'] -- this is the default since ADOdb 2.90

To use it, declare it before you incldue

define('ADODB_ASSOC_CASE', 2); # use native-case for ADODB_FETCH_ASSOC


Object that performs the connection to the database, executes SQL statements and has a set of utility functions for standardising the format of SQL statements for issues such as concatenation and date formats.

ADOConnection Fields

databaseType: Name of the database system we are connecting to. Eg. odbc or mssql or mysql.

dataProvider: The underlying mechanism used to connect to the database. Normally set to native, unless using odbc or ado.

host: Name of server or data source name (DSN) to connect to.

database: Name of the database or to connect to. If ado is used, it will hold the ado data provider.

user: Login id to connect to database. Password is not saved for security reasons.

raiseErrorFn: Allows you to define an error handling function. See for an example.

debug: Set to true to make debug statements to appear.

concat_operator: Set to '+' or '||' normally. The operator used to concatenate strings in SQL. Used by the Concat function.

fmtDate: The format used by the DBDate function to send dates to the database. is '#Y-m-d#' for Microsoft Access, and ''Y-m-d'' for MySQL.

fmtTimeStamp: The format used by the DBTimeStamp function to send timestamps to the database.

true: The value used to represent true.Eg. '.T.'. for Foxpro, '1' for Microsoft SQL.

false: The value used to represent false. Eg. '.F.'. for Foxpro, '0' for Microsoft SQL.

replaceQuote: The string used to escape quotes. Eg. double single-quotes for Microsoft SQL, and backslash-quote for MySQL. Used by qstr.

autoCommit: indicates whether automatic commit is enabled. Default is true.

charSet: set the default charset to use. Currently only interbase supports this.

dialect: set the default sql dialect to use. Currently only interbase supports this.

metaTablesSQL: SQL statement to return a list of available tables. Eg. SHOW TABLES in MySQL.

genID: The latest id generated by GenID() if supported by the database.

cacheSecs: The number of seconds to cache recordsets if CacheExecute() or CacheSelectLimit() omit the $secs2cache parameter. Defaults to 60 minutes.

sysDate: String that holds the name of the database function to call to get the current date. Useful for inserts and updates.

sysTimeStamp: String that holds the name of the database function to call to get the current timestamp/datetime value.

leftOuter: String that holds operator for left outer join, if known. Otherwise set to false.

rightOuter: String that holds operator for left outer join, if known. Otherwise set to false.

ansiOuter: Boolean that if true indicates that ANSI style outer joins are permitted. Eg. select * from table1 left join table2 on p1=p2.

connectSID: Boolean that indicates whether to treat the $database parameter in connects as the SID for the oci8 driver. Defaults to false. Useful for Oracle 8.0.5 and earlier.

autoRollback: Persistent connections are auto-rollbacked in PConnect( ) if this is set to true. Default is false.

ADOConnection Main Functions

ADOConnection( )

Constructor function. Do not call this directly. Use ADONewConnection( ) instead.


Non-persistent connect to data source or server $host, using userid $user and password $password. If the server supports multiple databases, connect to database $database.

Returns true/false depending on connection.

ADO Note: If you are using a Microsoft ADO and not OLEDB, you can set the $database parameter to the OLEDB data provider you are using.

PostgreSQL: An alternative way of connecting to the database is to pass the standard PostgreSQL connection string in the first parameter $host, and the other parameters will be ignored.

For Oracle and Oci8, there are two ways to connect. First is to use the TNS name defined in your local tnsnames.ora (or ONAMES or HOSTNAMES). Place the name in the $database field, and set the $host field to false. Alternatively, set $host to the server, and $database to the database SID, this bypassed tnsnames.ora.


 # $oraname in tnsnames.ora/ONAMES/HOSTNAMES
 $conn->Connect(false, 'scott', 'tiger', $oraname); 
 $conn->Connect('server:1521', 'scott', 'tiger', 'ServiceName'); # bypass tnsnames.ora

There are many examples of connecting to a database at, and in the file included in the release.


Persistent connect to data source or server $host, using userid $user and password $password. If the server supports multiple databases, connect to database $database.

We now perform a rollback on persistent connection for selected databases since 2.21, as advised in the PHP manual. See change log or source code for which databases are affected.

Returns true/false depending on connection. See Connect( ) above for more info.

Since ADOdb 2.21, we also support autoRollback. If you set:

 $conn = &NewADOConnection('mysql');
 $conn->autoRollback = true; # default is false
 $conn->PConnect(...); # rollback here

Then when doing a persistent connection with PConnect( ), ADOdb will perform a rollback first. This is because it is documented that PHP is not guaranteed to rollback existing failed transactions when persistent connections are used. This is implemented in Oracle, MySQL, PgSQL, MSSQL, ODBC currently.

Since ADOdb 3.11, you can force non-persistent connections even if PConnect is called by defining the constant ADODB_NEVER_PERSIST before you call PConnect.


Always force a new connection. In contrast, PHP sometimes reuses connections when you use Connect() or PConnect(). Currently works only on mysql (PHP 4.3.0 or later), postgresql and oci8-derived drivers. For other drivers, NConnect() works like Connect().


Execute SQL statement $sql and return derived class of ADORecordSet if successful. Note that a record set is always returned on success, even if we are executing an insert or update statement. You can also pass in $sql a statement prepared in Prepare().

Returns derived class of ADORecordSet. Eg. if connecting via mysql, then ADORecordSet_mysql would be returned. False is returned if there was an error in executing the sql.

The $inputarr parameter can be used for binding variables to parameters. Below is an Oracle example:

 $conn->Execute("SELECT * FROM TABLE WHERE COND=:val", array('val'=> $val));

Another example, using ODBC,which uses the ? convention:

  $conn->Execute("SELECT * FROM TABLE WHERE COND=?", array($val));
Binding variables

Variable binding speeds the compilation and caching of SQL statements, leading to higher performance. Currently Oracle, Interbase and ODBC supports variable binding. Interbase/ODBC style ? binding is emulated in databases that do not support binding. Note that you do not have to quote strings if you use binding.

Variable binding in the odbc, interbase and oci8po drivers.

$rs = $db->Execute('select * from table where val=?', array('10'));
Variable binding in the oci8 driver:
$rs = $db->Execute('select name from table where val=:key', 
  array('key' => 10));
Bulk binding

Since ADOdb 3.80, we support bulk binding in Execute(), in which you pass in a 2-dimensional array to be bound to an INSERT/UPDATE or DELETE statement.

$arr = array(
	array('Zulkifli', 24),
	array('Rosnah', 21)
$ok = $db->Execute('insert into table (name,age) values (?,?)',$arr);

This provides very high performance as the SQL statement is prepared first. The prepared statement is executed repeatedly for each array row until all rows are completed, or until the first error. Very useful for importing data.


Similar to Execute, except that the recordset is cached for $secs2cache seconds in the $ADODB_CACHE_DIR directory, and $inputarr only accepts 1-dimensional arrays. If CacheExecute() is called again with the same $sql, $inputarr, and also the same database, same userid, and the cached recordset has not expired, the cached recordset is returned.

  $ADODB_CACHE_DIR = '/usr/local/ADOdbcache';
  $conn = &ADONewConnection('mysql'); 
  $rs = $conn->CacheExecute(15, 'select * from table'); # cache 15 secs
  rs2html($rs); /* recordset to html table */  

Alternatively, since ADOdb 1.80, the $secs2cache parameter is optional:

   	$conn->cacheSecs = 3600*24; // cache 24 hours
	$rs = $conn->CacheExecute('select * from table');
If $secs2cache is omitted, we use the value in $connection->cacheSecs (default is 3600 seconds, or 1 hour). Use CacheExecute() only with SELECT statements.

Performance note: I have done some benchmarks and found that they vary so greatly that it's better to talk about when caching is of benefit. When your database server is much slower than your Web server or the database is very overloaded then ADOdb's caching is good because it reduces the load on your database server. If your database server is lightly loaded or much faster than your Web server, then caching could actually reduce performance.


Execute an Oracle stored procedure, and returns an Oracle REF cursor variable as a regular ADOdb recordset. Does not work with any other database except oci8. Thanks to Robert Tuttle for the design.

    $db = ADONewConnection("oci8"); 
    $db->Connect("", "uid", "pwd", "FOO"); 
    $rs = $db->ExecuteCursor("begin :cursorvar := getdata(:param1); end;", 
    # $rs is now just like any other ADOdb recordset object

ExecuteCursor() is a helper function that does the following internally:

	$stmt = $db->Prepare("BEGIN :RS := SP_FOO(); END;", true); 
	$db->Parameter($stmt, $cur, 'RS', false, -1, OCI_B_CURSOR);
	$rs = $db->Execute($stmt);


Returns a recordset if successful. Returns false otherwise. Performs a select statement, simulating PostgreSQL's SELECT statement, LIMIT $numrows OFFSET $offset clause.

In PostgreSQL, SELECT * FROM TABLE LIMIT 3 will return the first 3 records only. The equivalent is $connection->SelectLimit('SELECT * FROM TABLE',3). This functionality is simulated for databases that do not possess this feature.

And SELECT * FROM TABLE LIMIT 3 OFFSET 2 will return records 3, 4 and 5 (eg. after record 2, return 3 rows). The equivalent in ADOdb is $connection->SelectLimit('SELECT * FROM TABLE',3,2).

Note that this is the opposite of MySQL's LIMIT clause. You can also set $connection->SelectLimit('SELECT * FROM TABLE',-1,10) to get rows 11 to the last row.

The last parameter $inputarr is for databases that support variable binding such as Oracle oci8. This substantially reduces SQL compilation overhead. Below is an Oracle example:

 $conn->SelectLimit("SELECT * FROM TABLE WHERE COND=:val", 100,-1,array('val'=> $val));

The oci8po driver (oracle portable driver) uses the more standard bind variable of ?:

 $conn->SelectLimit("SELECT * FROM TABLE WHERE COND=?", 100,-1,array('val'=> $val));

Ron Wilson reports that SelectLimit does not work with UNIONs.

CacheSelectLimit([$secs2cache,] $sql, $numrows=-1,$offset=-1,$inputarr=false)

Similar to SelectLimit, except that the recordset returned is cached for $secs2cache seconds in the $ADODB_CACHE_DIR directory.

Since 1.80, $secs2cache has been optional, and you can define the caching time in $connection->cacheSecs.

   $conn->cacheSecs = 3600*24; // cache 24 hours
	$rs = $conn->CacheSelectLimit('select * from table',10);


Flush (delete) any cached recordsets for the SQL statement $sql in $ADODB_CACHE_DIR.

If no parameter is passed in, then all adodb_*.cache files are deleted.

If you want to flush all cached recordsets manually, execute the following PHP code (works only under Unix):
  system("rm -f `find ".$ADODB_CACHE_DIR." -name adodb_*.cache`");

For general cleanup of all expired files, you should use crontab on Unix, or at.exe on Windows, and a shell script similar to the following:
# This particular example deletes files in the TMPPATH
# directory with the string ".cache" in their name that
# are more than 7 days old.
find ${TMPPATH} -mtime +$AGED | grep "\.cache" | xargs rm -f


Returns a virtualized error number, based on PEAR DB's error number system. You might need to include before you call this function. The parameter $errno is the native error number you want to convert. If you do not pass any parameter, MetaError will call ErrorNo() for you and convert it. If the error number cannot be virtualized, MetaError will return -1 (DB_ERROR).


Pass the error number returned by MetaError() for the equivalent textual error message.


Returns the last status or error message. The error message is reset after every call to Execute().

This can return a string even if no error occurs. In general you do not need to call this function unless an ADOdb function returns false on an error.

Note: If debug is enabled, the SQL error message is always displayed when the Execute function is called.


Returns the last error number. The error number is reset after every call to Execute(). If 0 is returned, no error occurred.

Note that old versions of PHP (pre 4.0.6) do not support error number for ODBC. In general you do not need to call this function unless an ADOdb function returns false on an error.


Sets the current fetch mode for the connection and stores it in $db->fetchMode. Legal modes are ADODB_FETCH_ASSOC and ADODB_FETCH_NUM. For more info, see $ADODB_FETCH_MODE.

Returns the previous fetch mode, which could be false if SetFetchMode( ) has not been called before.

CreateSequence($seqName = 'adodbseq',$startID=1)

Create a sequence. The next time GenID( ) is called, the value returned will be $startID. Added in 2.60.

DropSequenceD($seqName = 'adodbseq')

Delete a sequence. Added in 2.60.

GenID($seqName = 'adodbseq',$startID=1)

Generate a sequence number . Works for interbase, mysql, postgresql, oci8, oci8po, mssql, ODBC based (access,vfp,db2,etc) drivers currently. Uses $seqName as the name of the sequence. GenID() will automatically create the sequence for you if it does not exist (provided the userid has permission to do so). Otherwise you will have to create the sequence yourself.

If your database driver emulates sequences, the name of the table is the sequence name. The table has one column, "id" which should be of type integer, or if you need something larger - numeric(16).

For ODBC and databases that do not support sequences natively (eg mssql, mysql), we create a table for each sequence. If the sequence has not been defined earlier, it is created with the starting value set in $startID.

Note that the mssql driver's GenID() before 1.90 used to generate 16 byte GUID's.


Allows you to store a blob (in $val) into $table into $column in a row at $where.


	# for oracle
	$conn->Execute('INSERT INTO blobtable (id, blobcol) VALUES (1, empty_blob())');
	# non oracle databases
	$conn->Execute('INSERT INTO blobtable (id, blobcol) VALUES (1, null)');

Returns true if succesful, false otherwise. Supported by MySQL, PostgreSQL, Oci8, Oci8po and Interbase drivers. Other drivers might work, depending on the state of development.

Note that when an Interbase blob is retrieved using SELECT, it still needs to be decoded using $connection->DecodeBlob($blob); to derive the original value in versions of PHP before 4.1.0.

For PostgreSQL, you can store your blob using blob oid's or as a bytea field. You can use bytea fields but not blob oid's currently with UpdateBlob( ). Conversely UpdateBlobFile( ) supports oid's, but not bytea data.

If you do not pass in an oid, then UpdateBlob() assumes that you are storing in bytea fields.


Allows you to store a clob (in $val) into $table into $column in a row at $where. Similar to UpdateBlob (see above), but for Character Large OBjects.


	# for oracle
	$conn->Execute('INSERT INTO clobtable (id, clobcol) VALUES (1, empty_clob())');
	# non oracle databases
	$conn->Execute('INSERT INTO clobtable (id, clobcol) VALUES (1, null)');


Similar to UpdateBlob, except that we pass in a file path to where the blob resides.

For PostgreSQL, if you are using blob oid's, use this interface. This interface does not support bytea fields.

Returns true if successful, false otherwise.


Some databases require blob's to be encoded manually before upload. Note if you use UpdateBlob( ) or UpdateBlobFile( ) the conversion is done automatically for you and you do not have to call this function. For PostgreSQL, currently, BlobEncode() can only be used for bytea fields.

Returns the encoded blob value.

Note that there is a connection property called blobEncodeType which has 3 legal values:

false - no need to perform encoding or decoding.
'I' - blob encoding required, and returned encoded blob is a numeric value (no need to quote).
'C' - blob encoding required, and returned encoded blob is a character value (requires quoting).

This is purely for documentation purposes, so that programs that accept multiple database drivers know what is the right thing to do when processing blobs.


Some databases require blob's to be decoded manually after doing a select statement. If the database does not require decoding, then this function will return the blob unchanged. Currently BlobDecode is only required for one database, PostgreSQL, and only if you are using blob oid's (if you are using bytea fields, we auto-decode for you).

$rs = $db->Execute("select bloboid from postgres_table where id=$key");
$blob = $db->BlobDecode( reset($rs->fields) );

Replace($table, $arrFields, $keyCols,$autoQuote=false)

Try to update a record, and if the record is not found, an insert statement is generated and executed. Returns 0 on failure, 1 if update statement worked, 2 if no record was found and the insert was executed successfully. This differs from MySQL's replace which deletes the record and inserts a new record. This also means you cannot update the primary key. The only exception to this is Interbase and its derivitives, which uses delete and insert because of some Interbase API limitations.

The parameters are $table which is the table name, the $keyCols which is an associative array where the keys are the field names, and keyCols is the name of the primary key, or an array of field names if it is a compound key. If $autoQuote is set to true, then Replace() will quote all values that are non-numeric; auto-quoting will not quote nulls. Note that auto-quoting will not work if you use SQL functions or operators.


# single field primary key
$ret = $db->Replace('atable', 
	'id',$autoquote = true);	
# generates UPDATE atable SET firstname='Harun',lastname='Al-Rashid' WHERE id=1000
# or INSERT INTO atable (id,firstname,lastname) VALUES (1000,'Harun','Al-Rashid')

# compound key
$ret = $db->Replace('atable2', 
	array('firstname'=>'Harun','lastname'=>'Al-Rashid', 'age' => 33, 'birthday' => 'null'),
	$autoquote = true);

# no auto-quoting
$ret = $db->Replace('atable2', 
	array('firstname'=>"'Harun'",'lastname'=>"'Al-Rashid'", 'age' => 'null'),

GetUpdateSQL(&$rs, $arrFields, $forceUpdate=false,$magicq=false)

Generate SQL to update a table given a recordset $rs, and the modified fields of the array $arrFields (which must be an associative array holding the column names and the new values) are compared with the current recordset. If $forceUpdate is true, then we also generate the SQL even if $arrFields is identical to $rs->fields. Requires the recordset to be associative. $magicq is used to indicate whether magic quotes are enabled (see qstr()). The field names in the array are case-insensitive.

Since 3.61, define('ADODB_FORCE_NULLS',1) and all PHP nulls will be auto-converted to SQL nulls.

GetInsertSQL(&$rs, $arrFields,$magicq=false)

Generate SQL to insert into a table given a recordset $rs. Requires the query to be associative. $magicq is used to indicate whether magic quotes are enabled (for qstr()). The field names in the array are case-insensitive.

Since 3.61, define('ADODB_FORCE_NULLS',1) and all PHP nulls will be auto-converted to SQL nulls.

PageExecute($sql, $nrows, $page, $inputarr=false)

Used for pagination of recordset. $page is 1-based. See Example 8.

CachePageExecute($secs2cache, $sql, $nrows, $page, $inputarr=false)

Used for pagination of recordset. $page is 1-based. See Example 8. Caching version of PageExecute.

Close( )

Close the database connection. PHP4 proudly states that we no longer have to clean up at the end of the connection because the reference counting mechanism of PHP4 will automatically clean up for us.

StartTrans( )

Start a monitored transaction. As SQL statements are executed, ADOdb will monitor for SQL errors, and if any are detected, when CompleteTrans() is called, we auto-rollback.

To understand why StartTrans() is superior to BeginTrans(), let us examine a few ways of using BeginTrans(). The following is the wrong way to use transactions:

$DB->Execute("update table1 set val=$val1 where id=$id");
$DB->Execute("update table2 set val=$val2 where id=$id");

because you perform no error checking. It is possible to update table1 and for the update on table2 to fail. Here is a better way:

$ok = $DB->Execute("update table1 set val=$val1 where id=$id");
if ($ok) $ok = $DB->Execute("update table2 set val=$val2 where id=$id");
if ($ok) $DB->CommitTrans();
else $DB->RollbackTrans();

Another way is (since ADOdb 2.0):

$ok = $DB->Execute("update table1 set val=$val1 where id=$id");
if ($ok) $ok = $DB->Execute("update table2 set val=$val2 where id=$id");

Now it is a headache monitoring $ok all over the place. StartTrans() is an improvement because it monitors all SQL errors for you. This is particularly useful if you are calling black-box functions in which SQL queries might be executed. Also all BeginTrans, CommitTrans and RollbackTrans calls inside a StartTrans block will be disabled, so even if the black box function does a commit, it will be ignored.

$DB->Execute("update table1 set val=$val1 where id=$id");
$DB->Execute("update table2 set val=$val2 where id=$id");

Note that a StartTrans blocks are nestable, the inner blocks are ignored.


Complete a transaction called with StartTrans(). This function monitors for SQL errors, and will commit if no errors have occured, otherwise it will rollback. Returns true on commit, false on rollback. If the parameter $autoComplete is true monitor sql errors and commit and rollback as appropriate. Set $autoComplete to false to force rollback even if no SQL error detected.

FailTrans( )

Fail a transaction started with StartTrans(). The rollback will only occur when CompleteTrans() is called.

HasFailedTrans( )

Check whether smart transaction has failed, eg. returns true if there was an error in SQL execution or FailTrans() was called. If not within smart transaction, returns false.

BeginTrans( )

Begin a transaction. Turns off autoCommit. Returns true if successful. Some databases will always return false if transaction support is not available. Any open transactions will be rolled back when the connection is closed. Among the databases that support transactions are Oracle, PostgreSQL, Interbase, MSSQL, certain versions of MySQL, DB2, Informix, Sybase, etc.

Note that StartTrans() and CompleteTrans() is a superior method of handling transactions, available since ADOdb 3.40. For a explanation, see the StartTrans() documentation.

You can also use the ADOdb error handler to die and rollback your transactions for you transparently. Some buggy database extensions are known to commit all outstanding tranasactions, so you might want to explicitly do a $DB->RollbackTrans() in your error handler for safety.

Detecting Transactions

Since ADOdb 2.50, you are able to detect when you are inside a transaction. Check that $connection->transCnt > 0. This variable is incremented whenever BeginTrans() is called, and decremented whenever RollbackTrans() or CommitTrans() is called.


End a transaction successfully. Returns true if successful. If the database does not support transactions, will return true also as data is always committed.

If you pass the parameter $ok=false, the data is rolled back. See example in BeginTrans().

RollbackTrans( )

End a transaction, rollback all changes. Returns true if successful. If the database does not support transactions, will return false as data is never rollbacked.


Returns an associative array for the given query $sql with optional bind parameters in $inputarr. If the number of columns returned is greater to two, a 2-dimensional array is returned, with the first column of the recordset becomes the keys to the rest of the rows. If the columns is equal to two, a 1-dimensional array is created, where the the keys directly map to the values (unless $force_array is set to true, when an array is created for each value).


We have the following data in a recordset:

row1: Apple, Fruit, Edible
row2: Cactus, Plant, Inedible
row3: Rose, Flower, Edible

GetAssoc will generate the following 2-dimensional associative array:

Apple => array[Fruit, Edible]
Cactus => array[Plant, Inedible]
Rose => array[Flower,Edible]

If the dataset is:

row1: Apple, Fruit
row2: Cactus, Plant
row3: Rose, Flower

GetAssoc will generate the following 1-dimensional associative array (with $force_array==false):

Apple => Fruit

The function returns:

The associative array, or false if an error occurs.

CacheGetAssoc([$secs2cache,] $sql,$inputarr=false,$force_array=false,$first2cols=false)

Caching version of GetAssoc function above.


Executes the SQL and returns the first field of the first row. The recordset and remaining rows are discarded for you automatically. If an error occur, false is returned.


Executes the SQL and returns the first row as an array. The recordset and remaining rows are discarded for you automatically. If an error occurs, false is returned.


Executes the SQL and returns the all the rows as a 2-dimensional array. The recordset is discarded for you automatically. If an error occurs, false is returned.


Executes the SQL and returns all elements of the first column as a 1-dimensional array. The recordset is discarded for you automatically. If an error occurs, false is returned.

CacheGetOne([$secs2cache,] $sql,$inputarr=false), CacheGetRow([$secs2cache,] $sql,$inputarr=false), CacheGetAll([$secs2cache,] $sql,$inputarr=false), CacheGetCol([$secs2cache,] $sql,$inputarr=false,$trim=false)

Similar to above Get* functions, except that the recordset is serialized and cached in the $ADODB_CACHE_DIR directory for $secs2cache seconds. Good for speeding up queries on rarely changing data. Note that the $secs2cache parameter is optional. If omitted, we use the value in $connection->cacheSecs (default is 3600 seconds, or 1 hour).

Prepare($sql )

Prepares (compiles) an SQL query for repeated execution. Bind parameters are denoted by ?, except for the oci8 driver, which uses the traditional Oracle :varname convention.

Returns an array containing the original sql statement in the first array element; the remaining elements of the array are driver dependent. If there is an error, or we are emulating Prepare( ), we return the original $sql string. This is because all error-handling has been centralized in Execute( ).

Prepare( ) cannot be used with functions that use SQL query rewriting techniques, e.g. PageExecute( ) and SelectLimit( ).


$stmt = $DB->Prepare('insert into table (col1,col2) values (?,?)');
for ($i=0; $i < $max; $i++)
$DB->Execute($stmt,array((string) rand(), $i));

Also see InParameter(), OutParameter() and PrepareSP() below. Only supported internally by interbase, oci8 and selected ODBC-based drivers, otherwise it is emulated. There is no performance advantage to using Prepare() with emulation.

Important: Due to limitations or bugs in PHP, if you are getting errors when you using prepared queries, try setting $ADODB_COUNTRECS = false before preparing. This behaviour has been observed with ODBC.

IfNull($field, $nullReplacementValue)

Portable IFNULL function (NVL in Oracle). Returns a string that represents the function that checks whether a $field is null for the given database, and if null, change the value returned to $nullReplacementValue. Eg.

$sql = 'SELECT '.$db->IfNull('name', "'- unknown -'"). ' FROM table';


This is not a function, but a property. Some databases have "length" and others "len" as the function to measure the length of a string. To use this property:

  $sql = "SELECT ".$db->length."(field) from table";
  $rs = $db->Execute($sql);


This is not a function, but a property. This is a string that holds the sql to generate a random number between 0.0 and 1.0 inclusive.


This is not a function, but a property. Some databases have "substr" and others "substring" as the function to retrieve a sub-string. To use this property:

  $sql = "SELECT ".$db->substr."(field, $offset, $length) from table";
  $rs = $db->Execute($sql);

For all databases, the 1st parameter of substr is the field, the 2nd is the offset (1-based) to the beginning of the sub-string, and the 3rd is the length of the sub-string.


Generates a bind placeholder portably. For most databases, the bind placeholder is "?". However some databases use named bind parameters such as Oracle, eg ":somevar". This allows us to portably define an SQL statement with bind parameters:

$sql = 'insert into table (col1,col2) values ('.$DB->Param('a').','.$DB->Param('b').')';
# generates 'insert into table (col1,col2) values (?,?)'
# or        'insert into table (col1,col2) values (:a,:b)'
$stmt = $DB->Prepare($sql);
$stmt = $DB->Execute($stmt,array('one','two'));

PrepareSP($sql, $cursor=false )

When calling stored procedures in mssql and oci8 (oracle), and you might want to directly bind to parameters that return values, or for special LOB handling. PrepareSP() allows you to do so.

Returns the same array or $sql string as Prepare( ) above. If you do not need to bind to return values, you should use Prepare( ) instead.

The 2nd parameter, $cursor is not used except with oci8. Setting it to true will force OCINewCursor to be called; this is to support output REF CURSORs.

For examples of usage of PrepareSP( ), see InParameter( ) below.

Note: in the mssql driver, preparing stored procedures requires a special function call, mssql_init( ), which is called by this function. PrepareSP( ) is available in all other drivers, and is emulated by calling Prepare( ).

InParameter($stmt, $var, $name, $maxLen = 4000, $type = false )

Binds a PHP variable as input to a stored procedure variable. The parameter $stmt is the value returned by PrepareSP(), $var is the PHP variable you want to bind, $name is the name of the stored procedure variable. Optional is $maxLen, the maximum length of the data to bind, and $type which is database dependant. Consult mssql_bind and ocibindbyname docs at for more info on legal values for $type.

InParameter() is a wrapper function that calls Parameter() with $isOutput=false. The advantage of this function is that it is self-documenting, because the $isOutput parameter is no longer needed. Only for mssql and oci8 currently.

Here is an example using oci8:

# For oracle, Prepare and PrepareSP are identical
$stmt = $db->PrepareSP(
	"declare RETVAL integer; 
	:RETVAL := SP_RUNSOMETHING(:myid,:group);

The same example using mssql:

# @RETVAL = SP_RUNSOMETHING @myid,@group
$stmt = $db->PrepareSP('SP_RUNSOMETHING'); 
# note that the parameter name does not have @ in front! $db->InParameter($stmt,$id,'myid'); $db->InParameter($stmt,$group,'group',64); # return value in mssql - RETVAL is hard-coded name $db->OutParameter($stmt,$ret,'RETVAL'); $db->Execute($stmt);

Note that the only difference between the oci8 and mssql implementations is $sql.

If $type parameter is set to false, in mssql, $type will be dynamicly determined based on the type of the PHP variable passed (string => SQLCHAR, boolean =>SQLINT1, integer =>SQLINT4 or float/double=>SQLFLT8).

In oci8, $type can be set to OCI_B_FILE (Binary-File), OCI_B_CFILE (Character-File), OCI_B_CLOB (Character-LOB), OCI_B_BLOB (Binary-LOB) and OCI_B_ROWID (ROWID). To pass in a null, use $db->Parameter($stmt, $null=null, 'param').

OutParameter($stmt, $var, $name, $maxLen = 4000, $type = false )

Binds a PHP variable as output from a stored procedure variable. The parameter $stmt is the value returned by PrepareSP(), $var is the PHP variable you want to bind, $name is the name of the stored procedure variable. Optional is $maxLen, the maximum length of the data to bind, and $type which is database dependant.

OutParameter() is a wrapper function that calls Parameter() with $isOutput=true. The advantage of this function is that it is self-documenting, because the $isOutput parameter is no longer needed. Only for mssql and oci8 currently.

For an example, see InParameter.

Parameter($stmt, $var, $name, $isOutput=false, $maxLen = 4000, $type = false )

Note: This function is deprecated, because of the new InParameter() and OutParameter() functions. These are superior because they are self-documenting, unlike Parameter().

Adds a bind parameter suitable for return values or special data handling (eg. LOBs) after a statement has been prepared using PrepareSP(). Only for mssql and oci8 currently. The parameters are:

$stmt Statement returned by Prepare() or PrepareSP().
$var PHP variable to bind to. Make sure you pre-initialize it!
$name Name of stored procedure variable name to bind to.
[$isOutput] Indicates direction of parameter 0/false=IN 1=OUT 2= IN/OUT. This is ignored in oci8 as this driver auto-detects the direction.
[$maxLen] Maximum length of the parameter variable.
[$type] Consult mssql_bind and ocibindbyname docs at for more info on legal values for type.

Lastly, in oci8, bind parameters can be reused without calling PrepareSP( ) or Parameters again. This is not possible with mssql. An oci8 example:

$id = 0; $i = 0;
$stmt = $db->PrepareSP( "update table set val=:i where id=:id");
$db->Parameter($stmt,$i, 'i');
for ($cnt=0; $cnt < 1000; $cnt++) {
	$id = $cnt; 
$i = $cnt * $cnt; # works with oci8! $db->Execute($stmt);

Bind($stmt, $var, $size=4001, $type=false, $name=false)

This is a low-level function supported only by the oci8 driver. Avoid using unless you only want to support Oracle. The Parameter( ) function is the recommended way to go with bind variables.

Bind( ) allows you to use bind variables in your sql statement. This binds a PHP variable to a name defined in an Oracle sql statement that was previously prepared using Prepare(). Oracle named variables begin with a colon, and ADOdb requires the named variables be called :0, :1, :2, :3, etc. The first invocation of Bind() will match :0, the second invocation will match :1, etc. Binding can provide 100% speedups for insert, select and update statements.

The other variables, $size sets the buffer size for data storage, $type is the optional descriptor type OCI_B_FILE (Binary-File), OCI_B_CFILE (Character-File), OCI_B_CLOB (Character-LOB), OCI_B_BLOB (Binary-LOB) and OCI_B_ROWID (ROWID). Lastly, instead of using the default :0, :1, etc names, you can define your own bind-name using $name.

The following example shows 3 bind variables being used: p1, p2 and p3. These variables are bound to :0, :1 and :2.

$stmt = $DB->Prepare("insert into table (col0, col1, col2) values (:0, :1, :2)");
$DB->Bind($stmt, $p1);
$DB->Bind($stmt, $p2);
$DB->Bind($stmt, $p3);
for ($i = 0; $i < $max; $i++) { 
   $p1 = ?; $p2 = ?; $p3 = ?;

You can also use named variables:

$stmt = $DB->Prepare("insert into table (col0, col1, col2) values (:name0, :name1, :name2)");
$DB->Bind($stmt, $p1, "name0");
$DB->Bind($stmt, $p2, "name1");
$DB->Bind($stmt, $p3, "name2");
for ($i = 0; $i < $max; $i++) { 
   $p1 = ?; $p2 = ?; $p3 = ?;


Call this method to install a SQL logging and timing function (using fnExecute). Then all SQL statements are logged into an adodb_logsql table in a database. If the adodb_logsql table does not exist, ADOdb will create the table if you have the appropriate permissions. Returns the previous logging value (true for enabled, false for disabled). Here are samples of the DDL for selected databases:

		CREATE TABLE adodb_logsql (
		  created datetime NOT NULL,
		  sql0 varchar(250) NOT NULL,
		  sql1 text NOT NULL,
		  params text NOT NULL,
		  tracer text NOT NULL,
		  timer decimal(16,6) NOT NULL
		CREATE TABLE adodb_logsql (
		  created timestamp NOT NULL,
		  sql0 varchar(250) NOT NULL,
		  sql1 text NOT NULL,
		  params text NOT NULL,
		  tracer text NOT NULL,
		  timer decimal(16,6) NOT NULL
		CREATE TABLE adodb_logsql (
		  created datetime NOT NULL,
		  sql0 varchar(250) NOT NULL,
		  sql1 varchar(4000) NOT NULL,
		  params varchar(3000) NOT NULL,
		  tracer varchar(500) NOT NULL,
		  timer decimal(16,6) NOT NULL
		CREATE TABLE adodb_logsql (
		  created date NOT NULL,
		  sql0 varchar(250) NOT NULL,
		  sql1 varchar(4000) NOT NULL,
		  params varchar(4000),
		  tracer varchar(4000),
		  timer decimal(16,6) NOT NULL
	$conn->LogSQL(); // turn on logging
	$conn->LogSQL(false); // turn off logging
	# output summary of SQL logging results
	$perf = NewPerfMonitor($conn);
	echo $perf->SuspiciousSQL();
	echo $perf->ExpensiveSQL();

One limitation of logging is that rollback also prevents SQL from being logged.

If you prefer to use another name for the table used to store the SQL, you can override it by calling adodb_perf::table($tablename), where $tablename is the new table name (you will still need to manually create the table yourself). An example:

Also see Performance Monitor.

fnExecute and fnCacheExecute properties

These two properties allow you to define bottleneck functions for all sql statements processed by ADOdb. This allows you to perform statistical analysis and query-rewriting of your sql.

Examples of fnExecute

Here is an example of using fnExecute, to count all cached queries and non-cached queries, you can do this:

# $db is the connection object
function CountExecs($db, $sql, $inputarray)
global $EXECS;

if (!is_array(inputarray)) $EXECS++;
# handle 2-dimensional input arrays
else if (is_array(reset($inputarray))) $EXECS += sizeof($inputarray);
else $EXECS++;

# $db is the connection object
function CountCachedExecs($db, $secs2cache, $sql, $inputarray)
global $CACHED; $CACHED++; }
$db = NewADOConnection('mysql'); $db->Connect(...); $db->fnExecute = 'CountExecs'; $db->fnCacheExecute = 'CountCachedExecs'; : :
# After many sql statements:` printf("<p>Total queries=%d; total cached=%d</p>",$EXECS+$CACHED, $CACHED);

The fnExecute function is called before the sql is parsed and executed, so you can perform a query rewrite. If you are passing in a prepared statement, then $sql is an array (see Prepare). The fnCacheExecute function is only called if the recordset returned was cached. The function parameters match the Execute and CacheExecute functions respectively, except that $this (the connection object) is passed as the first parameter.

Since ADOdb 3.91, the behaviour of fnExecute varies depending on whether the defined function returns a value. If it does not return a value, then the $sql is executed as before. This is useful for query rewriting or counting sql queries.

On the other hand, you might want to replace the Execute function with one of your own design. If this is the case, then have your function return a value. If a value is returned, that value is returned immediately, without any further processing. This is used internally by ADOdb to implement LogSQL() functionality.

ADOConnection Utility Functions


No longer available - removed since 1.99.


Generates the sql string used to concatenate $s1, $s2, etc together. Uses the string in the concat_operator field to generate the concatenation. Override this function if a concatenation operator is not used, eg. MySQL.

Returns the concatenated string.


Format the $date in the format the database accepts. This is used in INSERT/UPDATE statements; for SELECT statements, use SQLDate. The $date parameter can be a Unix integer timestamp or an ISO format Y-m-d. Uses the fmtDate field, which holds the format to use. If null or false or '' is passed in, it will be converted to an SQL null.

Returns the date as a quoted string.


Format the timestamp $ts in the format the database accepts; this can be a Unix integer timestamp or an ISO format Y-m-d H:i:s. Uses the fmtTimeStamp field, which holds the format to use. If null or false or '' is passed in, it will be converted to an SQL null.

Returns the timestamp as a quoted string.


Quotes a string to be sent to the database. The $magic_quotes_enabled parameter may look funny, but the idea is if you are quoting a string extracted from a POST/GET variable, then pass get_magic_quotes_gpc() as the second parameter. This will ensure that the variable is not quoted twice, once by qstr and once by the magic_quotes_gpc.

Eg. $s = $db->qstr(HTTP_GET_VARS['name'],get_magic_quotes_gpc());

Returns the quoted string.


Quotes the string $s, escaping the database specific quote character as appropriate. Formerly checked magic quotes setting, but this was disabled since 3.31 for compatibility with PEAR DB.

Affected_Rows( )

Returns the number of rows affected by a update or delete statement. Returns false if function not supported.

Not supported by interbase/firebird currently.

Insert_ID( )

Returns the last autonumbering ID inserted. Returns false if function not supported.

Only supported by databases that support auto-increment or object id's, such as PostgreSQL, MySQL and MS SQL Server currently. PostgreSQL returns the OID, which can change on a database reload.


Lock a table row for the duration of a transaction. For example to lock record $id in table1:


Supported in db2, interbase, informix, mssql, oci8, postgres, sybase.


Returns a list of databases available on the server as an array. You have to connect to the server first. Only available for ODBC, MySQL and ADO.

MetaTables($ttype = false, $showSchema = false, $mask=false)

Returns an array of tables and views for the current database as an array. The array should exclude system catalog tables if possible. To only show tables, use $db->MetaTables('TABLES'). To show only views, use $db->MetaTables('VIEWS'). The $showSchema parameter currently works only for DB2, and when set to true, will add the schema name to the table, eg. "SCHEMA.TABLE".

You can define a mask for matching. For example, setting $mask = 'TMP%' will match all tables that begin with 'TMP'. Currently only mssql, oci8, odbc_mssql and postgres* support $mask.


Returns an array of ADOFieldObject's, one field object for every column of $table. A field object is a class instance with (name, type, max_length) defined. Currently Sybase does not recognise date types, and ADO cannot identify the correct data type (so we default to varchar).

The $toupper parameter determines whether we uppercase the table name (required for some databases).

For schema support, pass in the $table parameter, "$schema.$tablename". This is only supported for selected databases.


Returns an array of column names for $table.

MetaPrimaryKeys($table, $owner=false)

Returns an array containing column names that are the primary keys of $table. Supported by mysql, odbc (including db2, odbc_mssql, etc), mssql, postgres, interbase/firebird, oci8 currently.

Views (and some tables) have primary keys, but sometimes this information is not available from the database. You can define a function ADODB_View_PrimaryKeys($databaseType, $database, $view, $owner) that should return an array containing the fields that make up the primary key. If that function exists, it will be called when MetaPrimaryKeys() cannot find a primary key for a table or view.

// In this example: dbtype = 'oci8', $db = 'mydb', $view = 'dataView', $owner = false 
function ADODB_View_PrimaryKeys($dbtype,$db,$view,$owner)
	switch(strtoupper($view)) {
	case 'DATAVIEW': return array('DATAID');
	default: return false;

$db = NewADOConnection('oci8');


Returns an array of containing two elements 'description' and 'version'. The 'description' element contains the string description of the database. The 'version' naturally holds the version number (which is also a string).

MetaForeignKeys($table, $owner=false, $upper=false)

Returns an associate array of foreign keys, or false if not supported. For example, if table employee has a foreign key where employee.deptkey points to dept_table.deptid, and employee.posn=posn_table.postionid and employee.poscategory=posn_table.category, then $conn->MetaForeignKeys('employee') will return

		'dept_table' => array('deptkey=deptid'),
		'posn_table' => array('posn=positionid','poscategory=category')

The optional schema or owner can be defined in $owner. If $upper is true, then the table names (array keys) are upper-cased.


When an SQL statement successfully is executed by ADOConnection->Execute($sql),an ADORecordSet object is returned. This object contains a virtual cursor so we can move from row to row, functions to obtain information about the columns and column types, and helper functions to deal with formating the results to show to the user.

ADORecordSet Fields

fields: Array containing the current row. This is not associative, but is an indexed array from 0 to columns-1. See also the function Fields, which behaves like an associative array.

dataProvider: The underlying mechanism used to connect to the database. Normally set to native, unless using odbc or ado.

blobSize: Maximum size of a char, string or varchar object before it is treated as a Blob (Blob's should be shown with textarea's). See the MetaType function.

sql: Holds the sql statement used to generate this record set.

canSeek: Set to true if Move( ) function works.

EOF: True if we have scrolled the cursor past the last record.

ADORecordSet Functions

ADORecordSet( )

Constructer. Normally you never call this function yourself.


Generates an associative array from the recordset. Note that is this function is also available in the connection object. More details can be found there.


Generate a 2-dimensional array of records from the current cursor position, indexed from 0 to $number_of_rows - 1. If $number_of_rows is undefined, till EOF.


Generate a 2-dimensional array of records from the current cursor position. Synonym for GetArray() for compatibility with Microsoft ADO.

GetMenu($name, [$default_str=''], [$blank1stItem=true], [$multiple_select=false], [$size=0], [$moreAttr=''])

Generate a HTML menu (<select><option><option></select>). The first column of the recordset (fields[0]) will hold the string to display in the option tags. If the recordset has more than 1 column, the second column (fields[1]) is the value to send back to the web server.. The menu will be given the name $name.

If $default_str is defined, then if $default_str == fields[0], that field is selected. If $blank1stItem is true, the first option is empty. You can also set the first option strings by setting $blank1stItem = "$value:$text".

$Default_str can be array for a multiple select listbox.

To get a listbox, set the $size to a non-zero value (or pass $default_str as an array). If $multiple_select is true then a listbox will be generated with $size items (or if $size==0, then 5 items) visible, and we will return an array to a server. Lastly use $moreAttr to add additional attributes such as javascript or styles.

Menu Example 1: GetMenu('menu1','A',true) will generate a menu: for the data (A,1), (B,2), (C,3). Also see example 5.

Menu Example 2: For the same data, GetMenu('menu1',array('A','B'),false) will generate a menu with both A and B selected:

GetMenu2($name, [$default_str=''], [$blank1stItem=true], [$multiple_select=false], [$size=0], [$moreAttr=''])

This is nearly identical to GetMenu, except that the $default_str is matched to fields[1] (the option values).

Menu Example 3: Given the data in menu example 2, GetMenu2('menu1',array('1','2'),false) will generate a menu with both A and B selected in menu example 2, but this time the selection is based on the 2nd column, which holds the values to return to the Web server.

UserDate($str, [$fmt])

Converts the date string $str to another format. The date format is Y-m-d, or Unix timestamp format. The default $fmt is Y-m-d.

UserTimeStamp($str, [$fmt])

Converts the timestamp string $str to another format. The timestamp format is Y-m-d H:i:s, as in '2002-02-28 23:00:12', or Unix timestamp format. UserTimeStamp calls UnixTimeStamp to parse $str, and $fmt defaults to Y-m-d H:i:s if not defined.


Parses the date string $str and returns it in unix mktime format (eg. a number indicating the seconds after January 1st, 1970). Expects the date to be in Y-m-d H:i:s format, except for Sybase and Microsoft SQL Server, where M d Y is also accepted (the 3 letter month strings are controlled by a global array, which might need localisation).

This function is available in both ADORecordSet and ADOConnection since 1.91.


Parses the timestamp string $str and returns it in unix mktime format (eg. a number indicating the seconds after January 1st, 1970). Expects the date to be in "Y-m-d, H:i:s" (1970-12-24, 00:00:00) or "Y-m-d H:i:s" (1970-12-24 00:00:00) or "YmdHis" (19701225000000) format, except for Sybase and Microsoft SQL Server, where "M d Y h:i:sA" (Dec 25 1970 00:00:00AM) is also accepted (the 3 letter month strings are controlled by a global array, which might need localisation).

This function is available in both ADORecordSet and ADOConnection since 1.91.

OffsetDate($dayFraction, $basedate=false)

Returns a string with the native SQL functions to calculate future and past dates based on $basedate in a portable fashion. If $basedate is not defined, then the current date (at 12 midnight) is used. Returns the SQL string that performs the calculation when passed to Execute().

For example, in Oracle, to find the date and time that is 2.5 days from today, you can use:

# get date one week from now
$fld = $conn->OffsetDate(7); // returns "(trunc(sysdate)+7")
# get date and time that is 60 hours from current date and time
$fld = $conn->OffsetDate(2.5, $conn->sysTimeStamp);	// returns "(sysdate+2.5)"

$conn->Execute("UPDATE TABLE SET dodate=$fld WHERE ID=$id");

This function is available for mysql, mssql, oracle, oci8 and postgresql drivers since 2.13. It might work with other drivers provided they allow performing numeric day arithmetic on dates.

SQLDate($dateFormat, $basedate=false)

Returns a string which contains the native SQL functions to format a date or date column $basedate. This is used in SELECT statements. For INSERT/UPDATE statements, use DBDate. It uses a case-sensitive $dateFormat, which supports:
 Y: 4-digit Year
 Q: Quarter (1-4)
 M: Month (Jan-Dec)
 m: Month (01-12)
 d: Day (01-31)
 H: Hour (00-23)
 h: Hour (1-12)
 i: Minute (00-59)
 s: Second (00-60)
 A: AM/PM indicator

All other characters are treated as strings. You can also use \ to escape characters. Available on selected databases, including mysql, postgresql, mssql, oci8 and DB2.

This is useful in writing portable sql statements that GROUP BY on dates. For example to display total cost of goods sold broken by quarter (dates are stored in a field called postdate):

 $sqlfn = $db->SQLDate('Y-\QQ','postdate'); # get sql that formats postdate to output 2002-Q1
 $sql = "SELECT $sqlfn,SUM(cogs) FROM table GROUP BY $sqlfn ORDER BY 1 desc";

MoveNext( )

Move the internal cursor to the next row. The $this->fields array is automatically updated. Returns false if unable to do so (normally because EOF has been reached), otherwise true.

If EOF is reached, then the $this->fields array is set to false (this was only implemented consistently in ADOdb 3.30). For the pre-3.30 behaviour of $this->fields (at EOF), set the global variable $ADODB_COMPAT_FETCH = true.


$rs = $db->Execute($sql);
if ($rs) 
	while (!$rs->EOF) {


Moves the internal cursor to a specific row $to. Rows are zero-based eg. 0 is the first row. The fields array is automatically updated. For databases that do not support scrolling internally, ADOdb will simulate forward scrolling. Some databases do not support backward scrolling. If the $to position is after the EOF, $to will move to the end of the RecordSet for most databases. Some obscure databases using odbc might not behave this way.

Note: This function uses absolute positioning, unlike Microsoft's ADO.

Returns true or false. If false, the internal cursor is not moved in most implementations, so AbsolutePosition( ) will return the last cursor position before the Move( ).


Internally calls Move(0). Note that some databases do not support this function.


Internally calls Move(RecordCount()-1). Note that some databases do not support this function.


Returns an associative array containing the current row. The keys to the array are the column names. The column names are upper-cased for easy access. To get the next row, you will still need to call MoveNext().

For example:
Array ( [ID] => 1 [FIRSTNAME] => Caroline [LASTNAME] => Miranda [CREATED] => 2001-07-05 )

Note: do not use GetRowAssoc() with $ADODB_FETCH_MODE = ADODB_FETCH_ASSOC. Because they have the same functionality, they will interfere with each other.


Returns the current page. Requires PageExecute()/CachePageExecute() to be called. See Example 8.


Returns true if at first page (1-based). Requires PageExecute()/CachePageExecute() to be called. See Example 8.


Returns true if at last page (1-based). Requires PageExecute()/CachePageExecute() to be called. See Example 8.


Returns the value of the associated column $colname for the current row. The column name is case-insensitive.

This is a convenience function. For higher performance, use $ADODB_FETCH_MODE.


Returns array containing current row, or false if EOF. FetchRow( ) internally moves to the next record after returning the current row.

Warning: Do not mix using FetchRow() with MoveNext().


$rs = $db->Execute($sql);
if ($rs)
	while ($arr = $rs->FetchRow()) {
	     # process $arr	


Sets $array to the current row. Returns PEAR_Error object if EOF, 1 if ok (DB_OK constant). If PEAR is undefined, false is returned when EOF. FetchInto( ) internally moves to the next record after returning the current row.

FetchRow() is easier to use. See above.


Returns an object containing the name, type and max_length of the associated field. If the max_length cannot be determined reliably, it will be set to -1. The column numbers are zero-based. See example 2.

FieldCount( )

Returns the number of fields (columns) in the record set.

RecordCount( )

Returns the number of rows in the record set. If the number of records returned cannot be determined from the database driver API, we will buffer all rows and return a count of the rows after all the records have been retrieved. This buffering can be disabled (for performance reasons) by setting the global variable $ADODB_COUNTRECS = false. When disabled, RecordCount( ) will return -1 for certain databases. See the supported databases list above for more details.

RowCount is a synonym for RecordCount.

PO_RecordCount($table, $where)

Returns the number of rows in the record set. If the database does not support this, it will perform a SELECT COUNT(*) on the table $table, with the given $where condition to return an estimate of the recordset size.

$numrows = $rs->PO_RecordCount("articles_table", "group=$group");


For databases that allow multiple recordsets to be returned in one query, this function allows you to switch to the next recordset. Currently only supported by mssql driver.

$rs = $db->Execute('execute return_multiple_rs');
$arr1 = $rs->GetArray();
$arr2 = $rs->GetArray();


Returns the current row as an object. If you set $toupper to true, then the object fields are set to upper-case. Note: The newer FetchNextObject() is the recommended way of accessing rows as objects. See below.


Gets the current row as an object and moves to the next row automatically. Returns false if at end-of-file. If you set $toupper to true, then the object fields are set to upper-case.

$rs = $db->Execute('select firstname,lastname from table');
if ($rs) {
	while ($o = $rs->FetchNextObject()) {
		print "$o->FIRSTNAME, $o->LASTNAME<BR>";

There is some trade-off in speed in using FetchNextObject(). If performance is important, you should access rows with the fields[] array. FetchObj()

Returns the current record as an object. Fields are not upper-cased, unlike FetchObject.


Returns the current record as an object and moves to the next record. If EOF, false is returned. Fields are not upper-cased, unlike FetctNextObject.

CurrentRow( )

Returns the current row of the record set. 0 is the first row.

AbsolutePosition( )

Synonym for CurrentRow for compatibility with ADO. Returns the current row of the record set. 0 is the first row.


Determine what generic meta type a database field type is given its native type $nativeDBType as a string and the length of the field $field_max_length. Note that field_max_length can be -1 if it is not known. The field object returned by FetchField() can be passed in $fieldobj or as the 1st parameter $nativeDBType. This is useful for databases such as mysql which has additional properties in the field object such as primary_key.

Uses the field blobSize and compares it with $field_max_length to determine whether the character field is actually a blob.

For example, $db->MetaType('char') will return 'C'.


  • C: Character fields that should be shown in a <input type="text"> tag.
  • X: Clob (character large objects), or large text fields that should be shown in a <textarea>
  • D: Date field
  • T: Timestamp field
  • L: Logical field (boolean or bit-field)
  • N: Numeric field. Includes decimal, numeric, floating point, and real.
  • I:  Integer field.
  • R: Counter or Autoincrement field. Must be numeric.
  • B: Blob, or binary large objects.

Since ADOdb 3.0, MetaType accepts $fieldobj as the first parameter, instead of $nativeDBType.

Close( )

Closes the recordset, cleaning all memory and resources associated with the recordset.

If memory management is not an issue, you do not need to call this function as recordsets are closed for you by PHP at the end of the script. SQL statements such as INSERT/UPDATE/DELETE do not really return a recordset, so you do not have to call Close() for such SQL statements.

function rs2html($adorecordset,[$tableheader_attributes], [$col_titles])

This is a standalone function (rs2html = recordset to html) that is similar to PHP's odbc_result_all function, it prints a ADORecordSet, $adorecordset as a HTML table. $tableheader_attributes allow you to control the table cellpadding, cellspacing and border attributes. Lastly you can replace the database column names with your own column titles with the array $col_titles. This is designed more as a quick debugging mechanism, not a production table recordset viewer.

You will need to include the file

Example of rs2html:

include(''); # load code common to ADOdb 
include(''); # load code common to ADOdb 
$conn = &ADONewConnection('mysql');   # create a connection 
$conn->PConnect('localhost','userid','','agora');# connect to MySQL, agora db
$sql = 'select CustomerName, CustomerID from customers'; 
$rs   = $conn->Execute($sql); 
rs2html($rs,'border=2 cellpadding=3',array('Customer Name','Customer ID'));

Differences between this ADOdb library and Microsoft ADO

  1. ADOdb only supports recordsets created by a connection object. Recordsets cannot be created independently.
  2. ADO properties are implemented as functions in ADOdb. This makes it easier to implement any enhanced ADO functionality in the future.
  3. ADOdb's ADORecordSet->Move() uses absolute positioning, not relative. Bookmarks are not supported.
  4. ADORecordSet->AbsolutePosition() cannot be used to move the record cursor.
  5. ADO Parameter objects are not supported. Instead we have the ADOConnection::Parameter( ) function, which provides a simpler interface for calling preparing parameters and calling stored procedures.
  6. Recordset properties for paging records are available, but implemented as in Example 8.

Database Driver Guide

This describes how to create a class to connect to a new database. To ensure there is no duplication of work, kindly email me at if you decide to create such a class.

First decide on a name in lower case to call the database type. Let's say we call it xbase.

Then we need to create two classes ADODB_xbase and ADORecordSet_xbase in the file

The simplest form of database driver is an adaptation of an existing ODBC driver. Then we just need to create the class ADODB_xbase extends ADODB_odbc to support the new date and timestamp formats, the concatenation operator used, true and false. For the ADORecordSet_xbase extends ADORecordSet_odbc we need to change the MetaType function. See as an example.

More complicated is a totally new database driver that connects to a new PHP extension. Then you will need to implement several functions. Fortunately, you do not have to modify most of the complex code. You only need to override a few stub functions. See for example.

The default date format of ADOdb internally is YYYY-MM-DD (Ansi-92). All dates should be converted to that format when passing to an ADOdb date function. See Oracle for an example how we use ALTER SESSION to change the default date format in _pconnect _connect.

ADOConnection Functions to Override

Defining a constructor for your ADOConnection derived function is optional. There is no need to call the base class constructor.

_connect: Low level implementation of Connect. Returns true or false. Should set the _connectionID.

_pconnect: Low level implemention of PConnect. Returns true or false. Should set the _connectionID.

_query: Execute a query. Returns the queryID, or false.

_close: Close the connection -- PHP should clean up all recordsets.

ErrorMsg: Stores the error message in the private variable _errorMsg.

ADOConnection Fields to Set

_bindInputArray: Set to true if binding of parameters for SQL inserts and updates is allowed using ?, eg. as with ODBC.







hasLimit support SELECT * FROM TABLE LIMIT 10 of MySQL.

hasTop support Microsoft style SELECT TOP 10 * FROM TABLE.

ADORecordSet Functions to Override

You will need to define a constructor for your ADORecordSet derived class that calls the parent class constructor.

FetchField: as documented above in ADORecordSet

_initrs: low level initialization of the recordset: setup the _numOfRows and _numOfFields fields -- called by the constructor.

_seek: seek to a particular row. Do not load the data into the fields array. This is done by _fetch. Returns true or false. Note that some implementations such as Interbase do not support seek. Set canSeek to false.

_fetch: fetch a row using the database extension function and then move to the next row. Sets the fields array. If the parameter $ignore_fields is true then there is no need to populate the fields array, just move to the next row. then Returns true or false.

_close: close the recordset

Fields: If the array row returned by the PHP extension is not an associative one, you will have to override this. See for an example. For databases such as MySQL and MSSQL where an associative array is returned, there is no need to override this function.

ADOConnection Fields to Set

canSeek: Set to true if the _seek function works.


See the RoadMap article.

Also see the ADOdb proxy article for bridging Windows and Unix databases using http remote procedure calls. For your education, visit for database info, and read this article on Optimizing PHP.

Change Log

4.21 20 Mar 2004

Pim Koeman contributed dutch language file

Rick Hickerson added CLOB support to db2 datadict.

Added odbtp driver. Thx to "stefan bogdan"

Changed PrepareSP() 2nd parameter, $cursor, to default to true (formerly false). Fixes oci8 backward compat problems with OUT params.

Fixed month calculation error in 2102-June-01 appeared as 2102-May-32.

Updated PHP5 RC1 iterator support. API changed, hasMore() renamed to valid().

Changed internal format of serialized cache recordsets. As we store a version number, this should be backward compatible.

Error handling when driver file not found was flawed in ADOLoadCode(). Fixed.

4.20 27 Feb 2004

Updated to AXMLS 1.01.

MetaForeignKeys for postgres7 modified by Edward Jaramilla, works on pg 7.4.

Now numbers accepts function calls or sequences for GetInsertSQL/GetUpdateSQL numeric fields.

Changed quotes of 'delete from $perf_table' to "". Thx Kehui (

Added ServerInfo() for ifx, and putenv trim fix. Thx Fernando Ortiz.

Added addq(), which is analogous to addslashes().

Tested with php5b4. Fix some php5 compat problems with exceptions and sybase.

Carl-Christian Salvesen added patch to mssql _query to support binds greater than 4000 chars.

Mike suggested patch to PHP5 exception handler. $errno must be numeric.

Added double quotes (") to ADODB_TABLE_REGEX.

For oci8, Prepare(...,$cursor), $cursor's meaning was accidentally inverted in 4.11. This causes problems with ExecuteCursor() too, which calls Prepare() internally. Thx to William Lovaton.

Now dateHasTime property in connection object renamed to datetime for consistency. This could break bc.

Csongor Halmai reports that db2 SelectLimit with input array is not working. Fixed..

4.11 27 Jan 2004

Csongor Halmai reports db2 binding not working. Reverted back to emulated binding.

Dan Cech modifies datadict code. Adds support for DropIndex. Minor cleanups.

Table misspelt in Changed v$conn_cache_advice to v$db_cache_advice. Reported by Steve W.

UserTimeStamp and DBTimeStamp did not handle YYYYMMDDHHMMSS format properly. Reported by Mike Muir. Fixed.

Changed oci8 Prepare(). Does not auto-allocate OCINewCursor automatically, unless 2nd param is set to true. This will break backward compat, if Prepare/Execute is used instead of ExecuteCursor. Reported by Chris Jones.

Added InParameter() and OutParameter(). Wrapper functions to Parameter(), but nicer because they are self-documenting.

Added 'R' handling in ActualType() to

Added ADOConnection::SerializableRS($rs). Returns a recordset that can be serialized in a session.

Added "Run SQL" to performance UI().

Misc spelling corrections in, and, from Heinz Hombergs.

MetaIndexes() for ibase contributed by Heinz Hombergs.

4.10 12 Jan 2004

Dan Cech contributed extensive changes to data dictionary to support name quoting (with `), and drop table/index.

Informix added cursorType property. Default remains IFX_SCROLL, but you can change to 0 (non-scrollable cursor) for performance.

Added ADODB_View_PrimaryKeys() for returning view primary keys to MetaPrimaryKeys().

Simplified chinese file, from cysoft.

Added check for ctype_alnum in Thx to Jason Judge.

Added connection parameter to ibase Prepare(). Fix by Daniel Hassan.

Added nameQuote for quoting identifiers and names to connection obj. Requested by Jason Judge. Also the data dictionary parser now detects `field name` and generates column names with spaces correctly.

BOOL type not recognised correctly as L. Fixed.

Fixed paths in ADODB_DIR for session files, and back-ported it to 4.05 (15 Dec 2003)

Added Schema to postgresql MetaTables. Thx to

Empty postgresql recordsets that had blob fields did not set EOF properly. Fixed.

CacheSelectLimit internal parameters to SelectLimit were wrong. Thx to Nio.

Modified adodb_pr() and adodb_backtrace() to support command-line usage (eg. no html).

Fixed some fr and it lang errors. Thx to Gaetano G.

Added contrib directory, with adodb rs to xmlrpc convertor by Gaetano G.

Fixed array recordset bugs when _skiprow1 is true. Thx to Gaetano G.

Fixed pivot table code when count is false.

4.05 13 Dec 2003

Added MetaIndexes - thx to Dan Cech.

Rewritten session code by Ross Smith. Moved code to adodb/session directory.

Added function exists check on connecting to most drivers, so we don't crash with the unknown function error.

Smart Transactions failed with GenID() when it no seq table has been created because the sql statement fails. Fix by Mark Newnham.

Added $db->length, which holds name of function that returns strlen.

Fixed error handling for bad driver in ADONewConnection - passed too few params to error-handler.

Datadict did not handle types like 16.0 properly in _GetSize. Fixed.

Oci8 driver SelectLimit() bug &= instead of =& used. Thx to Swen Thümmler.

Jesse Mullan suggested not flushing outp when output buffering enabled. Due to Apache 2.0 bug. Added.

MetaTables/MetaColumns return ref bug with PHP5 fixed in

New mysqli driver contributed by Arjen de Rijke. Based on adodb 3.40 driver. Then jlim added BeginTrans, CommitTrans, RollbackTrans, IfNull, SQLDate. Also fixed return ref bug.

$ADODB_FLUSH added, if true then force flush in debugging outp. Default is false. In earlier versions, outp defaulted to flush, which is not compat with apache 2.0.

Mysql driver's GenID() function did not work when when sql logging is on. Fixed.

$ADODB_SESSION_TBL not declared as global var. Not available if included in function. Fixed.

The input array not passed to Execute() in _adodb_getcount(). Fixed.

4.04 13 Nov 2003

Switched back to foreach - faster than list-each.

Fixed bug in ado driver - wiping out $this->fields with date fields.

Performance Monitor, View SQL, Explain Plan did not work if strlen($SQL)>max($_GET length). Fixed.

Performance monitor, oci8 driver added memory sort ratio.

Added random property, returns SQL to generate a floating point number between 0 and 1;

4.03 6 Nov 2003

The path to and was not setup properly.

Patched SQLDate in interbase to support hours/mins/secs. Thx to ari kuorikoski.

Force autorollback for pgsql persistent connections - apparently pgsql did not autorollback properly before 4.3.4. See

4.02 5 Nov 2003

Some errors in adodb_error_pg() fixed. Thx to Styve.

Spurious Insert_ID() error was generated by LogSQL(). Fixed.

Insert_ID was interfering with Affected_Rows() and Replace() when LogSQL() enabled. Fixed.

More foreach loops optimized with list/each.

Null dates not handled properly in ADO driver (it becomes 31 Dec 1969!).

Heinz Hombergs contributed patches for mysql MetaColumns - adding scale, made interbase MetaColumns work with firebird/interbase, and added lang/

Added INFORMIXSERVER environment variable.

Added $ADODB_ANSI_PADDING_OFF for interbase/firebird.

PHP 5 beta 2 compat check. Foreach (Iterator) support. Exceptions support.

4.01 23 Oct 2003

Fixed bug in rs2html(),, that generated blank table cells.

Fixed insert_id() incorrectly generated when logsql() enabled.

Modified PostgreSQL _fixblobs to use list/each instead of foreach.

Informix ErrorNo() implemented correctly.

Modified several places to use list/each, including GetRowAssoc().

Added UserTimeStamp() to connection class.

Added $ADODB_ANSI_PADDING_OFF for oci8po.

4.00 20 Oct 2003

Upgraded adodb-xmlschema to 1 Oct 2003 snapshot.

Fix to rs2html warning message. Thx to Filo.

Fix for odbc_mssql/mssql SQLDate(), hours was wrong.

Added MetaColumns and MetaPrimaryKeys for sybase. Thx to Chris Phillipson.

Added autoquoting to datadict for MySQL and PostgreSQL. Suggestion by Karsten Dambekalns

3.94 11 Oct 2003

Create trigger in did not work, because all cr/lf's must be removed.

ErrorMsg()/ErrorNo() did not work for many databases when logging enabled. Fixed.

Removed global variable $ADODB_LOGSQL as it does not work properly with multiple connections.

Added SQLDate support for sybase. Thx to Chris Phillipson

Postgresql checking of pgsql resultset resource was incorrect. Fix by Bharat Mediratta Same patch applied to _insertid and _affectedrows for

Added support for NConnect for postgresql.

Added Sybase data dict support. Thx to Chris Phillipson

Extensive improvements in $perf->UI(), eg. Explain now opens in new window, we show scripts which call sql, etc.

Perf Monitor UI works with magic quotes enabled.

rsPrefix was declared twice. Removed.

Oci8 stored procedure support, eg. "begin func(); end;" was incorrect in _query. Fixed.

Tiraboschi Massimiliano contributed italian language file.

Fernando Ortiz,, contributed informix performance monitor.

Added _varchar (varchar arrays) support for postgresql. Reported by PREVOT Stéphane.

3.92 22 Sept 2003

Added GetAssoc and CacheGetAssoc to connection object.

Removed TextMax and CharMax functions from

HasFailedTrans() returned false when trans failed. Fixed.

Moved perf driver classes into adodb/perf/*.php.

Misc improvements to performance monitoring, including UI().

RETVAL in mssql Parameter(), we do not append @ now.

Added Param($name) to connection class, returns '?' or ":$name", for defining bind parameters portably.

LogSQL traps affected_rows() and saves its value properly now. Also fixed oci8 _stmt and _affectedrows() bugs.

Session code timestamp check for oci8 works now. Formerly default NLS_DATE_FORMAT stripped off time portion. Thx to Tony Blair ( Also added new $conn->datetime field to oci8, controls whether MetaType() returns 'D' ($this->datetime==false) or 'T' ($this->datetime == true) for DATE type.

Fixed bugs in and

Fixed misc bugs in adodb_key_exists, GetInsertSQL() and GetUpdateSQL().

Tuned include_once handling to reduce file-system checking overhead.

3.91 9 Sept 2003

Only released to InterAkt

Added LogSQL() for sql logging and $ADODB_NEWCONNECTION to override factory for driver instantiation.

Added IfNull($field,$ifNull) function, thx to

Added portable substr support.

Now rs2html() has new parameter, $echo. Set to false to return $html instead of echoing it.

3.90 5 Sept 2003

First beta of performance monitoring released.

MySQL supports MetaTable() masking.

Fixed key_exists() bug in

Added sp_executesql Prepare() support to mssql.

Added bind support to db2.

Added swedish language file - Christian Tiberg"

Bug in drop index for mssql data dict fixed. Thx to Gert-Rainer Bitterlich.

Left join setting for oci8 was wrong. Thx to

3.80 27 Aug 2003

Patch for PHP 4.3.3 cached recordset csv2rs() fread loop incompatibility.

Added matching mask for MetaTables. Only for oci8, mssql and postgres currently.

Rewrite of "oracle" driver connection code, merging with "oci8", by Gaetano.

Added better debugging for Smart Transactions.

Postgres DBTimeStamp() was wrongly using TO_DATE. Changed to TO_TIMESTAMP.

ADODB_FETCH_CASE check pushed to ADONewConnection to allow people to define it after including

Added portugese (brazilian) to languages. Thx to "Levi Fukumori".

Removed arg3 parameter from Execute/SelectLimit/Cache* functions.

Execute() now accepts 2-d array as $inputarray. Also changed docs of fnExecute() to note change in sql query counting with 2-d arrays.

Added MONEY to MetaType in PostgreSQL.

Added more debugging output to CacheFlush().

3.72 9 Aug 2003

Added qmagic($str), which is a qstr($str) that auto-checks for magic quotes and does the right thing...

Fixed CacheFlush() bug - Thx to

Walt Boring contributed MetaForeignKeys for postgres7.

_fetch() called _BlobDecode() wrongly in interbase. Fixed.

adodb_time bug fixed with dates after 2038 fixed by Jason Pell.

3.71 4 Aug 2003

The oci8 driver, MetaPrimaryKeys() did not check the owner correctly when $owner == false.

Russian language file contributed by "Cyrill Malevanov"

Spanish language file contributed by "Horacio Degiorgi"

Error handling in oci8 bugfix - if there was an error in Execute(), then when calling ErrorNo() and/or ErrorMsg(), the 1st call would return the error, but the 2nd call would return no error.

Error handling in odbc bugfix. ODBC would always return the last error, even if it happened 5 queries ago. Now we reset the errormsg to '' and errorno to 0 everytime before CacheExecute() and Execute().

3.70 29 July 2003

Added new SQLite driver. Tested on PHP 4.3 and PHP 5.

Added limited "sapdb" driver support - mainly date support.

The oci8 driver did not identify NUMBER with no defined precision correctly.

Added ADODB_FORCE_NULLS, if set, then PHP nulls are converted to SQL nulls in GetInsertSQL/GetUpdateSQL.

DBDate() and DBTimeStamp() format for postgresql had problems. Fixed.

Added tableoptions to ChangeTableSQL(). Thx to Mike Benoit.

Added charset support to postgresql. Thx to Julian Tarkhanov.

Changed OS check for MS-Windows to prevent confusion with darWIN (MacOS)

Timestamp format for db2 was wrong. Changed to

adodb-cryptsession.php includes wrong. Fixed.

Added MetaForeignKeys(). Supported by mssql, odbc_mssql and oci8.

Fixed some oci8 MetaColumns/MetaPrimaryKeys bugs. Thx to Walt Boring.

adodb_getcount() did not init qryRecs to 0. Missing "WHERE" clause checking in GetUpdateSQL fixed. Thx to Sebastiaan van Stijn.

Added support for only 'VIEWS' and "TABLES" in MetaTables. From Walt Boring.

Upgraded to 0.0.2.

NConnect for mysql now returns value. Thx to Dennis Verspuij.

ADODB_FETCH_BOTH support added to interbase/firebird.

Czech language file contributed by Kamil Jakubovic

PostgreSQL BlobDecode did not use _connectionID properly. Thx to Juraj Chlebec.

Added some new initialization stuff for Informix. Thx to "Andrea Pinnisi"

ADODB_ASSOC_CASE constant wrong in sybase _fetch(). Fixed.

3.60 16 June 2003

We now SET CONCAT_NULL_YIELDS_NULL OFF for odbc_mssql driver to be compat with mssql driver.

The property $emptyDate missing from connection class. Also changed 1903 to constant (TIMESTAMP_FIRST_YEAR=100). Thx to Sebastiaan van Stijn.

ADOdb speedup optimization - we now return all arrays by reference.

Now DBDate() and DBTimeStamp() now accepts the string 'null' as a parameter. Suggested by vincent.

Added GetArray() to connection class.

Added not_null check in informix metacolumns().

Connection parameters for postgresql did not work correctly when port was defined.

DB2 is now a tested driver, making adodb 100% compatible. Extensive changes to odbc driver for DB2, including implementing serverinfo() and SQLDate(), switching to SQL_CUR_USE_ODBC as the cursor mode, and lastAffectedRows and SelectLimit() fixes.

The odbc driver's FetchField() field names did not obey ADODB_ASSOC_CASE. Fixed.

Some bugs in adodb_backtrace() fixed.

Added "INT IDENTITY" type to adorecordset::MetaType() to support odbc_mssql properly.

MetaColumns() for oci8, mssql, odbc revised to support scale. Also minor revisions to odbc MetaColumns() for vfp and db2 compat.

Added unsigned support to mysql datadict class. Thx to iamsure.

Infinite loop in mssql MoveNext() fixed when ADODB_FETCH_ASSOC used. Thx to Josh R,

ChangeTableSQL contributed by Florian Buzin.

The odbc_mssql driver now sets CONCAT_NULL_YIELDS_NULL OFF for compat with mssql driver.

0.10 Sept 9 2000 First release

Old changelog history moved to old-changelog.htm.