File: documentation-3.html

package info (click to toggle)
phplib 1%3A7.3dev-3.1
links: PTS
area: main
in suites: potato
size: 1,752 kB
ctags: 247
sloc: php: 6,659; perl: 323; pascal: 157; makefile: 102; sh: 7
file content (2347 lines) | stat: -rw-r--r-- 94,013 bytes
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2 Final//EN">
<HTML>
<HEAD>
 <META NAME="GENERATOR" CONTENT="SGML-Tools 1.0.9">
 <TITLE>PHP Base Library Documentation, Release phplib_7_2: Core Functionality</TITLE>
 <LINK HREF="documentation-4.html" REL=next>
 <LINK HREF="documentation-2.html" REL=previous>
 <LINK HREF="documentation.html#toc3" REL=contents>
</HEAD>
<BODY>
<A HREF="documentation-4.html">Next</A>
<A HREF="documentation-2.html">Previous</A>
<A HREF="documentation.html#toc3">Contents</A>
<HR>
<H2><A NAME="s3">3. Core Functionality</A></H2>

<P>Each class contains instance variables and instance methods. Some of these
variables and methods are available for customization, some are internal to
the classes themselves. All are documented, but tampering with internal
variables and methods is not supported. Internal interfaces are subject to
change without notice from one version of the library to another.
<P>
<P>This section covers PHPLIB core functionality in reference form.
Classes are presented in order of dependency, though, because
the core structure is easier understood in this order. You will
need to understand the complete core structure to successfully
use all of PHPLIB's features.
<P>
<H2><A NAME="ss3.1">3.1 DB_Sql</A>
</H2>

<P><CODE>DB_Sql</CODE> is used by <CODE>CT_Sql</CODE> and <CODE>Auth</CODE> to access a
SQL database. You are encouraged to use it directly, too.
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
Host </TD><TD>Host where your SQL server is running on.</TD></TR><TR><TD>
Database</TD><TD>Name of database or instance to use on that server.</TD></TR><TR><TD>
User </TD><TD>The username used in connections attempts.</TD></TR><TR><TD>
Password</TD><TD>The password used in connection attempts.</TD></TR><TR><TD>
Row </TD><TD>Number of the current result row starting at 0.</TD></TR><TR><TD>
Errno </TD><TD>Integer: error number of last database operation.</TD></TR><TR><TD>
Error </TD><TD>String: error message of last database operation.</TD></TR><TR><TD>
Halt_On_Error</TD><TD>One of "yes", "no", "report". If set to"yes" (the default), the database interface will report any errors and haltthe program. If set to "report", the database interface will still reportany errors, but continue reporting "false" back to the application, withErrno and Error set appropriately. If set to "no", the database interfacewill not report any errors, but silently report "false" back to application,with Errno and Error set appropriately.</TD></TR><TR><TD>
Auto_Free</TD><TD>Boolean: In some DB interfaces a flag for earlyresult memory release.</TD></TR><TR><TD>
Debug </TD><TD>Boolean: If set, the database class will output all queriesand additional diagnostic output.</TD></TR><TR><TD>
type </TD><TD>Contant string: The name of the database interface, e.g."mysql" or "oracle"</TD></TR><TR><TD>
revision</TD><TD>Contant version string: The version of the database API(e.g. 1.2), NOT the CVS revision of the file implementing the API.Sql_Table</TD><TD>string: The name of the table used by the<CODE>nextid()</CODE> API function.</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
Record </TD><TD>In some DB interfaces a hash of the current table result row.</TD></TR><TR><TD>
Link_ID </TD><TD>SQL Link ID.</TD></TR><TR><TD>
Query_ID</TD><TD>SQL Result ID.</TD></TR><TR><TD>

<CAPTION>Internal instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Instance methods</H3>

<P>
<P>
<H3>Accessible instance methods</H3>

<P>
<P>
<DL>
<DT><B>DB_Sql($query = "")</B><DD><P>Constructor. When creating an instance, you may optionally supply 
a query string.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
$db = new DB_Sql_Subclass("select * from mytable");)
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<DT><B>query($query_string)</B><DD><P><CODE>query_string</CODE> is a SQL statement that is sent to the database.
After sending the statement, <CODE>Error</CODE> and <CODE>Errno</CODE> are updated.
If the query is syntactically incorrect (no valid result id is
being produced), <CODE>halt()</CODE> is called with a meaningful error
message.
<P>If there is no active link to the database, a <CODE>pconnect()</CODE> is
made using the information from the <CODE>Host</CODE>, <CODE>Database</CODE>,
<CODE>User</CODE> and <CODE>Password</CODE> instance variables.
<P>Returns the result of the <CODE>query()</CODE> statement, which is
guaranteed to be a valid result id (or false, if
Halt_On_Error isn't "yes").
<P>
<DT><B>next_record()</B><DD><P><CODE>next_record()</CODE> advances the cursor through the
current query result and updates the <CODE>Record</CODE>, <CODE>Row</CODE>,
<CODE>Errno</CODE> and <CODE>Error</CODE> instance variables.
<P>Returns true, if there is a new result record. Returns false, if
done with the current result set.  If <CODE>Auto_Free</CODE> is true,
<CODE>free_result()</CODE> is called automatically before false
is returned.
<P>
<DT><B>num_rows(), nf()</B><DD><P>Returns the number of rows returned by the current SELECT query.
<P><EM>Note:</EM> This information is not available in all database
interfaces. Some of the more advanced databases begin to return
query results asynchronously while the backend is still
appending result rows. In such environments the complete size of
the result set is never known. 
<P>You should duplicate your WHERE clause of the query in such
environments and ask for the COUNT(*). This will be less
inefficient as it seems as the query path and query result have
been cached by the database.
<P>
<DT><B>affected_rows()</B><DD><P>Returns the number of rows affected by the current INSERT,
UPDATE or DELETE query.
<P>
<DT><B>num_fields()</B><DD><P>Returns the number of columns returned by the current query.
<P>
<DT><B>np()</B><DD><P>Prints the number of rows returned by the current query.
<P>
<DT><B>f($field)</B><DD><P>Identical to accessing <CODE>Record[$field]</CODE>.
<P>
<DT><B>p($field)</B><DD><P>Identical to printing Record[$field].
<P>
<DT><B>haltmsg($msg)</B><DD><P>This function is called by <CODE>halt()</CODE> and will actually print
the database error message. You may override this method in your
subclass of <CODE>DB_Sql</CODE> and format the error message to
be consistent with the layout of the rest of your application.
You may also add additional error handling such as informing the
application operator by mail that a database error has occured.
<P>
<DT><B>seek($pos)</B><DD><P>Positions the <CODE>Row</CODE> pointer within the result set. Useful for
reading the same result set twice or otherwise jumping around
within the result. <CODE>$pos</CODE> is not checked in any way for
validity.
<P><EM>Note:</EM> If <CODE>Auto_Free</CODE> is true, <CODE>seek()</CODE> may not be
useable, because the result set has already been free'ed when
<CODE>next_record()</CODE> when behind the last record of the result
set.
<P><EM>Note:</EM> Not all database interfaces provide a cursor that is
capable of seeking. This function will be unavailable in such
environments.
<P>
<DT><B>link_id()</B><DD><P>This function will return the current link ID, as returned by
the <CODE>pconnect()</CODE> executed internally by the database class.
<P>You should not need this information.
<P>
<DT><B>query_id()</B><DD><P>This function will return the current result ID, as returned by
the <CODE>query()</CODE> executed internally by the database class.
<P>You should not need this information.
<P>
<DT><B>metadata($table = "", $full = false)</B><DD><P><CODE>$table</CODE> is a SQL table name in the current database. The
function returns an array of hashes indexed on the (0 based)
column number of <CODE>$table</CODE>. Each hash is indexed by <CODE>table</CODE>
(table of which this column is part of), <CODE>name</CODE> (name of this
column), <CODE>type</CODE> (column data type), <CODE>len</CODE> (column width)
and <CODE>flags</CODE> (database specific column flags, if applicable)
with one row per table column. Each row describes a column in
your table.
<P>The data returned by <CODE>metadata()</CODE> is suitable for passing
it to the Table class. If you specify the <CODE>full</CODE> parameter,
an additional column <CODE>meta</CODE> is added, which is indexed by
field name and returns the field number of that name. Also, a
column <CODE>num_fields</CODE> is added, containing the width
of the table.
<P>If <CODE>$table</CODE> is omitted, the function returns metadata on the result
of the last executed query.  <EM>Note:</EM> This is currently implemented only
for the MySQL interface.  You are encouraged to implement this feature
for other interfaces.
<P><EM>NOTE:</EM> At the moment, the PostgreSQL and ODBC interface only report
the <CODE>table</CODE>, <CODE>name</CODE> and <CODE>type</CODE> data reliably. You are
encouraged to fix this.
<P>
<DT><B>table_names()</B><DD><P>Returns an array with table name and tablespace name.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
table name      : $return[$i]["table_name"]
tablespace_name : $return[$i]["tablespace_name"]
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Tables are from $i=0 to last table;
<P>Implemented in db_oracle.inc,db_oci8.inc,db_mysql.inc,db_pgsql.inc
<P>
<DT><B>nextid($sequence_name)</B><DD><P>This function will return a sequence number from the sequence
named by <CODE>$sequence_name</CODE>. This number is guaranteed
to be obtained in an atomic manner and can be used as a primary
key.
</DL>
<P>
<H3>Internal instance methods</H3>

<P>
<P>
<DL>
<DT><B>connect()</B><DD><P>Used internally to generate a <CODE>Link_ID</CODE>, if necessary. Link
creation is implicit, there is no need to call <CODE>connect()</CODE>
manually, ever.
<P>
<DT><B>halt($msg)</B><DD><P>Used by <CODE>query()</CODE> if the initial database connection cannot
be made or the target database does not exist. Depending on
the setting of <CODE>Halt_On_Error</CODE>, this method
will call <CODE>haltmsg()</CODE> to report the error.
<P>
<DT><B>free()</B><DD><P>Used internally by <CODE>next_record()</CODE> to free the result
set, if so configured.
</DL>
<P>
<H3>Example</H3>

<P>Use a subclass to provide the appropriate parameters for a
database connect. You may overwrite <CODE>halt()</CODE> to customize the
error message, although a sensible default is provided.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class DB_Article extends DB_Sql {
  var $classname = "DB_Article";

  var $Host     = "sales.doma.in";
  var $Database = "shop_project";
  var $User     = "webuser";
  var $Password = "";

  function haltmsg($msg) {
    printf("&lt;/td&gt;&lt;/table&gt;&lt;b&gt;Database error:&lt;/b&gt; %s&lt;br&gt;\n", $msg);
    printf("&lt;b&gt;MySQL Error&lt;/b&gt;: %s (%s)&lt;br&gt;\n",
      $this-&gt;Errno, $this-&gt;Error);
    printf("Please contact shopmaster@doma.in and report the ");
    printf("exact error message.&lt;br&gt;\n");
  }
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Use an instance of the subclass to manage your queries:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
$q = new DB_Article;

$query = sprintf("select * from articles where article like '%%%s%%'",
              $searchword);
$q-&gt;query($query);

while($q-&gt;next_record()) {
  printf("&lt;tr&gt;&lt;td&gt;%s&lt;/td&gt;&lt;td&gt;%s&lt;/td&gt;&lt;/tr&gt;\n",
    $q-&gt;f("art_id"),
    $q-&gt;f("article"));
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<H3>Additional information about database connections</H3>

<P>
<P>PHP reuses connections, if possible. When a connection is being
made to the same Host with the same Username and Password as an
existing connection, no second connection is being made by PHP.
Instead the existing connection is returned to the caller. This
is true for both, the *_connect() and *_pconnect() calls of all
PHP database interfaces.
<P>This has implications for MySQL users: Never use the MySQL "use"
command to change the current database. If you do, session
management will fail to operate properly. Instead, create
all PHPLIB tables as part of your application.
<P>Some databases (for example Oracle) have very expensive
connect() operations. For these databases, performance is
dramatically improved if you switch from CGI PHP to mod_php.
This is, because PHPLIB uses the "*_pconnect()" method to
connect to your database. In mod_php, the database connection is
kept around by the web server process after the page has been
processed and is reused if a further connect requires a
connection with the same Host/Username/Password pattern.
<P>This means that there will be at most "number of web server
processes" times "number of Host/Username/Password-combinations"
many simultaneous connections to your database server. Keep that
in mind when planning licenses and server load. Using CGI PHP
will probably reduce the number of concurrent connects to your
database server at the expense of connection setup time. For
database servers where connection setup time is negligible
(MySQL for example) this is a viable solution (don't try it with
Oracle) though.
<P>
<H3>Using <CODE>nextid()</CODE></H3>

<P>
<P>The <CODE>nextid()</CODE> function can be used to obtain a sequence
number which can be used as a primary key. The function manages
an arbitrary number of named sequences, you have to provide the
name of a sequence upon call.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
$db = new DB_Article;

$artnr = $db-&gt;nextid("article_sequence");
$query = sprintf("insert into articles ( artnr, ...) values ('%s', ...)", 
   $artnr, ...);
$db-&gt;query($query);

reset($articles);
while(list($itemnr, $itemdesc) = each($articles)) {
  $itemnr = $db-&gt;nextid("item_sequence");
  $query = sprintf("insert into items (artnr, itemnr, ...) values ('%s', '%s', ...)",
    $artnr, $itemnr, ...);
  $db-&gt;query($query);
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>

<P>
<H2><A NAME="ss3.2">3.2 Page Management</A>
</H2>

<P>
<P>
<H3>Accessible Functions</H3>

<P>
<P>Page Management currently consists a collection of functions:
<P>
<DL>
<DT><B>page_open(array("feature" => "classname"))</B><DD><P>This function is to be called with an array of page
features/classname pairs. Valid features are at the moment:
<P>
<DL>
<DT><B>sess</B><DD><P>This page makes use of session variables.
<P>
<DT><B>auth</B><DD><P>This page uses session authentication. If you
specify the <CODE>auth</CODE> feature, you MUST specify
the <CODE>sess</CODE> feature, also.
<P>
<DT><B>perm</B><DD><P>This page is protected by permissions and only
accessible to authenticated users with matching rights.
If you specify the <CODE>perm</CODE> feature, you MUST specify
the <CODE>auth</CODE> and <CODE>sess</CODE> features, also.
<P>
<DT><B>user</B><DD><P>This page makes use of user variables. If you specify
the <CODE>user</CODE> feature, you MUST specify the <CODE>auth</CODE> and
<CODE>sess</CODE> features, also.
</DL>
<P>Each feature specifies the name of the class that implements that feature,
for example
<P>
<HR>
<PRE>
  page_open(array("sess" =&gt; "Shop_Session"));
</PRE>
<HR>
<P>The function creates an instance of <CODE>Shop_Session</CODE> as
<CODE>$sess</CODE> and initializes it. It also checks feature
dependencies. Note that you are expected to provide an
implementation of the class <CODE>Shop_Session</CODE>. This is
usually done in <CODE>local.inc</CODE> and usually you do so by
extending the provided <CODE>Session</CODE> class.
<P>Examples on how to do this is given in the documentation below
when the classes are introduced.
<P>
<DT><B>page_close()</B><DD><P>
<P>At the end of your page (after all results have been calculated)
you have to call <CODE>page_close()</CODE>. This will save all
page state, session and user variables into database. Changes to
session or user variables after <CODE>page_close()</CODE> has
been called are not recorded. Currently it is allowed to call
<CODE>page_close()</CODE> multiple times on a single page (not
guaranteed for future versions!). Each time session state will be
saved.
<P><EM>Note:</EM> This is going to change. When we introduce record
locking, it is important that you call <CODE>page_close()</CODE> only
once per page, because that will implicitly unlock your session
record. Also, it is important that you call <CODE>page_close()</CODE> as
early as possible on a page so that the locking time is kept
minimal.
<P>
<DT><B>sess_load(array("var" => "classname")</B><DD><P> 
<P><EM>Advanced feature</EM>. Some applications have need to manually
load data belonging to one or multiple session classes. @@TODO
<P>
<DT><B>sess_save(array("var" => "classname"))</B><DD><P><EM>Advanced feature</EM>. @@TODO
<P>
</DL>
<P>
<H3>Example</H3>

<P>
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
&lt;?php
  page_open(array("sess" =&gt; "Shop_Session"));
  $sess-&gt;register("s");  // See "Session" below for explanation.
 ?&gt;
&lt;html&gt;
&lt;h1&gt;&lt;?php print ++$s ?&gt;&lt;/h1&gt;
&lt;/html&gt;
&lt;?php page_close(); ?&gt;
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<H3>The "cart" feature is gone</H3>

<P> 
There used to be a feature "cart" for <CODE>page_open()</CODE> in
versions of PHPLIB up to release-5. The cart has been removed
from the core functionality of PHPLIB to keep the library small,
maintainable and structured. Consequently the "cart" feature is
gone.
<P>The <CODE>Cart</CODE> class is still present and exists as an extended
feature. You have to include and instantiate your cart manually
on that pages that use it, though. See the <CODE>Cart</CODE> class for
more information.
<P>
<P>
<H2><A NAME="ss3.3">3.3 CT_Sql</A>
</H2>

<P>
<P>The <CODE>Session</CODE> class used to contain a bit of SQL to read
and write session data from and to a database. To make sessions
database independent, this SQL has been isolated and put in
a separate class, <CODE>CT_Sql</CODE>. <CODE>Session</CODE> now makes
all storage accesses through a container class, which may or
may not be an SQL container.
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
database_table</TD><TD>The name of the database table which should be used</TD></TR><TR><TD>
database_class</TD><TD>A classname. CT_Sql uses this class to store and retrieve data</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Example</H3>

<P>
<P>Use a subclass to provide the appropriate parameters 
to your container. Usually your subclass looks like this: 
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Sql extends CT_Sql {
        var $classname = "My_Sql";
        var $database_table = "active_sessions";
        var $database_class = "DB_Session";
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>You can then use My_Sql in class Session. Reference it
by putting "My_Sql" in the "that_class" variable.
<P>
<P>
<H2><A NAME="ss3.4">3.4 CT_Split_Sql</A>
</H2>

<P>
<P>The <CODE>Session</CODE> class used to contain a bit of SQL to read and
write session data from and to a database. To make sessions
database independent, <CODE>Session</CODE> now makes all storage
accesses through a container class. The
<CODE>CT_split_sql</CODE> container is very similar to
<CODE>CT_Sql</CODE> container, with the difference that if
serialized data exceeds a specified amount of bytes, multiple
rows will be used to memorized the entire field.
<P>This class is NOT compatible with <CODE>CT_Sql</CODE> class, since
table layout is different and column names are different in order to
avoid reserved words in various database implementation. This uses a
<CODE>DB_Sql</CODE> like class so you can access all supported databases
with this container.
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
database_table</TD><TD>The name of the database table which should be used</TD></TR><TR><TD>
database_class</TD><TD>A classname. CT_Sql uses this class to store and retrieve data</TD></TR><TR><TD>
split_length</TD><TD>A number. This specifies the maximum amount of bytessaved in each row of the table.

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Example</H3>

<P>
<P>Use a subclass to provide the appropriate parameters 
to your container. Usually your subclass looks like this: 
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Sql extends CT_Split_Sql {
        var $classname = "My_Sql";
        var $database_table = "active_sessions_split";
        var $database_class = "DB_Session";
        var $split_length = 4096;
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>You can then use My_Sql in class Session. Reference it
by putting "My_Sql" in the "that_class" variable.
<P>
<P>
<H2><A NAME="ss3.5">3.5 CT_Shm</A>
</H2>

<P>
<P>The <CODE>Session</CODE> class used to contain a bit of SQL to read and
write session data from and to a database. To make sessions
database independent, <CODE>Session</CODE> now makes all storage
accesses through a container class. To let <CODE>Session</CODE> use shared
memory as container, you use <CODE>CT_Shm</CODE>.
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
max_sessions</TD><TD>The maximum number of concurrent sessions supported by this container.</TD></TR><TR><TD>
shm_key</TD><TD>The unique (important!) key of the shared memory&nbsp;segment you want to use.</TD></TR><TR><TD>
shm_size</TD><TD>The size of the shared memory segment. The size is&nbsp;set, when the segment is accessed for the first time. If you do not use &nbsp;too many session variables, the formula shm_size = max_sessions * 600&nbsp;should be sufficient.</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Example</H3>

<P>
<P>Use a subclass to provide the appropriate parameters 
to your container. Usually your subclass looks like this: 
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Shm extends CT_Shm {
        var $classname = "My_Shm";
        var $max_sessions = 500;
        var $shm_key = 0x1234232;
        var $shm_size = 64000;
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>You can then use My_Shm in class Session. Reference it
by putting "My_Shm" in the "that_class" variable.
<P>
<P>
<H2><A NAME="ss3.6">3.6 CT_Dbm</A>
</H2>

<P>
<P>The <CODE>Session</CODE> class used to contain a bit of SQL to read and
write session data from and to a database. To make sessions
database independent, <CODE>Session</CODE> now makes all storage
accesses through a container class. To let <CODE>Session</CODE> use a DBM
database file as a container, you use <CODE>CT_Dbm</CODE>.
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
dbm_file</TD><TD>The path to the dbm file (must exist already AND must be writable by the server process)</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Example</H3>

<P>
<P>Use a subclass to provide the appropriate parameters 
to your container. Usually your subclass looks like this: 
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Dbm extends CT_Dbm {
        var $dbm_file = "data/session.dbm";
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>You can then use My_Dbm in class Session. Reference it
by putting "My_Dbm" in the "that_class" variable.
<P>
<P>
<H2><A NAME="ss3.7">3.7 CT_Ldap</A>
</H2>

<P>
<P>The <CODE>Session</CODE> class used to contain a bit of SQL to read and
write session data from and to a database. To make sessions
database independent, <CODE>Session</CODE> now makes all storage
accesses through a container class. To let <CODE>Session</CODE> use a
LDAP database as a container, you use <CODE>CT_Ldap</CODE>.
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
ldap_host</TD><TD>This is the hostname of the LDAP server to connect to</TD></TR><TR><TD>
ldap_port</TD><TD>And this is its port (LDAP default is 389)</TD></TR><TR><TD>
basedn</TD><TD>This is the basedn</TD></TR><TR><TD>
rootdn</TD><TD>This is the rootdn which is required to modify the database</TD></TR><TR><TD>
rootpw</TD><TD>The respective password for rootdn</TD></TR><TR><TD>
objclass</TD><TD>The objectclass for PHPLIB's data</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Example</H3>

<P>
<P>Use a subclass to provide the appropriate parameters 
to your container. Usually your subclass looks like this: 
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Ldap extends CT_Ldap {
        var $classname = "My_Ldap";
        var $ldap_host = "localhost";
        var $ldap_port = 389;
        var $basedn = "dc=your-domain, dc=com";
        var $rootdn = "cn=root, dc=your-domain, dc=com";
        var $rootpw = "secret";
        var $objclass = "phplibdata";
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>You can then use My_Ldap in class Session. Reference it
by putting "My_Ldap" in the "that_class" variable.
<P>
<P>
<H2><A NAME="ss3.8">3.8 Session</A>
</H2>

<P>
<P>The session class keeps a list of global variable names and
provides a set of functions to load and save these variables
from and to a data storage container (we will call it container
for shortness). The named variables may be scalar
variables (strings, integers and floats) or arrays. Objects are
handled as well, provided they implement two instance variables
naming their class and enumerating their (persistent) slots.
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
classname</TD><TD>Serialization helper: The name of this class.</TD></TR><TR><TD>
magic</TD><TD>A secret string used in ID creation. Change it!</TD></TR><TR><TD>
mode </TD><TD>Mode of Session ID propagation. Either <CODE>cookie</CODE> or <CODE>get</CODE>.</TD></TR><TR><TD>
fallback_mode</TD><TD>Mode of Session ID propagation should <CODE>$mode</CODE> not work. Set <CODE>$mode</CODE> to <CODE>cookie</CODE> and <CODE>$fallback_mode</CODE> to <CODE>get</CODE>.</TD></TR><TR><TD>
lifetime </TD><TD>Lifetime of the session cookie in minutes or 0 to use session cookies.</TD></TR><TR><TD>
gc_time</TD><TD>Garbage collection tuning parameter, see below.</TD></TR><TR><TD>
gc_probability</TD><TD>Garbage collection tuning parameter, see below.</TD></TR><TR><TD>
allowcache</TD><TD>Control caching of session pages, if set to <CODE>no</CODE> (also the default), the page is not cached under HTTP/1.1 or HTTP/1.0; if set to <CODE>public</CODE> , the page is publically cached under HTTP/1.1 and HTTP/1.0; if set to <CODE>private</CODE> , the page is privately cached under HTTP/1.1 and not cached under HTTP/1.0</TD></TR><TR><TD>
allowcache_expires</TD><TD>When caching is allowed, the pages can be cached for this many minutes.</TD></TR><TR><TD>
that_class</TD><TD>A classname. Session uses this class to store and retrieve data.</TD></TR><TR><TD>
auto_init</TD><TD>The file to be loaded on session establishment.</TD></TR><TR><TD>
secure_auto_init</TD><TD>Set to 0, if all pages always callpage_close() (This is never the case!).</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
pt</TD><TD>Internal array of names of persistent variables.</TD></TR><TR><TD>
in</TD><TD>Flag: If set, auto_init has been executed.</TD></TR><TR><TD>
name</TD><TD>A tag (name) for the session type.</TD></TR><TR><TD>
id</TD><TD>Id of the current session.</TD></TR><TR><TD>
that</TD><TD>Container object instance.</TD></TR><TR><TD>

<CAPTION>Internal instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Instance methods</H3>

<P>
<P>
<H3>Accessible instance methods</H3>

<P>
<P>
<DL>
<DT><B>register($varname)</B><DD><P>Registers a global variable name as a session variable. The
name may identify a scalar variable, an array or an object.
If an object is to be made persistent, it must have two
instance variables:
<P>
<DL>
<DT><B>classname</B><DD><P>A string with the name of the objects class.
<DT><B> persistent_slots</B><DD><P>An array with the names of all object slots to save.
</DL>
<P>
<DT><B>unregister($varname)</B><DD><P>Unregisters a global variable name as a session variable.
The variable is not deleted, but its value will be lost
at the end of a page. It is no longer saved to the database.
<P>
<DT><B>is_registered($varname)</B><DD><P>Returns true if the variable named $varname is registered
with the session, false otherwise.
<P>
<DT><B>delete()</B><DD><P>Destroy the current session and put_id() the current session
id.
<P>After <CODE>delete()</CODE> has been executed, all session data has
been removed from the database. Also, the session object is
unusable on this page. Consequently, <CODE>page_close()</CODE> may
not be called for this session. Session variables are still
available on this page, even after the <CODE>delete()</CODE>, but
will be lost on the following pages.
<P>In cookie mode, it is possible to <CODE>page_open()</CODE> a new
session after <CODE>delete()</CODE> has been called, if no HTML has
been output so far so that the new cookie can be set. If you
do this, you can also re-register some of the previous
session variables and can call <CODE>page_close()</CODE> for the new
session.  This allows you to change the session on the fly
and selectively carry over session data from the previous
session.
<P>
<DT><B>url($url)</B><DD><P>Return an URL referencing the current session. If in <CODE>get</CODE>
mode, the current session id is attached to this URL, else the
URL is returned unmodified.
<P>
<DT><B>purl($url)</B><DD><P>A shorthand for <CODE>print $this-&gt;url($url);</CODE>
<P>
<DT><B>self_url()</B><DD><P>Return an URL referencing the current page, including
<CODE>PHP_SELF</CODE> and <CODE>QUERY_STRING</CODE> information.
If in <CODE>get</CODE> mode, the session id is included.
<P>
<DT><B>pself_url()</B><DD><P>A shorthand for <CODE>print $this-&gt;self_url()</CODE>.
<P>
<DT><B>hidden_session()</B><DD><P>Prints a hidden form element containing the session name and id.
This function will print the form element unconditionally, which
is useful to propagate a session id from one web server to
another, if both share the same stable storage for session data.
This is useful to share a session id between a HTTPS and HTTP
server or between several servers in different domains.
<P>
<P>
<DT><B>get_hidden_session()</B><DD><P>Returns a hidden element containing the session name and id.
This function will return the form element unconditionally,
which is useful to propagate a session id from one web server to
another, if both share the same stable storage for session data.
This is useful to share a session id between a HTTPS and HTTP
server or between several servers in different domains.
<P>
<DT><B>hidden_id()</B><DD><P>Adds a hidden form element containing the session name and id,
if currently in get-mode. Else the function prints nothing.
This function can be used to generate a form element which will
propagate the session id in a POST form.
<P>
<DT><B>get_hidden_session()</B><DD><P>Returns a hidden element containing the session name and id, if
currently in get-mode. Else the functions returns the empty
string. This function can be used to get a string with a form
element which can propagate the session id in a POST form.
<P>
<P>
<DT><B>add_query($qarray)</B><DD><P>
<P>Return string to be appended to the current URL for parameters
in GET query format. Intended usage is like this:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>

&lt;a href="&lt;&lt;?
$sess->pself_url().$sess-&gt;padd_query(array("again"=&gt;"yes"))
?&gt;"&gt; Reload&lt;/a&gt; and log in?
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<DT><B>padd_query($qarray)</B><DD><P>
<P>A shorthand for <CODE>print $this-&gt; add_query($qarray)</CODE>.
<P>
<DT><B>reimport_get_vars()</B><DD><P>
<P>When a <CODE>FORM</CODE> variable is made persistent, that form variable is
imported into PHP, then page_open() is being called and
the new variable value is overwritten from the database. The
<CODE>FORM</CODE> value is lost.
<P>
<P>If you had enabled <CODE>track_vars</CODE> and were accessing
<CODE>HTTP_GET_VARS</CODE> directly, which is recommended,
this were not a problem. Some legacy scripts rely on persistent
<CODE>FORM</CODE> input variables, though.
<P>
<P>These scripts may call the appropriate
<CODE>reimport</CODE>_x_<CODE>vars()</CODE> functions. These
functions will re-read the tracked variable arrays and
reinitialize the appropriate global variables after session
variables have been restored.
<P>
<P>Use of this function is discouraged.
<P>
<DT><B>reimport_post_vars()</B><DD><P>See <CODE>reimport_get_vars()</CODE>.
<P>
<DT><B>reimport_cookie_vars()</B><DD><P>See <CODE>reimport_get_vars()</CODE>.
<P>
<DT><B>set_container()</B><DD><P>You shall not call this function directly. It is called back by the
<CODE>start()</CODE> function of <CODE>Session()</CODE> during initializiation.
It is documented so that you can override its implementation in
your subclass of <CODE>Session</CODE> if you know what you are doing.
<P>This function creates and starts the container class used by
this instance of session.
<P>
<DT><B>set_tokenname()</B><DD><P>You shall not call this function directly. It is called back by the
<CODE>start()</CODE> function of <CODE>Session()</CODE> during initializiation.
It is documented so that you can override its implementation in
your subclass of <CODE>Session</CODE> if you know what you are doing.
<P>This function determines and sets the internal session name.
<P>
<DT><B>release_token()</B><DD><P>You shall not call this function directly. It is called back by the
<CODE>start()</CODE> function of <CODE>Session()</CODE> during initializiation.
It is documented so that you can override its implementation in
your subclass of <CODE>Session</CODE> if you know what you are doing.
<P>This function determines the current method of session
propagation and determines if a new session token has to be
generated.
<P>
<DT><B>put_headers()</B><DD><P>You shall not call this function directly. It is called back by the
<CODE>start()</CODE> function of <CODE>Session()</CODE> during initializiation.
It is documented so that you can override its implementation in
your subclass of <CODE>Session</CODE> if you know what you are doing.
<P>This function determines which header lines are to be generated
by the session, including cache control headers.
</DL>
<P>
<H3>Internal instance methods</H3>

<P>
<P>
<DL>
<DT><B>get_id()</B><DD><P>See <CODE>get_id()</CODE>.
<P>
<DT><B>get_id($id_to_use)</B><DD><P>get_id() is used internally to determine a session
identifier.  Currently, a session identifier is a hex number
of 32 characters (128 bits) and it is generated by
md5(uniqid($this->magic)) to make it hard to guess.
<P>get_id() may be called with an optional session id to use as
a parameter. This is useful if you want to change a session
id without breaking the session (taking over an old, left
over session).
<P>get_id() can be overwritten by a subclass, if you want a
different system to create session ids. For example, some
applications want to use a constant session id that is not
propagated to the client to use a shared pool of persistent
variables (a guestbook for example). These applications need
locking (to be implemented soon).
<P>
<DT><B>put_id()</B><DD><P>put_id() is used internally to "unuse" a session it. At the
moment it deletes the client side cookie and deletes
$HTTP_COOKIE_VAR[$this->name] for that cookie. The variable
${$this->name} is <EM>not</EM> deleted.
<P>
<DT><B>serialize($prefix, &amp;$str)</B><DD><P>serialize() is used internally to append to str all PHP
code needed to reconstruct the variable named in prefix.
<P>
<DT><B>freeze()</B><DD><P>freeze() serializes all register()ed variables and writes
the resulting code into the database, tagged with the
current session id and the current session name.
<P>
<DT><B>thaw()</B><DD><P>thaw() loads a set of freeze()ed variables for the current
session id and session name out of the database and
recreates them.
<P>
<DT><B>gc()</B><DD><P>The <CODE>active_sessions</CODE> table contains one row for
each session. That row is uniquely identified by the <CODE>sid</CODE>
and <CODE>name</CODE> values (<CODE>name</CODE> is the name of the session
class that has written the row). Each time that row is written,
the column <CODE>changed</CODE> is updated with the current time.
<P>The gc() function deletes all rows that are older than
<CODE>gc_time</CODE> minutes and have a matching <CODE>name</CODE> 
field. For speed reasons, gc() is not not called every time
an update to <CODE>active_sessions</CODE> is being made.
Instead it is called randomly with a probability of
<CODE>gc_probability</CODE>.
<P>
<DT><B>reimport_any_vars($arrayname)</B><DD><P>Used to implement the three official reimport functions.
<P>
<DT><B>start()</B><DD><P>Initialization function, to be called after object
instantiation. Calls get_id() to get the current session id,
creates a database connection, then calls thaw() to load all
session variables. Randomly activates gc(). Checks <CODE>allowcache</CODE>
to send proper headers to control browser caching.
<P>
</DL>
<P>
<P>
<H3>Example</H3>

<P>Use a subclass to provide the appropriate parameters to your
session. Usually your subclass looks like this:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Session extends Session {
  var $classname = "My_Session"; ## Persistence support
  
  var $mode      = "cookie";
  var $lifetime  = 0;            ## use session cookies
  
  ## which container to use
  var $that_class = "Session_sql";
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Remember that you have to provide a <CODE>DB_Sql</CODE> subclass
with the parameters needed to access your database.
<P>Use the page management functions (see above) to use your
session subclass. The feature name for session management is
<CODE>sess</CODE>; provide the name of your session subclass as a
parameter to the sess feature:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
  page_open(array("sess" =&gt; "My_Session"));
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Use the <CODE>register()</CODE> instance method to register variables as
persistent. If <CODE>$sess</CODE> is your session object, use
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
$sess-&gt;register("s");
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>to make the global variable <CODE>$s</CODE> persistent. <CODE>$s</CODE> may be a
scalar value, an array or an object with persistence support
slots.
<P>Do not use the instance methods <CODE>freeze()</CODE> and <CODE>thaw()</CODE>
directly, but use the page management functions instead.
<P>To have some pages cached and others not cached, use multiple
instances of the session object. For example, for those pages
that should be cached, use a session object instance like
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Cached_Session extends My_Session {
  ## pages that use this session instance are cached.
  var $allowcache = "private";
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Be careful when using the <CODE>public</CODE> cache option. Publically cached pages
may be accessible to unauthenticated users. The <CODE>private</CODE> cache option
prevents unauthenticated access, but is only functional in HTTP/1.1 browsers.
<P>
<H3>Using "auto_init"</H3>

<P>
<P>You may define <CODE>$sess->auto_init</CODE> to the name of an include
file in your extension of session. Per convention, the name
<CODE>setup.inc</CODE> is being used.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Session extends Session {
  var $classname = "My_Session";
  var $magic     = "Calvin+Hobbes";
  var $mode      = "cookie";
  var $gc_probability = 5;

  var $auto_init = "setup.inc";   // name of auto_init file.
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Whenever a new session is established, that is, a user without a
session id connects to your application, the auto_init file is
included and executed exactly once. The file is executed from
within the context of the <CODE>page_open()</CODE> function, that is,
<EM>not</EM> within a global context. To define or access global
variables from the auto_init file, you have to <CODE>global</CODE> them.
<P>When auto_init is being executed, all features of your page
already exist and are available globally.
That is, you can safely rely on
the existence of the <CODE>$sess</CODE>, <CODE>$auth</CODE>, <CODE>$perm</CODE> and
<CODE>$user</CODE> variables, if your application specifies them.
<EM>Note</EM> that you cannot in general know which particular page
triggered the execution of auto_init, though. If you have some
pages that request authentication and others that don't, you
cannot rely on the presence of the <CODE>$auth</CODE> object in general,
but have to test for it with <CODE>is_object($auth)</CODE> before
accessing it.
<P>The auto_init file is the appropriate place to initialize and
register all your session variables. A sample <CODE>setup.inc</CODE> may
look like this:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
&lt;?php
global $lang;   // application language
$lang = "de";   // german by default
$sess-&gt;register("lang");

global $cur;   // application currency
$cur = "EUR";   // Euro by default
$sess-&gt;register("cur");

global $cart;
$cart = new Shop_Cart;      // Create a shopping cart object as defined in local.inc
$sess-&gt;register("cart"); // register it.
?&gt;
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P><EM>Note:</EM> If you don't use a fallback_mode and you get users
that turn off cookies, these users will force a new session each
time they hit any page of your application. Of course this will
force inclusion and execution of <CODE>setup.inc</CODE> for each page
they visit, too. Nothing can be done about this.
<P>
<H3>Unregistering variables and deleting sessions</H3>

<P>To get rid of a persistent variable, call
<CODE>$sess-&gt;unregister()</CODE> with the name of that variable. The
value of the formerly registered variable is still available
after the call to unregister, but the variable is no longer
persistent and will be lost at the end of the current page.
<P>To get rid of all session related data including the session
record in the database, the current session id and the session
cookie in the users browser, call <CODE>$sess-&gt;delete()</CODE>. In
shopping applications this is commonly done when the user
commits his order to get rid of the current shopping cart and
everything else. You may want to remember selected information
about that user, though, as shown below.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
&lt;?php
  page_open(array("sess" =&gt; "Shop_Session"));

  // send order as mail
  mail_order($shopowner, $user, $cart);

  // delete the current session
  $sess->delete();

  // now get a new session id, but retain the users
  // address and name:
  page_open(array("sess" =&gt; "Shop_Session")); // will force auto_init again!
  $sess->register("user");  // could be done in auto_init as well

?&gt;
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<H3>Reading and understanding session data for debugging</H3>

<P>When debugging PHPLIB applications, it is often useful to be
able to read and understand the contents of the active_sessions
table. Each session is represented by a single line in this
table. The primary key to this table is the pair <CODE>name</CODE> and
<CODE>sid</CODE>. <CODE>name</CODE> is the content of <CODE>$this-&gt;name</CODE> and
is usually the classname of your session class. <CODE>sid</CODE> is the
content of <CODE>$this-&gt;id</CODE> and is usually the MD5 hash of a
uniqid and some magic string.
<P>By choosing a pair, it is possible for PHPLIB to have more than
one session type (for example, session and user data, see the
<CODE>User</CODE> class below) per application and store all this data
in a single table. If you are debugging a session class, for
example <CODE>Example_Session</CODE>, only records where <CODE>name =
"Example_Session"</CODE> are of interest to you. Determine the current
session id of your <CODE>Example_Session</CODE> by printing <CODE>$sess->id</CODE>
and select the record with that <CODE>name</CODE> and <CODE>sid</CODE> from the
database.
<P>The <CODE>changed</CODE> field indicates when this record has been
updated the last time. It is a 14 character (Y2K compliant)
string of the format YYYYMMDDhhmmss. Ordering by <CODE>changed</CODE>
desc will show you the most current session records first (the
MySQL "limit" clause may come in handy here).
<P>The <CODE>val</CODE> column of a session record contains a PHP program
that can be safely fed to <CODE>stripslashes()</CODE> first and
<CODE>eval()</CODE> after that. The PHP program consists entirely of
assignments and contains all instructions necessary to recreate
the persistent variables. The structure and order of
instructions within this program is always the same.
<P>First item is always an assignment to <CODE>$this-&gt;in</CODE>. If set
to 1, auto_init has been executed by this session. If
<EM>not</EM> set to 1, auto_init has not been executed, yet.
This may be because no auto_init file is defined for
that session.
<P>After that comes code like this: <CODE>$this-&gt;pt = array();</CODE>
followed by a bunch of assignments like
<CODE>$this-&gt;pt["somestring"] = 1;</CODE>. Each somestring is the
name of a registered variable. Variable registrations are
persistent themselves and are saved with the <CODE>$this-&gt;pt</CODE>
array. Even if the variable in question is not set, it may be
registered and stays so until it is unregistered or the session
is deleted. Check the contents of the pt array is you want to
see which variables are currently registered with your session.
<P>Finally, the actual contents of your variables are saved. This
is always done by accessing the $GLOBALS array and always by
enumerating the scalar values that make up the persistent
variable. For a scalar, you will see code like
<CODE>$GLOBALS[somevar] = "value";</CODE>. 
<P>For an array, first <CODE>$GLOBALS[someary] = array();</CODE>
is generated. Then the scalars that make up the array, if any,
are written out, generating code that looks like
<CODE>$GLOBALS[someary][index] = "value"</CODE>.
<P>And for objects, code to create an object instance is saved:
<CODE>$GLOBALS[someobj] = new Classname;</CODE>. "Classname"
is taken from the objects <CODE>$classname</CODE> slot, which <EM>must</EM>
be present and accurate. Then the scalars that are to be saved
are written out, according to the contents of the objects
<CODE>persistent_slots</CODE> array:
<CODE>$GLOBALS[someobj]-&gt;slot = "value";</CODE> is written.
<P>If you want to see what values have been saved to the
database, you just have to look at the <CODE>$GLOBALS</CODE> assignments
for that session.
<P>
<H3>How "serialize()" operates</H3>

<P>
<P>The following information is applicable only to library
developers, that is, programmers that want to change the
internal workings of PHPLIB. You may safely skip this section;
some information here requires advanced understanding of the PHP
language.
<P>The heart of the session class is the <CODE>serialize()</CODE> internal
function. This function takes an expression called prefix and
generates PHP code that will assign the value of that expression
to the expression when executed. For example, if the expression
is <CODE>$GLOBALS["a"]</CODE> and the global variable <CODE>$a</CODE>
has the value <CODE>17</CODE>, then serialize will create the PHP
program <CODE>$GLOBALS["a"] = "17";</CODE>. To save memory,
<CODE>serialize()</CODE> operates on a reference parameter <CODE>$str</CODE>,
where is will append the code generated.
<P>First thing <CODE>serialize()</CODE> does is to determine the type of
the current expression using the PHP <CODE>gettype()</CODE> function.
The current type is stored in <CODE>$t</CODE>. The type of the
expression may indicate either a scalar value (integer number,
float number or string), an array or an object.
<P>Scalar values are the easiest to handle: <CODE>serialize()</CODE> just
evaluates the current expression and remembers the result value
in <CODE>$l</CODE>. An assignment is generated that will assign the
current value to the current expression. Since the current value
may be a string and that string may contain bad characters (any
of backslash, double quotes or dollar sign), these characters
are backslashed. We are done, <CODE>serialize()</CODE> ends here for
scalars.
<P>In the case of <CODE>$t</CODE> indicating an array, code is generated to
create an empty array (<CODE>expression = array();</CODE>). Then the
keys of current expression are enumerated and for each key
<CODE>serialize()</CODE> is called recursively with the current key
appended to the expression. That will append code for each array
slot.
<P>Should <CODE>$t</CODE> indicate an object, code is generated to create
that object (<CODE>expression = new Classname;</CODE>). Since one cannot
find out the name of the class of an object for arbitrary
objects in PHP, objects handled by <CODE>serialize()</CODE> must have a
slot named <CODE>classname</CODE>. The object handler will then
enumerate the contents of the objects slot <CODE>persistent_slots</CODE>
and call <CODE>serialize()</CODE> recursively for each of these slots
with the appropriate prefix.
<P>Since many of the expressions used in <CODE>serialize()</CODE> require
variable variable names or even variable code, <CODE>eval()</CODE> is
used liberally. Unfortunately, this makes the code hard to read.
<P>
<H2><A NAME="ss3.9">3.9 Auth</A>
</H2>

<P>
<P>Authentication management can be used to authenticate a session,
that is, to identify the user at the client side of the session.
<P>Authentication is done inline, with HTML forms, <EM>not</EM> with
HTTP authentication (that's the browser popup you get when you
hit a page protected with htaccess). Inline authentication has
several advantages over HTTP authentication:
<P>
<UL>
<LI>It can be undone: A session can be un-authenticated, the
user can "log out".</LI>
<LI>It can expire: A session can automatically be
un-authenticated after a given idle time.</LI>
<LI>It can be customized: You are not limited to user/password
pairs. Instead you could use a customer number, operator id
and a password to log in. Also, you have full control over
the login screen, which is a normal HTML page with logos,
help and forms as you see fit.</LI>
<LI>It is database based. Authentication is being done against
a database of your design, not a htpasswd text file.</LI>
<LI>It is per page. You decide on a per-page basis which pages
are authenticated and which aren't.</LI>
<LI>It can be user authenticating and optionally self
registering. In <EM>registration</EM> mode, a user without a valid login is
encouraged to register and an account is created for this
user.</LI>
<LI>It works with CGI PHP. HTTP authentication is available
only in mod_php.</LI>
<LI>It is integrated with a permission checking scheme.</LI>
</UL>
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
classname</TD><TD>Serialization helper: The name of this class.</TD></TR><TR><TD>
persistent_slots</TD><TD>Serialization helper: The names of all persistent slots.</TD></TR><TR><TD>
lifetime</TD><TD>Maximum allowed idle time before the authentication expires. If set to 0, The authentication never expires (as long as the session remains active).</TD></TR><TR><TD>
refresh</TD><TD>Maximum allowed time before the authentication info (perms and alike) are re-read from the database calling <CODE>auth_refreshlogin()</CODE> method. If set to 0 authentication info are read only at the login stage.</TD></TR><TR><TD>
mode</TD><TD>Authentication mode: <CODE>log</CODE> or <CODE>reg</CODE> (see below).</TD></TR><TR><TD>
database_class</TD><TD>A classname. Auth uses this class to make a database connection.</TD></TR><TR><TD>
database_table</TD><TD>Database table used to keep the session variables.</TD></TR><TR><TD>
magic</TD><TD>An arbitrary value used in uniqid generation.</TD></TR><TR><TD>
nobody</TD><TD>Flag: If true, we use default authentication.</TD></TR><TR><TD>
cancel&thinsp;login</TD><TD>The name of a button that can be used to cancel a login form</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
db</TD><TD>Internal: The database connection object instance.</TD></TR><TR><TD>
auth</TD><TD>Internal: User authentication information, see below.</TD></TR><TR><TD>
in</TD><TD>Internal: Used in default authentication mode.</TD></TR><TR><TD>

<CAPTION>Internal instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Instance methods</H3>

<P>
<P>
<H3>Accessible instance methods</H3>

<P>
<P>
<DL>
<DT><B>url()</B><DD><P>  
A function that can be used in <CODE>auth_loginform()</CODE>a and
<CODE>auth_registerform</CODE>. It returns the appropriate "action="
attribute to the form tag.
<P>
<DT><B>purl()</B><DD><P>  
A function that can be used in <CODE>auth_loginform()</CODE>a and
<CODE>auth_registerform</CODE>. It prints the appropriate "action="
attribute to the form tag.
<P>
<DT><B>login_if($t)</B><DD><P>A function that can be used to change the current user
identity. 
See the section and example on using default authentication
below.
<P>
<DT><B>unauth($nobody = false)</B><DD><P>This function destroys the authentication information in
<CODE>$this-&gt;auth</CODE>, forcing the user to relogin the next time
a protected page is being loaded.
<P><CODE>$this-&gt;auth["uname"]</CODE> is being kept, so that the
correct username is available as a default.
<P>Since V6: To give the user the credentials of `nobody', pass
true as the first parameter to unauth. This will also change
<CODE>$this-&gt;auth["uname"]</CODE>.
<P>Since V7.2: Passing $nobody to this method is deprecated.
<P>
<DT><B>logout($nobody = $this-&gt;nobody)</B><DD><P>This function destroy all authentication information
in  <CODE>$this-&gt;auth</CODE>, forcing the user to relogin
the next time a protected page is being loaded.
<P>Most applications want to use <CODE>$this-&gt;unauth()</CODE>
instead.
<P>Since V6: To give the user the credentials of `nobody', pass
true as the first parameter to logout. This defaults to the 
value you set in the class definition (<CODE>$nobody</CODE>).
<CODE>logout()</CODE> will call <CODE>unauth()</CODE> (passing <CODE>$nobody</CODE>),
so the behaviour is identical (except <CODE>logout()</CODE> will always
clear <CODE>$this-&gt;auth["uname"]</CODE> and unregister the auth class).
<P>Since V7.2: Passing $nobody to this method is deprecated.
<P>
<DT><B>is_authenticated()</B><DD><P>Will return false, if the current authentication is
invalid or expired. Will return the authenticated uid
otherwise.
<P>
<DT><B>auth_preauth()</B><DD><P>This function can be overridden in a subclass to Auth. It
is being called as the very first step in the authentication
process and has the opportunity to authenticate the user
without a loginform being displayed (by deriving all necessary
information telepathically, or by using cookies, or divining
the user identities from the incestines of a dead squirrel).
<P>If it returns a UID value, the user is authenticated and neither
auth_loginform() nor auth_validatelogin() are
called. If it returns false, all goes on as usual.
<P>
<DT><B>auth_loginform()</B><DD><P>  
This function must be overridden by a subclass to Auth. It
should output HTML that creates a login screen for the user.
We recommend that you use an <CODE>include()</CODE> statement to include
your HTML file.
<P>
<DT><B>auth_validatelogin()</B><DD><P>This function is called when the user submits the login form
created by <CODE>auth_loginform()</CODE>. It must validate the user input.
<P>If the user authenticated successfully, it must set up
several fields within the <CODE>$auth[]</CODE> instance variable:
<P>
<DL>
<DT><B>"uid"</B><DD><P>must contain the user id associated with that user.
<DT><B>"uname"</B><DD><P>must contain the user name as entered by the user.
<DT><B>"exp"</B><DD><P>must not be tampered with (field is maintained by
<CODE>start()</CODE>, contains the time when the login expires).
<DT><B>"perm"</B><DD><P>if you want to use the permission feature, you
must store the permissions of the validated user here.
(Hint: due to a name conflict with sybase, "perm" is called "perms"
in all the databases tables. Look for this small difference!)
</DL>
<P>See the example below for more information.
<P>
<DT><B>auth_refreshlogin()</B><DD><P>This function is called every <CODE>refresh</CODE> minutes. It must refresh
the authentication informations stored in <CODE>auth</CODE> array by
<CODE>auth_validatelogin()</CODE> method. It is not called if the
user is logged in as nobody.
<P>It must return true on success, false otherwise (i.e.: the userid
is no longer valid).
<P>
<DT><B>auth_registerform()</B><DD><P>See auth_doregister().
<P>
<DT><B>auth_doregister()</B><DD><P>These functions mirror <CODE>auth_loginform()</CODE> and
<CODE>auth_validatelogin()</CODE> in registration mode.
</DL>
<P>
<H3>Internal instance methods</H3>

<P>
<P>
<DL>
<DT><B>start()</B><DD><P>
<P>Initialization function, does the authentication. If we are
in <CODE>log</CODE> (login) mode, <CODE>auth_loginform()</CODE> is
called to draw a login screen. When the login screen is
submitted back, <CODE>auth_validatelogin()</CODE> is called to
validate the login. If the validation was successful, the
actual page content is shown, otherwise we're back at
<CODE>auth_loginform()</CODE>.
<P>In <CODE>reg</CODE> mode, <CODE>auth_registerform()</CODE> is called to draw a
registration form. When the registration form is submitted
back, <CODE>auth_doregister()</CODE> is called to register the user and
to validate the session. If registration was successful, the
actual page content is shown, otherwise we're back at
<CODE>auth_registerform()</CODE>.
</DL>
<P>
<H3>Example</H3>

<P>Use a subclass of <CODE>Auth</CODE> to provide parameters for your
authentication class and to implement your own <CODE>auth_*</CODE> functions.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Auth extends Auth {
  var $classname        = "My_Auth"; # Object serialization support

  var $lifetime         =  15;
  
  ## DB_Sql subclass and database table to use
  var $database_class = "DB_Session";
  var $database_table = "auth_user";

  ## Some magic value to make our uids harder to guess.
  var $magic = "Abracadabra";

  ## Use an own login form
  function auth_loginform() {
    global $sess;
    include("loginform.ihtml");
  }
  
  function auth_validatelogin() {
    global $username, $password;    ## form variables from loginform.ihtml
    
    ## If authentication fails, loginform.html will
    ## find $this-&gt;auth["uname"] set and use it.
    $this-&gt;auth["uname"]=$username;
    
    ## Value to return in case auth fails.
    $uid   = false;
    
    ## Check the database for this user and password pair.
    $query = sprintf(
      "select * from %s where username = '%s' and password = '%s'",
      $this-&gt;database_table,
      addslashes($username),
      addslashes($password)
    );
    $this-&gt;db-&gt;query($query);
    
    ## If we found a matching user, grab the uid and permissions...
    while($this-&gt;db-&gt;next_record()) {
      ## Required.
      $uid = $this-&gt;db-&gt;f("uid");
      
      ## Optional, for the perm feature.
      $this-&gt;auth["perm"] = $this-&gt;db-&gt;f("perms");
      ## if you use perm feature be aware, that the db-field in our
      ## example table is called "perms" due to a name conflict with sybase
    }
    
    return $uid;
  }
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Your <CODE>loginform.ihtml</CODE> contains HTML and PHP code to draw a login
form. <CODE>$this-&gt;auth["uname"]</CODE> will be empty on the first login
attempt and set on all further login attempts. You can use this
to detect repeated login attempts and display an appropriate
error message. You must print the result of <CODE>$this-&gt;url()</CODE> to
create your forms action attribute.
<P>See the provided <CODE>loginform.ihtml</CODE> for an example.
<P>Use the page management functions (see above) to use your
authentication subclass. The feature name for authentication
management is <CODE>auth</CODE>; provide the name of your <CODE>Auth</CODE> subclass as
a parameter to the <CODE>auth</CODE> feature. The <CODE>auth</CODE> feature requires the
<CODE>sess</CODE> feature:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
  page_open(array("sess" =&gt; "My_Session", "auth" =&gt; "My_Auth"));
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<H3>Using default authentication</H3>

<P>Many applications want to use <CODE>$auth</CODE> and <CODE>$perm</CODE>
objects to protect functionality on a page, but do want to
make the unprotected part of this page available to users
with no account. This presents a kind of dilemma, because you
need <CODE>$auth</CODE> and <CODE>$perm</CODE> objects to protect
functionality on a page, but you don't want a login screen to
appear by default.
<P>Default authentication solves this dilemma by providing a
special uid and uname "nobody", which is guaranteed to fail
every permission check. If you set the <CODE>nobody</CODE> flag,
<CODE>$auth</CODE> will not create a login screen to force a user to
authenticate, but will authenticate the user silently as
<CODE>nobody</CODE>. The application must offer a login button or
other facility for users with accounts to change from that
id to their real user id.
<P>To use default authentication, create a subclass of <CODE>My_Auth</CODE>
as shown above with the <CODE>nobody</CODE> flag set (<EM>Note:</EM> No need
to extend in two steps. The only important thing here is that
the <CODE>nobody</CODE> flag is set.)
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Default_Auth extends My_Auth {
  var $classname = "My_Default_Auth";

  var $nobody = true;
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>To create a page that uses default authentication, use the page
management functions. Check for relogin requests with the
<CODE>login_if()</CODE> function. Create a relogin link on your page.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
&lt;?php
  // using Default Authentication
  page_open(array("sess" =&gt; "My_Session", "auth" =&gt; "My_Default_Auth"));
  $auth-&gt;login_if($again);

  if ($auth-&gt;auth["uid"] == "nobody"):
?&gt;
  &lt;A HREF="&lt;?php $sess-&gt;purl("$PHP_SELF?again=yes") ?&gt;"&gt;Relogin&lt;/A&gt;
  to this page.
&lt;?php endif ?&gt;
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<H3>Using Challenge-Response Authentication</H3>

<P>As distributed, <CODE>local.inc</CODE> contains an example class
named <CODE>Example_Challenge_Auth</CODE>, which uses a
Challenge-Response authentication scheme. If the client
browser supports Javascript, this login screen does not
transmit passwords in clear over the network. If the client
does not support Javascript, login is still possible, but
passwords are transmitted in clear, as regular <CODE>Example_Auth</CODE>
always does.
<P><CODE>Example_Challenge_Auth</CODE> is there to demonstrate advanced
usage of PHP and Javascript and to show off the flexibility
of the library base classes: The Challenge-Response
authentication scheme has been implemented completely and
naturally in local.inc by subclassing <CODE>Auth</CODE> with no
alteration of library code.
<P><CODE>Example_Challenge_Auth</CODE> includes <CODE>crloginform.ihtml</CODE>. It
also requires that the file <CODE>md5.js</CODE> is present in the
document root directory of your web server. That file contains
an implementation of the MD5 message digest algorithm done by
Henri Torgemane. The basic idea behind this authentication
scheme is simple: <CODE>$auth-&gt;auth_loginform()</CODE> creates a
challenge value which is incorporated into this form. When
the user tries to submit the form,
MD5("username:password:challenge") is calculated and filled
into the reply field. The password field is erased. The
server can calculate the expected reply from the username
received, the password in the database and the challenge,
which it knows. It can compare the expected reply to the
actual reply value. If they match, the user is authenticated.
<P>If the reply field is empty and password is set, the server
knows that the client cannot do Javascript. The user can still be
authenticated, but the password is visible on the network.    
<P>The class is a dropin-replacement for <CODE>Example_Auth</CODE>.
<P>
<H3>The complete guide to authentication and user variables</H3>

<P>
<P>This feature has originally been written for the PHPLIB mailing
list by Kristian Khntopp and was included into the
documentation later.
<P>
<H3>How is the <CODE>Auth</CODE> class used usually?</H3>

<P>
<P>Usually, you write code like this into the top of the page you
want to protect:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
&lt;?php
page_open(array(
    "sess" =&gt; "My_Session", 
    "auth" =&gt; "My_Auth",
    "perm" =&gt; "My_Perm")); // see Perm docs for specifics
$perm-&gt;check("admin");            // see Perm docs
?&gt;

&lt;!-- your code here --&gt;

&lt;?php
page_close()
?&gt;
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<H3>How does <CODE>$auth</CODE> work internally?</H3>

<P>
<P>When you access this page, the call <CODE>to page_open()</CODE> is being
made as the first thing on that page. <CODE>page_open()</CODE> creates
an instance of <CODE>My_Auth</CODE> named <CODE>$auth</CODE> and starts it.
<CODE>$auth</CODE> then detects that you are not authenticated (how it
does, I will explain below) and displays <CODE>loginform.ihtml</CODE>.
$auth then exits the interpreter, so that &lt;!-- your code here
--&gt; is never being executed or displayed.
<P>The user now sits in front of a <CODE>loginform.ihtml</CODE> screen,
which is shown under the URL of the page the user originally
tried to access. The loginform has an action URL, which just
points back to itself.
<P>When the user filled out the loginform and submits it, the very
same URL is requested and the above <CODE>page_open()</CODE> is
reexecuted, but this time a username and a password are
submitted. When the <CODE>$auth</CODE> object is created and started, it
detects these parameters and validates them, resulting in either
a NULL value or a valid user id. If the validation failed,
creating an empty user id, the loginform is displayed again and
the interpreter exits. Again &lt;!-- your code here --&gt; is not
executed.
<P>If a UID is returned, that UID and a timestamp are being made
persistent in that session and <CODE>$auth</CODE> returns control to
<CODE>page_open()</CODE>. When <CODE>page_open()</CODE> finishes, which it may
or may not do, depending on the presence and result of an
optional <CODE>$perm</CODE> check, &lt;!-- your code here --&gt; is being
executed or shown.
<P>Later calls to other pages or the same page check for the
presence of the UID and the timestamp in the sessions data. If
the UID is present and the timestamp is still valid, the UID is
retained and the timestamp is refreshed. On <CODE>page_close()</CODE>
both are written back to the user database (Note: Authenticated
pages REQUIRE that you <CODE>page_close()</CODE> them, even when you
access them read-only or the timestamp will not be refreshed).
<P>If the UID is not present (<CODE>$auth-&gt;logout()</CODE> or
<CODE>$auth-&gt;unauth()</CODE> have been called, for example) or the
timestamp has expired, <CODE>$auth</CODE> will again intercept page
display and draw the loginform.
<P>The only way to get into a page with an <CODE>$auth</CODE> object on it
is to have a UID and a valid timestamp in your session data
(Note: This is true even for default authentication. These
create a dummy UID and timestamp in your session data).
<P>
<H3>How do $sess and $auth interact?</H3>

<P>
<P>Your browser has a session cookie, named after your session
class. This is the only thing that is ever shipped between your
browser and PHPLIB, as far as core functionality is concerned.
The session cookie value is used as a reference into
<CODE>active_sessions</CODE>, to retrieve PHPLIB generated PHP code,
which is then <CODE>eval()</CODE>ed and recreates your session variables
within <CODE>page_open()</CODE>.
<P>Part of the <CODE>$auth</CODE> object now makes itself persistent and is
retrieved when the <CODE>$sess</CODE> part of <CODE>page_open()</CODE> is being
executed. This is just before the <CODE>$auth</CODE> part of
<CODE>page_open()</CODE> gets its turn, so that <CODE>$auth</CODE> can rely on its
persistent data being present when it is being called.
<P>From the PHPLIB source you all know that <CODE>$auth</CODE> has only one
persistent slot, called <CODE>$auth-&gt;auth[]</CODE>, of type hash. This
hash contains the slots <CODE>uid</CODE>, <CODE>exp</CODE> and <CODE>uname</CODE>.
<CODE>$auth-&gt;auth["uid"]</CODE> is the currently authenticated user id,
<CODE>$auth-&gt;auth["exp"]</CODE> is the currently active expiration
timestamp (Unix time_t format) for that uid.
<CODE>$auth-&gt;auth["uname"]</CODE> is completely irrelevant as far as the
regular PHPLIB <CODE>Auth</CODE> class is concerned. It is relevant in
the context of the supplied default <CODE>Auth</CODE> subclass
<CODE>Example_Auth</CODE>, though.
<P>So a session is authenticated, if it contains <CODE>$auth-&gt;auth["uid"] !=
false</CODE> and <CODE>time() &lt; $auth-&gt;auth["exp"]</CODE>.
<P>
<H3>Where is the beef?</H3>

<P>
<P>The original <CODE>Auth</CODE> class as included in PHPLIB makes no
assumptions at all on how a loginform looks or how and where
uids come from. There is no code at all in <CODE>Auth</CODE> that ever
checks anything but the above two conditions. It is your
responsibility to modifiy a subclass of Auth in a way that these
conditions can ever be met.
<P>Auth helps you in doing this by calling its own function
<CODE>$auth-&gt;auth_loginform()</CODE> when it wants to draw a loginform.
Unfortunately this function is empty in Auth itself, so you have
to provide an implementation for that. The suggested standard
implementation in <CODE>local.inc</CODE>s <CODE>Auth</CODE> subclass
<CODE>Example_Auth</CODE> is
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
function auth_loginform() {
  include("loginform.ihtml");
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>and you put your code into that file. We also provide sample
code for that file, but you are not limited to that code and may
write a <CODE>loginform.ihtml</CODE> as it meets your needs.
<P>When the loginform has been filled in and submitted back by the
user, <CODE>Auth</CODE> calls <CODE>$auth-&gt;auth_validatelogin()</CODE>. Again,
this function is empty in <CODE>Auth</CODE> itself and so <CODE>Auth</CODE> by
itself will never function correctly. You have to subclass
<CODE>Auth</CODE> and provide your own implementation of
<CODE>$auth-&gt;auth_validatelogin()</CODE> in <CODE>local.inc</CODE> to make it
work.
<P>What you actually do in that function is completely irrelevant
to <CODE>Auth</CODE> itself. It only exspects that you either return
false, if the user-supplied authentication data was invalid, or
a user id, if the user could be validated. <CODE>Auth</CODE> then takes
care to create the appropriate entries (<CODE>$auth-&gt;auth["uid"]</CODE>
and <CODE>$auth-&gt;auth["exp"]</CODE>) in the session record.
<P>
<H3>I still do not understand! What am I supposed to code?</H3>

<P>
<P>
<P>You write your code into <CODE>local.inc</CODE>, after you have removed
the classes <CODE>Example_Auth</CODE>, <CODE>Example_Default_Auth</CODE> and
<CODE>Example_Challenge_Auth</CODE> from that file (keep a copy around, just
for reference).
<P>You code a class called <CODE>My_Auth</CODE> and you use that name later
in your calls to <CODE>page_open</CODE> as an argument to the <CODE>auth</CODE>
feature, as show at the start of this message. Follow the
standard rules for deriving persistent classes in PHPLIB when
you create your code, that is, do it like this:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>

class My_Auth extends Auth {
var $classname = "My_Auth";
// we inherit $persistent_slots and do not need to modify it.

// later code is inserted here
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Now configure the lifetime of the authentication, that is, how
many minutes in the future shall the current value of
<CODE>$auth-&gt;auth["exp"]</CODE> be? Also, name a database connector
class and name the table that you will be using to check
usernames and passwords.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
  // insert this code as indicated above.
  var $lifetime = 15;
  var $database_class = "DB_Example";
  var $database_table = "my_special_user_table";

  // later code is inserted here
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Okay, now we have a basic implementation of <CODE>My_Auth</CODE>
that is only lacking the required functions
<CODE>auth_loginform()</CODE> and
<CODE>auth_validatelogin()</CODE>. Our implementation of
<CODE>auth_loginform()</CODE> will have access to all <CODE>$sess</CODE>
variables by globaling <CODE>$sess</CODE> into our context (because
these can come in handy) and to all <CODE>$auth</CODE> variables (via
<CODE>$this</CODE>).
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>

function auth_loginform() {
  global $sess;
  include("loginform.ihtml");
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>The loginform is free to do whatever it damn well pleases to
create a form for the user to supply the needed values for
authentication. It has access to anything <CODE>$sess</CODE> and
anything <CODE>$this</CODE> related.
<P>The loginform will display some input fields for the user, for
example a given name, a surname and a password. When the form is
submitted back, <CODE>auth_validatelogin()</CODE> is being called. The
form values are global variables (or <CODE>$HTTP_x_VARS[]</CODE>) and
must be imported into <CODE>$auth-&gt;auth_validatelogin()</CODE>. Then,
<CODE>$auth-&gt;auth_validatelogin()</CODE> is free to do whatever it must
do to produce a unique identifier for that user (or return
false).
<P>Suppose you created input fields named given_name, surname and
password. So go ahead, global <CODE>$given_name</CODE>, <CODE>$surname</CODE>
and <CODE>$password</CODE> and set <CODE>$uid</CODE> to false. Then create the
SQL needed to access you user table and retrieve the user record
from your database as indicated by <CODE>$given_name</CODE> and
<CODE>$surname</CODE> and <CODE>$password</CODE>.
<P>The query may succeed, if a record with matching
<CODE>$given_name</CODE>, <CODE>$surname</CODE> and <CODE>$password</CODE> is present.
In that case return the uid, which uniquely identifies exactly
that (given_name, surname) pair. Else return false.
<P>In code:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>

function auth_validatelogin() {
  // import authentication data
  global $given_name, $surname, $password;

  $uid = false;

  $query = sprintf("select uid
                      from %s
                     where given_name = '%s'
                       and surname = '%s'
                       and password = '%s'",
             $this-&gt;database_table,
             addslashes($given_name), addslashes($surname), 
             addslashes($password));

  // $auth-&gt;db is our DB_Example database connection
  $this-&gt;db-&gt;query($query);

  // now check for any results
  while($this-&gt;db-&gt;next_record()) {
    $uid = $this-&gt;db-&gt;f("uid");
  }

  // either $uid is false now (no results)
  // or set to the last retrieved value from the uid
  // column.

  // Anyway we are set now and can return control
  return $uid;
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Okay, that's all and useable now. There is room for some
improvements, though: First we did not retrieve permission data,
so this will not work, if we want to use the perm feature as
well.
<P>This is easily changed: Modify the query to <CODE>select uid,
perms</CODE> instead of <CODE>select uid</CODE> alone. Of course, you may call
your perm column whatever you like, just adapt the SQL
accordingly. Also, add a line after the
<CODE>$uid</CODE> assignment so that the code looks like this:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
  $uid = $this-&gt;db-&gt;f("uid");
  $this-&gt;auth["perm"] = $this-&gt;db-&gt;f("perms");
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>This will store the retrived <CODE>perms</CODE> value under the key
<CODE>perm</CODE> within the <CODE>$auth-&gt;auth[]</CODE> array. It will be kept
around in that place in case <CODE>$perm</CODE> is called and starts
looking for the current permissions of that user.
<P>Another possible improvement becomes apparent when you try to
login and fail to do so correctly: <CODE>auth_validatelogin()</CODE>
returns false and you hit the loginform again. Empty loginform
that is, because we did not remember what you typed into the
<CODE>given_name</CODE> and <CODE>surname</CODE> fields before. If we remembered
what you typed, we could easily supply these values back to you
so that you can correct them. We would also be able to detect if
this is a second, third, ... attempt to login and display an
appropriate error message somewhere in that loginform to inform
the user of his or her typo. A convenient place to store these
values is the <CODE>$auth-&gt;auth</CODE> array, which is persistent
anyway. 
<P>Standard <CODE>Example_Auth</CODE> uses the field <CODE>$auth-&gt;auth["uname"]</CODE>
to store that value, but you may use any field and as many
fields as you like as long as you make sure not to clash with
any of the three officially used fields, <CODE>uid</CODE>, <CODE>exp</CODE>, and
<CODE>perm</CODE>.
<P>Do not try to turn the global variables <CODE>$given_name</CODE> and
<CODE>$surname</CODE> into persistent variables by calling
<CODE>$sess-&gt;register("given_name")</CODE> and
<CODE>$sess-&gt;register("surname")</CODE>! Remember: These are form
variables! Never ever make form variables persistent and never
ever trust unvalidated user supplied from the Internet!
<P>So add the folling code just below the "global" line:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
  $this-&gt;auth["gname"] = $given_name;
  $this-&gt;auth["sname"] = $surname;
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>and check for these two variables in loginform.ihtml at the
appropriate places.
<P>
<H3>Ok, I did that and it works. I even understood it. Now, what exactly is that uid used for?</H3>

<P>
<P>It is simply a token to indicate that the user is authenticated.
We use a different token for each user, so that we can decide
which user we are currently dealing with. You can think of the
uid as a primary key for your <CODE>auth_user</CODE> table (or whatever
it is being called in your current application). The (
given_name, surname ) tuple would also be a possible primary
key, albeit a compound one. It is the external, human-readable
(and probably sometimes very long) representation of the
internal uid. The password field is functionally dependent on
either of both key candidates.
<P>The internal user id should never be presented to the user; the (
given_name, surname ) pair is much more natural to handle for the user
and easier to remember (A user who does not remember his or her name
would probably not be in a state of mind to operate the rest of the
application anyway :-).
<P>The internal user id should always be used to identify a user
internally within an application, though. That is, because the
uid is of a fixed length and has a known form and structure, so
you can make assumptions. A given_name or surname may be of any
length and may contain about any character, so you probably do
not want to use this as a user-reference internally.
<P>
<H3>But is the uid used internally by PHPLIB?</H3>

<P>
<P>Yes, if you make use of the <CODE>user</CODE> feature of
<CODE>page_open()</CODE>, that is, if you create user variables.
<P>The <CODE>User</CODE> class is actually a subclass of <CODE>Session</CODE>. That
is, user variables are just like session variables. They are
even stored in <CODE>active_sessions</CODE>. The only difference is that
the session has a different name (it is called <CODE>Example_User</CODE>
instead of <CODE>Example_Session</CODE>, if you use the classes and names
supplied in <CODE>local.inc</CODE>).
<P>And in <CODE>Example_User</CODE>, the user id of the authenticated user
becomes the session id in the <CODE>active_sessions</CODE> table. That
is the reason why we recommend <CODE>md5(uniqid("abracadabra"))</CODE>
style uids.
<P>
<H2><A NAME="ss3.10">3.10 Perm</A>
</H2>

<P>
<P>Permission management relies on an authenticated session. It
associates a set of required permissions with a page. The actual
page content is only visible to users with ALL matching
permissions; all other users are shown a screen of your design.
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
classname</TD><TD>Serialization helper: The name of this class.</TD></TR><TR><TD>
permissions</TD><TD>A hash of (name, permission bit) pairs.</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Instance methods</H3>

<P>
<P>
<H3>Accessible instance methods</H3>

<P>
<P>
<DL>
<DT><B>check($required)</B><DD><P>Checks that the currently authenticated user has all the
rights that are specified in <CODE>required</CODE>. If not,
<CODE>perm_invalid()</CODE> is called.
<P>If one or more of the required rights or user rights are
invalid (not to be found in the permissions hash),
<CODE>perm_invalid()</CODE> is called as well.
<P>
<DT><B>have_perm($required)</B><DD><P>Similar to <CODE>check()</CODE> in usage, only that it doesn't halt the
session if the user doesn't have the appropriate rights: This
function returns true, if the user has the required rights,
false otherwise.
<P>
<DT><B>perm_sel($name, $current = "", $class = "")</B><DD><P>This function returns a <CODE>SELECT</CODE>-tag with the given
<CODE>name</CODE>. Within this tag, all available permission values from
<CODE>$perm->permissions</CODE> are contained as <CODE>OPTION</CODE> tags.
<P>
<P>If you supply a value for <CODE>current</CODE>, the permission value
that matches <CODE>current</CODE> is <CODE>SELECTED</CODE>. If you supply a
value for <CODE>class</CODE>, the tags are marked with that CSS
stylesheet class.
<P>
</DL>
<P>
<H3>Internal instance methods</H3>

<P>   
<P>
<DL>
<DT><B>permsum($rights)</B><DD><P>Logically or's all the rights and returns a pair <CODE>(valid,
or_result)</CODE>. If valid is true, an <CODE>or_result</CODE>
is provided. If valid is false, the <CODE>or_result</CODE> is
undefined and one or more of the rights do not exist at all.
This is a severe error and the application should be halted at
once.
<P>
<DT><B>perm_invalid($does_have, $must_have)</B><DD><P>Called in case of an access violation. <CODE>does_have</CODE> is a string
listing the rights the user actually has. <CODE>must_have</CODE> are the
rights the page requires.
</DL>
<P>
<H3>Example</H3>

<P>
<P>Use a subclass of <CODE>Perm</CODE> to provide parameters for your
permission class and to implement your own <CODE>perm_invalid</CODE>
function.
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_Perm extends Perm {
  var $classname = "My_Perm";
  
  var $permissions = array (
    "user"          =&gt; 1,
    "author"        =&gt; 2,
    "editor"        =&gt; 4,
    "moderator"     =&gt; 8,
    "admin"         =&gt; 16
  );
  
  function perm_invalid($does_have, $must_have) {
    global $perm, $auth, $sess;
    
    include("perminvalid.ihtml");
  }
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Use the page management functions (see above) to use your
permission subclass. The feature name for permission
management is <CODE>perm</CODE>; provide the name of your <CODE>Perm</CODE> subclass as
a parameter to the <CODE>perm</CODE> feature. The <CODE>perm</CODE> feature requires the
<CODE>sess</CODE> feature and the <CODE>auth</CODE> feature:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
  page_open(array("sess" =&gt; "My_Session", "auth" =&gt; "My_Auth", "perm" =&gt; "My_Perm"));
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Use the <CODE>check()</CODE> instance method to protect your page:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
  $perm-&gt;check("admin");  ## This page is for users with admin rights only.
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Use <CODE>have_perm()</CODE> to create protected functionality on a
page:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
&lt;?php
  if ($perm-&gt;have_perm("admin")):
 ?&gt;
  &lt;h1&gt;Admin only functionality&lt;/h1&gt;
&lt;?php
  endif;
 ?&gt;
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>
<H3>How permissions work </H3>

<P> 
<P>Your subclass of <CODE>Perm</CODE> defines an array <CODE>$permissions</CODE>,
which translates permission names into bit patterns. For
example, the definition of <CODE>Example_Perm</CODE> in the distributed
<CODE>local.inc</CODE> defines the names <CODE>user</CODE>, <CODE>author</CODE>,
<CODE>editor</CODE>, <CODE>supervisor</CODE> and <CODE>admin</CODE>, all of which
translate into a bit pattern with a single bit set.
<P>A user may be assigned any number of permissions as a comma
separated list of permission names (no spaces!) in the
<CODE>perms</CODE> column of the <CODE>auth_user</CODE> table. The effective
permissions of the user are determined by logically OR'ing the
bit patterns of these permissions.
<P>A page may require any permissions as a comma separated list of
permission names (again no spaces!) with the
<CODE>$perm-&gt;check()</CODE> function. The required permissions are
again determined by logically OR'ing the bit patterns of these
permissions. Similarly, a page function may be protected by
requiring permissions with <CODE>$perm-&gt;check()</CODE>.
<P>Access is granted to a protected page or a protected page
function, if the effective permissions of the authenticated user
have all the required bits set, that is: If the effective
permissions of the user logically AND'ed with the required
permissions are equal to the required permissions.
<P>With the permission names as defined in <CODE>Example_Perm</CODE> from the
distribution, a user <CODE>kris</CODE> may be defined with <CODE>admin</CODE>
permission in the <CODE>auth_user</CODE> table. A page that requires
<CODE>admin,user</CODE> permission with
<CODE>$perm-&gt;check("user,admin")</CODE> is inaccessible to this user.
This is how it is calculated:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
Effective Permissions of User: admin
              translates into:    16

Required Permissions of Page : user,admin
              translates into:    1 OR 16 == 17

Permission Check: 
        Effective Permissions 16 
AND     Required Permissions  17
ARE     16 &amp; 17 =             16

MUST BE Required Permissions  17 -> access denied
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>The example permissions as defined in <CODE>Example_Perm</CODE> from the
distribution are called <EM>atomic</EM> permissions, because each of
them has only a single bit set. Atomic permissions are the
simplest of all schemes, because they allow for easy permission
checks: To access a page protected with <CODE>user,admin</CODE>, you
need to have at least <CODE>user,admin</CODE> rights in your
<CODE>auth_user</CODE> table.
<P>Another common scheme used in permission definitions are
<CODE>inclusive permissions</CODE>. In this scheme, each permission
definition has all bits of its predecessor set plus one addition
bit. For example
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class Inclusive_Perm extends Perm {
  var $classname = "Inclusive_Perm";

  var $permissions = array(
                            "user"       => 1,
                            "author"     => 3,
                            "editor"     => 7,
                            "supervisor" => 15,
                            "admin"      => 31
                     );
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>defines a set of inclusive permissions. In this example, a user
<CODE>kris</CODE> with <CODE>admin</CODE> permissions can easily access a page
protected with <CODE>editor</CODE> permissions. This is how it is
calculated:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
Effective Permissions of User: admin
              translates into:    31

Required Permissions of Page : editor
              translates into:     7

Permission Check:
        Effective Permissions 31
AND     Required Permissions   7
ARE     31 &amp; 7 =               7

MUST BE Required Permissions   7 -> access granted
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Inclusive Permissions are easy to deal with, too, because a user
with a <EM>higher</EM> access level may access all pages or page
functions with a <EM>lower</EM> access level.
<P>Due to limitations of your machines integer size you can only
define up to 31 permission levels.
<P>
<P>
<H2><A NAME="ss3.11">3.11 User</A>
</H2>

<P>
<P>The user class is an extension (a subclass) of the Session
class. It keeps a list of global variable names and provides a
set of functions to load and save these variables from and to a
database. The same restrictions as for session variables apply
to user variables.
<P>Unlike session variables, user variables are not lost when the
user stops and restarts the browser or moves to a different
workplace (the session id is then lost and consequently all
session variables are lost, since they are bound to the session
id).
<P>User variables require that the user logs in, because they
depend on the availability of a User id to bind variables to
this id. Thus, User is dependent on Auth.
<P>The User class is an extension of the Session class. It has all
instance variables and instance methods of Session, only that
some are implemented different. This documentation only
describes these differences.
<P>Note that Session and User can successfully share a single
<CODE>active_sessions</CODE> table in a database due to the
different values in the <CODE>name</CODE> column.
<P>
<P>
<H3>Instance variables</H3>

<P>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
classname</TD><TD>Serialization helper: The name of this class.magic</TD><TD>Not meaningful for User.</TD></TR><TR><TD>
mode</TD><TD>Not meaningful for User.</TD></TR><TR><TD>
fallback_mode</TD><TD>Not meaningful for User.</TD></TR><TR><TD>
lifetime </TD><TD>Not meaningful for User; see authentication lifetime in Auth instead.</TD></TR><TR><TD>
gc_time</TD><TD>Functional, but probably not useful in User.</TD></TR><TR><TD>
gc_probability</TD><TD>Functional, but should be set to 0 in User.</TD></TR><TR><TD>
that_class</TD><TD>A classname. User uses this class to store and retrieve data.</TD></TR><TR><TD>
auto_init</TD><TD>Not meaningful for User.</TD></TR><TR><TD>
secure_auto_init</TD><TD>Not meaningful for User.</TD></TR><TR><TD>

<CAPTION>Accessible instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<CENTER><TABLE BORDER><TR><TD>
<BR>
pt</TD><TD>Internal array of names of persistent variables.</TD></TR><TR><TD>
name</TD><TD>A tag (name) for the session type.</TD></TR><TR><TD>
id</TD><TD>Id of the current session.</TD></TR><TR><TD>
that</TD><TD>Container object instance.</TD></TR><TR><TD>

<CAPTION>Internal instance variables.</CAPTION>
</TD></TR></TABLE></CENTER>
<P>
<H3>Instance methods</H3>

<P>
<P>
<H3>Accessible instance methods</H3>

<P>
<P>
<DL>
<DT><B>register($varname)</B><DD><P>Works as expected.
<P>
<DT><B>unregister($varname)</B><DD><P>Works as expected.
<P>
<DT><B>delete()</B><DD><P>Works as expected.
<P>
<DT><B>url($url)</B><DD><P>Not useful with User.
<P>
<DT><B>purl($url)</B><DD><P>Not useful with User.
<P>
<DT><B>self_url()</B><DD><P>Not useful with User.
<P>
<DT><B>pself_url()</B><DD><P>Not useful with User.
<P>
<DT><B>reimport_get_vars()</B><DD><P>Works as expected.
<P>
<DT><B>reimport_post_vars()</B><DD><P>Works as expected.
<P>
<DT><B>reimport_cookie_vars()</B><DD><P>Works as expected.
<P>
</DL>
<P>
<H3>Internal instance methods</H3>

<P>
<P>
<DL>
<DT><B>get_id()</B><DD><P>This is only a stub implementation that depends on
the user id provided by the page management functions.
The page management functions will use
<CODE>$auth-&gt;auth["uid"]</CODE>, which is set up by <CODE>Auth</CODE>.
<P>
<DT><B>put_id()</B><DD><P>Empty. Not useful with User.
<P>
<DT><B>serialize($prefix, &amp;$str)</B><DD><P>Works as expected.
<P>
<DT><B>freeze()</B><DD><P>Works as expected.
<P>
<DT><B>thaw()</B><DD><P>Works as expected.
<P>
<DT><B>gc()</B><DD><P>Works as expected. You do not want to use it, though.
<P>
<DT><B>reimport_any_vars($arrayname)</B><DD><P>Works as expected.
<P>
<DT><B>start()</B><DD><P>Initialization function, to be called after object
instantiation. Calls get_id() to get the current session id,
creates a database connection, then calls thaw() to load all
session variables. <EM>Note:</EM> gc() activation  is commented out!
Remove the comments if you really want gc with User variables.
<P>
</DL>
<P>
<P>
<H3>Example</H3>

<P>Use a subclass to provide the appropriate parameters to your
user variables. Usually your subclass looks like this:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
class My_User extends User {
  var $classname = "My_User"; ## Persistence support
  
  var $that_class = "CT_Sql";
}
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Remember that you have to provide a <CODE>DB_Sql</CODE> subclass with the
parameters needed to access your database.
<P>Use the page management functions (see above) to use your
User subclass. The feature name for user variables is
<CODE>user</CODE>; provide the name of your User subclass as a parameter
to the user feature:
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
  page_open(array("sess" =&gt; "My_Session", "auth" =&gt; "My_Auth", "user" =&gt; "My_User"));
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>Use the <CODE>register()</CODE> instance method to register variables as
persistent. If <CODE>$user</CODE> is your user object, use
<P>
<BLOCKQUOTE><CODE>
<HR>
<PRE>
$user-&gt;register("u");
</PRE>
<HR>
</CODE></BLOCKQUOTE>
<P>to make the global variable <CODE>$u</CODE> persistent. <CODE>$u</CODE> may be a
scalar value, an array or an object with persistence support
slots.
<P>Do not use the instance methods <CODE>freeze()</CODE> and <CODE>thaw()</CODE>
directly, but use the page management functions instead.
<P><EM>Note:</EM> Using default authentication and user variables is
going to be a problem, because currently <CODE>User</CODE> does not do
any locking. This is, because the <CODE>DB_Sql</CODE> has currently no
portable locking mechanism.
<P>
<P>
<HR>
<A HREF="documentation-4.html">Next</A>
<A HREF="documentation-2.html">Previous</A>
<A HREF="documentation.html#toc3">Contents</A>
</BODY>
</HTML>