<?xml version="1.0" standalone="no"?>
<!DOCTYPE article PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN"
"http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd"
[
<!ENTITY listing1 SYSTEM "listing1.xml">
<!ENTITY listing2 SYSTEM "listing2.xml">
<!ENTITY listing3 SYSTEM "listing3.xml">
<!ENTITY listing4 SYSTEM "listing4.xml">
<!ENTITY listing5 SYSTEM "listing5.xml">
<!ENTITY listing6 SYSTEM "listing6.xml">
<!ENTITY listing7 SYSTEM "listing7.xml">
<!ENTITY listing8 SYSTEM "listing8.xml">
]>
<article>
<articleinfo>
<title>libConfuse tutorial</title>
<author><firstname>Martin</firstname> <surname>Hedenfalk</surname></author>
</articleinfo>
<sect1>
<title>Introducing libConfuse in an existing program</title>
<para>Consider this simple program:</para>
&listing1;
<para>
Simple enough, but we want to extend the program so we can greet
others. Maybe we don't want to greet the whole world, just our
neighbour. We use libConfuse to let the user decide whom to greet.
</para>
&listing2;
<para>
All programs using libConfuse must first include the
<filename>confuse.h</filename> header file. This is done on line
2.
</para>
<para>
On line 6 - 10, the options that should be recognized are defined in an
array of cfg_opt_t structs. This is passed to the
<function>cfg_init</function> function on line 13. The resulting
<structname>cfg_t</structname> context is used by
<function>cfg_parse()</function>, which reads the configuration file
"hello.conf". When reading the configuration file, only options defined in
the array of options passed to <function>cfg_init()</function> are
recognized.
</para>
<para>
The friendly greeting is now replaced with a parameter read from the
configuration file. The value of the <varname>target</varname> option is retrieved with
<function>cfg_getstr(cfg, "target")</function>.
</para>
<para>
Lets take a look at the configuration file hello.conf:
</para>
<programlisting>
# this is the configuration file for the hello program
target = "Neighbour"
</programlisting>
<para>
Here, the target option is set to the string value "Neighbour".
What if the configuration file was empty or didn't exist? Then the
default value for the <varname>target</varname> option would be
used. When we initialized our options, the second parameter to the
<function>CFG_STR()</function> macro specified the default value.
Thus, if no <varname>target</varname> option was specified in the
configuration file, the hello program would have printed the
standard greeting "Hello, World".
</para>
<sect2>
<title>Environment variables in values</title>
<para>
What else can we do in the configuration file? We can set the value to an
environment variable:
</para>
<programlisting>
target = ${USER}
</programlisting>
<para>
This results in the hello program greeting the user who runs it. On some
systems, the USER variable might not be available, so we want to specify a
default value in those cases:
</para>
<programlisting>
target = ${USER:-User}
</programlisting>
<para>
Now, if the USER environment variable is unset, the string "User" will be
used instead.
</para>
</sect2>
</sect1>
<sect1>
<title>Other types of options</title>
<para>
Of course, not only strings can be specified in the configuration file.
libConfuse can parse strings, integers, booleans and floating point values.
These are the fundamental values, and they are all also available as lists.
We'll talk more about lists in the next chapter.
</para>
<para>
The macros used to initialize a string, integer, boolean and a
float is, respectively, <function>CFG_STR()</function>,
<function>CFG_INT()</function>, <function>CFG_BOOL()</function>,
<function>CFG_FLOAT()</function> and
<function>CFG_PTR()</function>. All macros take three parameters:
the name of the option, a default value and flags. To retrieve the
values, use <function>cfg_getstr()</function>,
<function>cfg_getint()</function>,
<function>cfg_getbool()</function>,
<function>cfg_getfloat()</function> or
<function>cfg_getptr()</function>, respectively.
</para>
<para>
Let's introduce an integer option that tells us how many times to print the
greeting:
</para>
&listing3;
<para>
Here we have used the <function>CFG_INT()</function> macro to
initialize an integer option named "repeat". The default value is 1
as in the standard greeting. The value is retrieved with
<function>cfg_getint()</function>.
</para>
<para id="negative-repeat-problem">
But, wait a moment, what if the user specified a negative value for
"repeat"? Or a too large positive value? libConfuse can handle
that with a so-called validating callback. We'll come back to this
problem later on, but we will first take a look at lists.
</para>
</sect1>
<sect1>
<title>Introducing lists</title>
<para>
That was easy. Now let's extend the program a bit so we can greet more than one
"target". We'd like to be able to specify a list of targets to greet.
</para>
<para>
The list versions of the initialization macros are named
<function>CFG_STR_LIST()</function>,
<function>CFG_INT_LIST()</function>,
<function>CFG_BOOL_LIST()</function> and
<function>CFG_FLOAT_LIST()</function>. They take the same
parameters as the non-list versions, except the default value must
be a string surrounded by curly braces.
</para>
<para>
The modified program is shown below:
</para>
&listing4;
<para>
Three things are a bit different here. First, the macro to
initialize the "targets" option is
<function>CFG_STR_LIST()</function>. This tells libConfuse that
"targets" is a list of strings. Second, the default value in the
second parameter is surrounded by curly braces. This is needed to
indicate to libConfuse where the list of values ends.
</para>
<para>
The third change is in the printing of the greeting. First we print
the "Hello" string. Then we loop through all values found for the
"targets" option. The number of values is retrieved with the
<function>cfg_size()</function> function. The string values are
then retrieved with <function>cfg_getnstr()</function>, which is an
indexed version of <function>cfg_getstr()</function>. In fact,
<function>cfg_getstr()</function> is equivalent to
<function>cfg_getnstr()</function> with an index of zero.
</para>
<para>
In the configuration file hello.conf, we can now specify a list of targets to
greet:
</para>
<programlisting>
# this is the configuration file for the hello program
targets = {"Life", "Universe", "Everything"}
repeat = 1
</programlisting>
<para>
The output of the hello program, run with the above configuration file, is:
"Hello, Life, Universe, Everything!"
</para>
<para>
Again, if no targets were configured, the greeting would have been the standard
"Hello, World!".
</para>
</sect1>
<sect1>
<title>Using sections</title>
<para>
So far, we have only use a flat configuration file. libConfuse can also handle
sections to build a hierarchy of options. Sections can be used to group options
in logical blocks, and those blocks can (optionally) be specified multiple
times.
</para>
<para>
Sections are initialized with the <function>CFG_SEC()</function> macro. It also takes three
parameters: the name of the option, an array of options allowed in the section
and flags.
</para>
<para>
We'll extend the, now rather complex, hello program so we can do other kinds of
greetings, not just "Hello". Each greeting will have its own settings for
targets and repeat.
</para>
&listing5;
<para>
We have renamed the option array from "opts" to "greet_opts", and introduced a
new "opts" array that only has one option: a "greeting" section.
The second parameter of the <function>CFG_SEC()</function> macro
points to the old greeting options "targets" and "repeat".
</para>
<para>
We have also used a couple of flags to alter the behaviour of the
section: CFGF_TITLE means that a greeting section should have a
title and the CFGF_MULTI flag tells libConfuse that this section
may be specified multiple times in the configuration file. The
title of a section is retrieved with the
<function>cfg_title()</function> function.
</para>
<para>
The outmost loop (with index j) now loops through all given
sections in the configuration file. We retrieve a section with a
<function>cfg_getnsec()</function> call. The value returned is a
pointer to a cfg_t struct, the same type as returned by
<function>cfg_init()</function>. Thus we can use the ordinary value
retrieval functions <function>cfg_getstr()</function>,
<function>cfg_getint()</function> and so on to retrieve values of
options inside the section.
</para>
<para>
Ok, so how does the configuration file look like for this setup?
</para>
<programlisting>
# this is the configuration file for the hello program
greeting Hello
{
targets = {"Life", "Universe", "Everything"}
repeat = 1
}
greeting Bye
{
targets = {Adams}
repeat = 1
}
</programlisting>
<para>
The program will loop through the sections in the order specified
in the configuration file. First it will find the "Hello" section.
It prints the title of the section, "Hello", retrieved with
<function>cfg_title()</function>. Then the targets are printed just
as in the previous examples, but this time the values are retrieved
from the cfg_greet section. Next, the section titled "Bye" is
found, and the values are retrieved from that section.
</para>
<para>
When run, the program produces the following:
</para>
<programlisting>
$ ./listing5
Hello, Life, Universe, Everything!
Bye, Adams!
$
</programlisting>
</sect1>
<sect1>
<title>Parsing from internal buffers</title>
<para>
So far, we have only parsed configuration data from files.
libConfuse can also parse buffers, or in-memory character
strings. We will use this to fix a problem in the previous code.
</para>
<para>
The problem is that without a configuration file, the hello program
will not print anything. We want it to at least print the standard
greeting "Hello, World!" if no configuration file is available.
</para>
<para>
We can't have a default value for a section that can be specified
multiple times (ie, a section with the CFGF_MULTI flag set).
Instead we will parse a default configuration string if no section
has been parsed:
</para>
&listing6;
<para>
Only the changes from the previous code is shown here. We check if
the size of the "greeting" section is zero (ie, no section has been
defined). In that case we call <function>cfg_parse_buf()</function>
to parse a default in-memory string "greeting Hello {}". This
string defines a greeting section with title Hello, but without any
sub-options. This way we rely on the default values of the
(sub-)options "targets" and "repeat".
</para>
<para>
When this program is run, it issues the well-known standard greeting
"Hello, World!" if no configuration file is present.
</para>
</sect1>
<sect1>
<title>Validating callback functions</title>
<para>
Remember the problem about a negative or too large "repeat" value
in <xref linkend="negative-repeat-problem"/>? The code that prints
the greeting has those lines:
</para>
<programlisting>
...
repeat = cfg_getint(cfg_greet, "repeat");
while(repeat--)
...
</programlisting>
<para>
The repeat variable is defined as an int, a signed integer. If the user
specified a negative repeat value in the configuration file, this code
would continue to decrease the repeat variable until it eventually
underflowed.
</para>
<para>
We'll fix this by not allowing a negative value in the configuration
file. Of course we could first just check if the value is negative
and then abort, using <function>cfg_getint()</function> and a test.
But we will use a validating callback function instead. This way
<function>cfg_parse()</function> will return an error directly when
parsing the file, additionally indicating on which line the error
is.
</para>
<para>
A validating callback function is defined as:
</para>
<programlisting>
typedef int (*cfg_validate_callback_t)(cfg_t *cfg, cfg_opt_t *opt);
</programlisting>
<para>
This function takes two arguments: the section and the option. It
should return 0 on success (ie, the value validated ok). All other
values indicates an error, and the parsing is aborted. The callback
function should notify the error itself, for example by calling
<function>cfg_error()</function>.
</para>
<para>
Here is the code for the callback function:
</para>
&listing7;
<para>
Only the last value is validated, because libConfuse will call this
function once for every value corresponding to the option. Since
the "repeat" option is not a list, we could instead have used
<function>cfg_opt_getint(opt)</function> to retrieve the only
value. However, if we later want to use this callback to validate
an integer list, it is already lists-aware.
</para>
<sect2>
<title>Installing the callback</title>
<para>
The validating callback is installed with
<function>cfg_set_validate_func()</function>. It is called with
a string specifying which option is affected, and a pointer to
the callback function. To specify an option in a subsection,
the section and the option must be separated with a vertical
bar ("|").
</para>
<para>
We're now also looking at the return code from
<function>cfg_parse()</function> to verify that the parsing was
successful. The complete program is now:
</para>
&listing8;
</sect2>
</sect1>
<sect1>
<title>Value parsing callback</title>
<para>
A value parsing callback is another kind of callback function
available in libConfuse. This function is used to map a string into
some other value. One example is to extend a boolean option
to accept the values "yes", "no" and "ask" (or perhaps "true",
"false" and "maybe"). Those values should be mapped to the integers
1, 2 and 3.
</para>
<programlisting>
typedef int (*cfg_callback_t)(cfg_t *cfg, cfg_opt_t *opt,
const char *value, void *result);
</programlisting>
<para>
</para>
</sect1>
<sect1>
<title>Functions</title>
<para>
libConfuse supports functions to parse options that does not fit
well in the general syntax. Functions can be called with a variable
number of arguments. No data from the function or any arguments are
stored by libConfuse after the function has run. It is up to the caller
to process and/or save the data.
</para>
<para>
A function is defined with a <function>CFG_FUNC</function> macro.
It takes two arguments: the name of the function and a function
callback. The callback is defined as:
</para>
<programlisting>
typedef int (*cfg_func_t)(cfg_t *cfg, cfg_opt_t *opt,
int argc, const char **argv);
</programlisting>
<para>
</para>
<sect2>
<title>Predefined functions</title>
<para>
Currently there is only one pre-defined function:
<function>cfg_include()</function>. This function includes
another configuration file. Configuration data is immediately
read from the included file, and is returned to the position
right after the include() statement upon end of file.
</para>
<para>
To use this function, include a <function>CFG_FUNC()</function>
entry in your options:
</para>
<programlisting>
cfg_opt_t opts[] = {
CFG_FUNC("include", cfg_include),
CFG_END()
};
</programlisting>
<para>
In the configuration file, it is used in the following way:
</para>
<programlisting>
include("included.conf")
</programlisting>
</sect2>
</sect1>
<sect1>
<title>Saving configuration files</title>
<para>
</para>
<sect2>
<title>Altering the printing of certain options</title>
<para>
</para>
</sect2>
</sect1>
</article>
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>