Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

Fetching latest commit…

Cannot retrieve the latest commit at this time

..
Failed to load latest commit information.
Makefile.am
Makefile.in
Makefile.mak
README.scconf
internal.h
lex-parse.l
parse.c
scconf.c
scconf.h
sclex.c
test-conf.c
write.c

README.scconf

A short introduction to scconf as an API and a file format
==========================================================
written by Jamie Honan <jhonan@optusnet.com.au>

The scconf system is a small system library for handling
scconf files. Why should anyone care about scconf format?
It is a handy format for short pieces of structured data.

Handy because:
  - it is readable, which makes support easy
  - it is easy to parse and write
  - it is extensible, you can add fields without breaking things

It isn't
  - XML, so it doesn't need xml parsing
  - suitable for large amounts of data, like a database or text files

It doesn't have
  - anything else but data. No locking, no threads etc.

It has heirarchical data blocks, it has lists.

Similar, but different:
  - .ini files. scconf is block structured, has lists and arrays
  - xml. xml is more complete, but requires a lot of overhead
  - sexp. sexp resembles lisp with it's use of parenthesis. sexp
    has modes for binary. scconf really doesn't have binary
  - yaml. yaml is larger

What does it look like?
=======================

Like this:

transport_stream {
        id = 0x0009;
        original_network_id = 0x1000;
        sat_tuning_info {
                frequency = 12278000;
                symbol_rate = 30000000;
                polarization = 0;
        }
        service {
                id = 0x0064;
                pmt_pid = 0x0101;
                type = 144;
                name = "aGuide";
                provider_name = "A";
        }
        service {
                id = 0x238D;
                pmt_pid = 0x0623;
                type = 144;
                name = "aCar";
                provider_name = "A";
        }
}

Why doesn't it have X, why don't you use XML?
=============================================

Maybe it should. Maybe XML is the answer. Maybe a database
is more appropriate. It's all a trade-off. You choose.


API
===

There are four useful structures. scconf_block, scconf_list,
scconf_item, and a scconf_context.

A context is similar to a file, except in memory. Within
a context there is a root block. Within each block
there are one or more items. Items can be sub-blocks, lists, or
comments. Every item can have a name, or key.

A list can have one or more values; boolean, integer or string.

A context contains a root block, which contains one or more blocks.

A block is :

key [[,] name [[,] name ... ] ] {
	block_contents
}

block_contents is one or more block_items

block_items is one of

  # comment string \n
or
  key [[,] name [[,] name ... ] ] = value [[,] value ... ]];
or
  block

Initialising and file handling
==============================

Allocate scconf_context
The filename can be NULL. The file is not read at this point,
but in the function scconf_parse.

scconf_context *scconf_new(const char *filename);


Free scconf_context

void scconf_free(scconf_context * config);


Parse configuration
Returns 1 = ok, 0 = error, -1 = error opening config file

int scconf_parse(scconf_context * config);


 Write config to a file
 If the filename is NULL, use the config->filename
 Returns 0 = ok, else = errno

int scconf_write(scconf_context * config, const char *filename);


Finding items and blocks
========================

 Find a block by key
 If the block is NULL, the root block is used

const scconf_block *scconf_find_block(const scconf_context * config,
	const scconf_block * block, const char *item_name);

This finds a block in the given context. This function doesn't descend
the heirarchy, it only finds blocks in the top level of either
the context (the root block) or of the block given in the block
paramter (if not NULL).

The block pointer returned points to data held by the context, hence
the const qualifier.


 Find blocks by key and possibly name
 If the block is NULL, the root block is used
 The key can be used to specify what the blocks first name should be

scconf_block **scconf_find_blocks(const scconf_context * config,
	const scconf_block * block, const char *item_name, const char *key);

This function is similar to scconf_find_block above, except that an array
of pointers to matched blocks is returned. Each pointer points
to data held by the context. The last entry in the returned table
is the null pointer.

The table should be freed after use, but the individual pointers to blocks
point to data held by the context.

The key values for blocks is matched. If name is not NULL, the block
name must also match.


 Get a list of values for option

const scconf_list *scconf_find_list(const scconf_block * block, const char *option);

Find an item that has a value (i.e. is not a block nor a comment), and
return the values for that item as a list.

The list is held in memory owned by the context.


 Return the first string of the option
 If no option found, return def

const char *scconf_get_str(const scconf_block * block, const char *option, const char *def);

This is similar to scconf_find_list, but instead of returning the whole
list, just return the first value, as a string. If this is not possible,
return the default value.

Again the value returned is either a pointer the default value or to
memory held by the context.


 Return the first value of the option as integer
 If no option found, return def

int scconf_get_int(const scconf_block * block, const char *option, int def);

This is similar to scconf_get_str, but an integer value is returned.


 Return the first value of the option as boolean
 If no option found, return def

int scconf_get_bool(const scconf_block * block, const char *option, int def);

This completes the types that can be returned by a find.


For parsing blocks and items
============================

A table of scconf_entry values is used, terminated by a NULL name value.

This table is passed to the routine scconf_parse_entries. This
function walks the current context or block, and adds the data
to the scconf_entry table entries.
Sub-blocks can be walked, using SCCONF_BLOCK, and callbacks can be issued
using SCCONF_CALLBACK.

This is a handy method for accessing scconf data from within a program.

typedef struct _scconf_entry {
	const char *name;
			 * Look for blocks with this key, or check if this
			 * block has an item with this key. Run the block
			 * or blocks found against the rest of this entry
			 * Stop after the first one, unless
			 * SCCONF_ALL_BLOCKS is set in flags
	unsigned int type;
			 * SCCONF_CALLBACK
			 * 	parm contains a function ptr of type
			 *      int (*callback)(scconf_context* context,
			 *      	scconf_block* block,
			 *      	scconf_entry* entry,
			 *      	int depth);
			 *      run the callback with the block found
			 *
			 * SCCONF_BLOCK
			 *      param contains a pointer to another entry table
			 *      use the found block against every entry
			 *	in the pointed entry table
			 *
			 * SCCONF_LIST
			 * SCCONF_BOOLEAN
		  	 * SCCONF_INTEGER
			 * SCCONF_STRING
			 *      find the entry with the key given in name in
			 *      the found block. Return the value found
			 *      to parm as follows:
			 *    SCCONF_INTEGER:
			 *		if parm not NULL, then
			 *		points to integer location to put
			 *		the value
			 *    SCCONF_BOOLEAN:
			 *		if parm not NULL, then
			 *		points to integer location to put
			 *		the value
			 *    SCCONF_STRING:
			 *		if parm not NULL, then
			 *		  if flag bit SCCONF_ALLOC not set
			 *		     then parm points to a buffer
			 *		  else
			 *		     parm points to a pointer where
			 *		     the pointer to an allocated
			 *		     buffer should be stored.
			 *		     if arg is not NULL, points
			 *		     to a location where the buffer
			 *		     length (size_t) is to be stored
			 *    SCCONF_LIST:
			 *		if parm not NULL, then
			 *		  if flag bit SCCONF_ALLOC not set
			 *		     then parm points to a location
			 *		     where a pointer to the list
			 *		     can be stored
			 *		  else
			 *		     then parm points to a location
			 *		     where a pointer to a copy of list
			 *		     can be stored
			 *
			 *
	unsigned int flags;
			 * SCCONF_PRESENT
			 *       This bit is or'ed in when found
			 * SCCONF_MANDATORY
			 *       If not found, this is a fault
			 * SCCONF_ALLOC
			 *       C.f. type above
			 * SCCONF_ALL_BLOCKS
			 *       C.f. name above
			 * SCCONF_VERBOSE
			 *       For debugging
	void *parm;
	void *arg;
} scconf_entry;


For adding blocks and items
===========================

A table of scconf_entry values is used, terminated by a NULL name value.

This table is passed to the routine scconf_write_entries. This
function adds the scconf_entry table entries to the current block.
Sub-blocks can be added, and callbacks can be issued.

This is a handy method for adding scconf data from within a program.

typedef struct _scconf_entry {
	const char *name;
			 * key value for blocks and items *
	unsigned int type;
			 * SCCONF_CALLBACK
			 * 	parm contains a function ptr of type
			 *      int (*callback)(scconf_context* context,
			 *      	scconf_block* block,
			 *      	scconf_entry* entry,
			 *      	int depth);
			 *
			 * SCCONF_BLOCK
			 *      param contains a pointer to another entry table
			 *      the entry table is added as a block to the
			 *      current block, with name as the key, and
			 *      arg is a list of names
			 *
			 * SCCONF_LIST
			 * SCCONF_BOOLEAN
		  	 * SCCONF_INTEGER
			 * SCCONF_STRING
			 *      these add key=value pairs to the current
			 *      block. The value is in parm.
			 *
	unsigned int flags;
			 * SCCONF_PRESENT
			 *       This bit is or'ed in when item added
	void *parm;
	void *arg;
} scconf_entry;
Something went wrong with that request. Please try again.