<html><head>
<title>swim - MAKING INSTALLED SYSTEM DATABASES
</title>
<link rev=made href="mailto:mttrader@access.mountain.net">
</head><body>
<h1>
swim - chapter 5<br>
MAKING INSTALLED SYSTEM DATABASES
</h1>
<hr>
<h2><A name="s-initdb">
5.1 Initial database making, and Rebuilding for an
Installed system.
</A></h2>
<pre>
usage: swim --initdb
swim --rebuilddb
options: [--dbpath <dir>] [--root <dir>] [--lowmem]
[--split_data <lines>]
</pre>
<P>
An <em>installed Debian distribution</em> is one in which packages are
installed using <strong>dpkg</strong> or some front-end to
<strong>dpkg</strong> like <strong>apt</strong> or
<strong>dselect</strong>; <strong>swim</strong> supports installation
through <strong>apt</strong>. These major modes are for a computer with an
<em>installed Debian distribution</em> and make the databases which allow
querying and searching capabilities for the installed distribution.
<P>
<strong>--initdb</strong> is run when the databases do not exist yet,
<strong>--rebuilddb</strong> is run if the databases have become corrupt,
or you want to rebuild the databases instead of updating them.
<P>
<strong>--dbpath</strong> can be specified as an alternative location for
where the databases will be made. The default location is
``<em>/var/lib/dpkg</em>''. An argument like ``<em>/otherstuff</em>''
could be provided, and then the databases would be made here instead.
<P>
<strong>--root</strong> allows a database to be made for a Debian distribution installed on a
different partition. If the distribution is mounted on
<em>/New_Debian</em>, ``<em>/New_Debian</em>'' would be the argument to root. The databases would be made for the
Debian distribution installed on the ``<em>/New_Debian</em>'' partition.
<P>
<strong>--dbpath</strong> and <strong>--root</strong> can be used
together. Given the previous two examples, the databases would be made on
``<em>/New_Debian/otherstuff</em>'', assuming
``<em>/New_Debian/otherstuff</em>'' actually existed.
<P>
<strong>--lowmem</strong> uses a method which uses a small amount of
memory to make the databases. By default <strong>--initdb</strong> and
<strong>--rebuilddb</strong> use a method which fully takes advantage of
memory, this is a good thing, because it means the databases are made in a
quicker manner. On a computer with a K6-200 CPU, 64MB of memory, and 1500
installed packages, the databases can be made in 4.5 minutes using the
default method, and 11 minutes using the low memory method. The high
memory method is the default because in general the size of a distribution
is related to how much resources a computer has, and probably a large
installation is unusual. If you get an ``out of memory'' when you use the
default method, or if your system is overloaded to begin with, the
<strong>--lowmem method</strong> is the prefered way.
<P>
<strong>--split_data</strong> determines the size of the files in the
temporary directory used to contruct the database. The default is 25000
lines per file. If you are using the <strong>--lowmem method</strong> you
may want to provide a different argument to <strong>--split_data</strong>,
like ``<strong>--split_data 10000</strong>''. This is a subject of
experimentation.
<P>
<hr>
<h2><A name="s5.2">
5.2 UPDATING
</A></h2>
usage: <strong>swim --db</strong>
<P>
options: <strong>[--dbpath</strong> <<strong>dir</strong>><strong>]
[--root</strong> <<strong>dir</strong>><strong>] [--check]</strong>
<P>
<strong>--db</strong> allows you to update the databases by hand when
packages have been removed, added, or changed. swim will automatically run
<strong>--db</strong> under certain conditions.
<P>
<strong>--check</strong> prints out the changes to STDERR, and the
total to STDOUT without proceeding with the update.
<P>
See <A href="#s-initdb">Initial database making, and Rebuilding for an
Installed system.
, section 5.1</A> for options <strong>--dbpath</strong> and
<strong>--root</strong>.
<P>
<hr>
<h2><A name="s-searchy">
5.3 REBUILDING THE SEARCH
</A></h2>
usage: <strong>swim --rebuildflatdb</strong>
<P>
options: <strong>[--dbpath</strong> <<strong>dir</strong>><strong>]
[--root</strong> <<strong>dir</strong>><strong>]</strong>
<P>
swim makes the flat databases <em>searchindex.deb</em> and
<em>dirindex.deb</em> for doing <em>powersearches</em>. Instead of
rebuilding these databases everytime <strong>--db</strong> is run, new
information is just appended to these databases, and old information is
kept. Generally, this is not a problem because only files and directories
which the other databases actually know something about will be refered
to. But in a situation where a file has changed into a directory, the
<em>powersearch</em> may not work properly, because the old file name
remains in <em>searchindex.deb</em>, and the new directory name is now in
<em>dirindex.deb</em> directory. In general, it takes a lot of changes to
the installed system before it is really becomes necessary to rebuild the
flat databases. This process takes less than a minute on a K6-200 with
1500 packages.
<P>
See <A href="#s-initdb">Initial database making, and Rebuilding for an
Installed system.
, section 5.1</A> for options <strong>--dbpath</strong> and
<strong>--root</strong>.
<hr>
<h2><A name="s5.4">
5.4 FILES
</A></h2>
Databases which are made:
<P>
<pre> packages.deb
fileindex.deb
statusindex.deb
groupindex.deb
searchindex.deb
dirindex.deb</pre>
<hr>
swim
- <A href="index.html#copyright">
Copyright � 1999 Jonathan D. Rosenbaum
</A>
<br>
<A href="index.html#toc">Contents</A>; <A href="ch-important.html">next</A>; <A href="ch-history.html">back</A>.
<br>
<address>15 June 1999<br>
Jonathan D. Rosenbaum<A href="mailto:mttrader@access.mountain.net">mttrader@access.mountain.net</A></address>
</body></html>