<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<title>Dbstl persistence</title>
<link rel="stylesheet" href="gettingStarted.css" type="text/css" />
<meta name="generator" content="DocBook XSL Stylesheets V1.73.2" />
<link rel="start" href="index.html" title="Berkeley DB Programmer's Reference Guide" />
<link rel="up" href="stl.html" title="Chapter 7. Standard Template Library API" />
<link rel="prev" href="stl_complex_rw.html" title="Store and Retrieve data or objects of complex types" />
<link rel="next" href="stl_container_specific.html" title="Dbstl container specific notes" />
</head>
<body>
<div class="navheader">
<table width="100%" summary="Navigation header">
<tr>
<th colspan="3" align="center">Dbstl persistence</th>
</tr>
<tr>
<td width="20%" align="left"><a accesskey="p" href="stl_complex_rw.html">Prev</a> </td>
<th width="60%" align="center">Chapter 7. Standard Template Library API</th>
<td width="20%" align="right"> <a accesskey="n" href="stl_container_specific.html">Next</a></td>
</tr>
</table>
<hr />
</div>
<div class="sect1" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="stl_persistence"></a>Dbstl persistence</h2>
</div>
</div>
</div>
<div class="toc">
<dl>
<dt>
<span class="sect2">
<a href="stl_persistence.html#directdbget">Direct database get</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="stl_persistence.html#chg_persistence">Change persistence</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="stl_persistence.html#obj_life_persistence">Object life time and persistence </a>
</span>
</dt>
</dl>
</div>
<p>
The following sections provide information on how to achieve
persistence using dbstl.
</p>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="directdbget"></a>Direct database get</h3>
</div>
</div>
</div>
<p>
Each container has a <span class="bold"><strong>begin()</strong></span> method
which produces an iterator. These
<span class="bold"><strong>begin</strong></span> methods take a boolean parameter,
<span class="bold"><strong>directdb_get</strong></span>, which controls the
caching behavior of the iterator. The default value of this parameter
is <code class="literal">true</code>.
</p>
<p>
If <span class="bold"><strong>directdb_get</strong></span> is
<code class="literal">true</code>, then the persistent object is fetched anew
from the database each time the iterator is dereferenced as a pointer
by use of the star-operator
(<span class="bold"><strong>*iterator</strong></span>) or by use of the arrow-operator
(<span class="bold"><strong>iterator->member</strong></span>). If
<span class="bold"><strong>directdb_get</strong></span> is <code class="literal">false</code>, then
the first dereferencing of the iterator fetches the object from the
database, but later dereferences can return cached data.
</p>
<p>
With <span class="bold"><strong>directdb_get</strong></span> set to <code class="literal">true</code>, if you call:
</p>
<pre class="programlisting">(*iterator).datamember1=new-value1;
(*iterator).datamember2=new-value2; </pre>
<p>
then the assignment to <code class="literal">datamember1</code> will be lost,
because the second dereferencing of the iterator would cause the cached
copy of the object to be overwritten by the object's persistent data
from the database.
</p>
<p>
You also can use the arrow operator like this:
</p>
<pre class="programlisting">iterator->datamember1=new-value1;
iterator->datamember2=new-value2; </pre>
<p>
This works exactly the same way as <span class="bold"><strong>iterator::operator*</strong></span>.
For this reason, the same caching rules apply to arrow operators as they do
for star operators.
</p>
<p>
One way to avoid this problem is to create a reference to the object,
and use it to access the object:
</p>
<pre class="programlisting">container::value_type &ref = *iterator;
ref.datamember1=new-value1;
ref.datamember2=new-value2;
...// more member function calls and datamember assignments
ref._DB_STL_StoreElement(); </pre>
<p>
The above code will not lose the newly assigned value of <code class="literal">ref.datamember1</code>
in the way that the previous example did.
</p>
<p>
In order to avoid these complications, you can assign to the object
referenced by an iterator with another object of the same type like this:
</p>
<pre class="programlisting">container::value_type obj2;
obj2.datamember1 = new-value1;
obj2.datamember2 = new-value2;
*itr = obj2; </pre>
<p>
This code snippet causes the new values in <code class="literal">obj2</code> to
be stored into the underlying database.
</p>
<p>
If you have two iterators going through the same container like this:
</p>
<pre class="programlisting">
for (iterator1 = v.begin(), iterator2 = v.begin();
iterator1 != v.end();
++iterator1, ++iterator2) {
*iterator1 = new_value;
print(*iterator2);
} </pre>
<p>
then the printed value will depend on the value of
<span class="bold"><strong>directdb_get</strong></span> with which the
iterator had been created. If <span class="bold"><strong>directdb_get</strong></span>
is <code class="literal">false</code>, then the original, persistent value is
printed; otherwise the newly assigned value is returned from the
cache when <code class="literal">iterator2</code> is dereferenced. This
happens because each iterator has its own cached copy of the
persistent object, and the dereferencing of
<code class="literal">iterator2</code> refreshes
<code class="literal">iterator2</code>'s copy from the database, retrieving
the value stored by the assignment to
<code class="literal">*iterator1</code>.
</p>
<p>
Alternatively, you can set <span class="bold"><strong>directdb_get
</strong></span> to <code class="literal">false</code> and call
<code class="methodname">iterator2->refresh()</code> immediately before
the dereferencing of <code class="literal">iterator2</code>, so that
<code class="literal">iterator2</code>'s cached value is refreshed.
</p>
<p>
If <span class="bold"><strong>directdb_get</strong></span> is
<code class="literal">false</code>, a few of the tests in dbstl's test kit
will fail. This is because the above contrived case appears in
several of C++ STL tests. Consequently, the default value of the
<span class="bold"><strong>directdb_get</strong></span> parameter in the
<code class="methodname">container::begin()</code> methods is
<code class="literal">true</code>. If your use cases avoid such bizarre usage
of iterators, you can set it to <code class="literal">false</code>, which
makes the iterator read operation faster.
</p>
</div>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="chg_persistence"></a>Change persistence</h3>
</div>
</div>
</div>
<p>
If you modify the object to which an iterator refers by using one
of the following:
</p>
<pre class="programlisting">(*iterator).member_function_call()</pre>
<p>
or
</p>
<pre class="programlisting">(*iterator).data_member = new_value</pre>
<p>
then you should call
<code class="methodname">iterator->_DB_STL_StoreElement()</code> to
store the change. Otherwise the change is lost after the
iterator moves on to other elements.
</p>
<p>
If you are storing a sequence, and you modified some part of it, you
should also call
<code class="methodname">iterator->_DB_STL_StoreElement()</code>
before moving the iterator.
</p>
<p>
And in both cases, if <span class="bold"><strong>directdb_get</strong></span>
is <code class="literal">true</code> (this is the default value), you should
call <code class="methodname">_DB_STL_StoreElement()</code> after the
change and before the next iterator movement OR the next
dereferencing of the iterator by the star or arrow operators
(<code class="literal">iterator::operator*</code> or
<code class="literal">iterator::operator-></code>). Otherwise, you will
lose the change.
</p>
<p>
If you update the element by assigning to a dereferenced iterator like
this:
</p>
<pre class="programlisting">*iterator = new_element;</pre>
<p>
then you never have to call
<code class="methodname">_DB_STL_StoreElement()</code> because the change
is stored in the database automatically.
</p>
</div>
<div class="sect2" lang="en" xml:lang="en">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="obj_life_persistence"></a>Object life time and persistence </h3>
</div>
</div>
</div>
<p>
Dbstl is an interface to Berkeley DB, so it is used to store data
persistently. This is really a different purpose from that of
regular C++ STL. This difference in their goals has implications on
expected object lifetime: In standard STL, when you store an object
A of type ID into C++ stl vector V using V.push_back(A), if a
proper copy constructor is provided in A's class type, then the
copy of A (call it B) and everything in B, such as another object C
pointed to by B's data member B.c_ptr, will be stored in V and will
live as long as B is still in V and V is alive. B will be destroyed
when V is destroyed or B is erased from V.
</p>
<p>
This is not true for dbstl, which will copy A's data and store it
in the underlying database. The copy is by default a shallow copy,
but users can register their object marshalling and unmarshalling
functions using the <code class="classname">DbstlElemTraits</code> class
template. So if A is passed to a <code class="classname">db_vector</code>
container, <code class="literal">dv</code>, by using
<code class="literal">dv.push_back(A)</code>, then dbstl copies A's data
using the registered functions, and stores data into the underlying
database. Consequently, A will be valid, even if the container is
destroyed, because it is stored into the database.
</p>
<p>
If the copy is simply a shallow copy, and A is later destroyed, then
the pointer stored in the database will become invalid. The next time
we use the retrieved object, we will be using an invalid pointer, which
probably will result in errors. To avoid this, store the referred
object C rather than the pointer member A.c_ptr itself, by registering
the right marshalling/unmarshalling function with
<code class="classname">DbstlElemTraits</code>.
</p>
<p>
For example, consider the following example class declaration:
</p>
<pre class="programlisting">class ID
{
public:
string Name;
int Score;
}; </pre>
<p>
Here, the class ID has a data member <span class="bold"><strong>Name</strong></span>, which refers to a memory address of
the actual characters in the string. If we simply shallow copy an
object, <code class="literal">id</code>, of class ID to store it, then the
stored data, <code class="literal">idd</code>, is invalid when
<code class="literal">id</code> is destroyed. This is because
<code class="literal">idd</code> and <code class="literal">id</code> refer to a common
memory address which is the base address of the memory space storing
all characters in the string, and this memory space is released when
<code class="literal">id</code> is destroyed. So <code class="literal">idd</code> will be
referring to an invalid address. The next time we retrieve
<code class="literal">idd</code> and use it, there will probably be memory
corruption.
</p>
<p>
The way to store <code class="literal">id</code> is to write a marshal/unmarshal
function pair like this:
</p>
<pre class="programlisting">void copy_id(void *dest, const ID&elem)
{
memcpy(dest, &elem.Score, sizeof(elem.Score));
char *p = ((char *)dest) + sizeof(elem.Score);
strcpy(p, elem.Name.c_str());
}
void restore_id(ID& dest, const void *srcdata)
{
memcpy(&dest.Score, srcdata, sizeof(dest.Score));
const char *p = ((char *)srcdata) + sizeof(dest.Score);
dest.Name = p;
}
size_t size_id(const ID& elem)
{
return sizeof(elem.Score) + elem.Name.size() +
1;// store the '\0' char.
} </pre>
<p>
Then register the above functions before storing any instance of
<code class="classname">ID</code>:
</p>
<pre class="programlisting">DbstlElemTraits<ID>::instance()->set_copy_function(copy_id);
DbstlElemTraits<ID>::instance()->set_size_function(size_id);
DbstlElemTraits<ID>::instance()->set_restore_function(restore_id); </pre>
<p>
This way, the actual data of instances of ID are stored, and so the
data will persist even if the container itself is destroyed.
</p>
</div>
</div>
<div class="navfooter">
<hr />
<table width="100%" summary="Navigation footer">
<tr>
<td width="40%" align="left"><a accesskey="p" href="stl_complex_rw.html">Prev</a> </td>
<td width="20%" align="center">
<a accesskey="u" href="stl.html">Up</a>
</td>
<td width="40%" align="right"> <a accesskey="n" href="stl_container_specific.html">Next</a></td>
</tr>
<tr>
<td width="40%" align="left" valign="top">Store and Retrieve data or objects of complex types </td>
<td width="20%" align="center">
<a accesskey="h" href="index.html">Home</a>
</td>
<td width="40%" align="right" valign="top"> Dbstl container specific notes</td>
</tr>
</table>
</div>
</body>
</html>
