Repository files navigation

• Introduction
• Compilation and linking
• Getting started
  • Initialisation
  • Parameter registration
  • Function registration
  • Parsing command line options
  • Parsing configuration file
  • Result validation
  • Releasing memory
  • Error handling
  • Examples

This is a simple library written in C, for parsing command line options and plain configuration files. It can be used to retrieve variables and arrays into runtime memory, or call functions indicated by command line flags.

This library is compliant with the ISO C99 standard, and relies only on the C standard library. It is written by Cheng Zhao (赵成), and is distributed under theMIT license.

_[TOC]

Since this library is tiny and portable, it is recommended to compile the only two source files —libcfg.h andlibcfg.c — along with your own program. To this end, one only needs to include the headerlibcfg.h in the source file for parsing configurations:

#include"libcfg.h"

_[TOC]

Acfg_t type structure has to be initialised in the first place, for storing all the configuration information. This can be done using the functioncfg_init, e.g

cfg_t*cfg=cfg_init();

This function returns theNULL pointer on error.

_[TOC]

To retrieve variables and arrays from command line options or configuration files, they need to be registered as configuration parameters, which are represented by thecfg_param_t type structure:

typedefstruct {intopt;/* short command line option            */char*lopt;/* long command line option             */char*name;/* name of the parameter                */cfg_dtype_tdtype;/* data type of the parameter           */void*var;/* variable for the retrieved value     */}cfg_param_t;

The format of the attributes are:

• opt: a single case-sensitive letter or0;
• lopt: a string composed of characters with graphical representations (ensured byisgraph), or aNULL pointer;
• name: a string composed of case-sensitive letters, digits, and the underscore character, and starting with either a letter or an underscore;
• dtype: a pre-defined data type indicator;
• var: pointer to the address of the variable/array for holding the retrieved value, and no memory allocation is needed.

In particular, ifopt is set to0, orlopt is set toNULL, the value will not be retrieved from short or long command line options, respectively. For safety consideration, the length oflopt andname must be smaller than the pre-definedCFG_MAX_LOPT_LEN andCFG_MAX_NAME_LEN values respectively.

The supported data types and their indicators are listed below:

Once the configuration parameters are set, they can be registered using the function

intcfg_set_params(cfg_t*cfg,constcfg_param_t*params,constintnpar);

Here,cfg indicates the structure for storing all configuration information, or the entry for the registration.params denotes the address of the configuration parameter structure, andnpar indicates the number of parameters to be registered at once. This function returns0 on success, and a non-zero integer on error.

Note that thecfg_param_t type structure for parameter registration cannot be deconstructed until the command line options and configuration files containing this parameter are parsed (seeParsing command line options andParsing configuration file),

_[TOC]

Functions that can be called with command line flags must be declared with the prototype

voidfunc(void*args);

Similar to configuration parameters, functions can also be registered in the form of a structure:

typedefstruct {intopt;/* short command line option            */char*lopt;/* long command line option             */void (*func) (void*);/* pointer to the function              */void*args;/* pointer to the arguments             */}cfg_func_t;

Theopt andlopt variables are the short and long command line option for calling this function, respectively. And at least one of them has to be set, i.e., a case-sensitive letter foropt, or a string composed of graphical characters forlopt. Again, the length oflopt must be smaller than the pre-definedCFG_MAX_LOPT_LEN limit. The pointersfunc andargs are the address of the function to be called, and the corresponding arguments, respectively.

The functions can then be registered using

intcfg_set_funcs(cfg_t*cfg,constcfg_func_t*funcs,constintnfunc);

Here,cfg indicates the entry for the registration,funcs denotes the address to the structure holding the registration information of functions, andnfunc indicates the number of functions to be registered at once. This function returns0 on success, and a non-zero integer on error.

Note that thecfg_func_t type structure for function registration cannot be deconstructed until the command line options are parsed (seeParsing command line options).

As an example, a typical demand for calling functions via command line is to print the usage of a program, when there is the-h or--help flag. In this case, the help function and the corresponding structure can be defined as

voidhelp(void*arg) {printf("Display the help messages.\n");exit(0);}

constcfg_func_thelp_func= {'h',"help",help,NULL};

_[TOC]

Command line options are passed to themain function at program startup, as theargc andargv parameters. These two parameters can be used by the functioncfg_read_opts for parsing the options, and retrieving parameter values or calling functions. This function is defined as

intcfg_read_opts(cfg_t*cfg,constintargc,char*const*argv,constintpriority,int*optidx);

It returns0 on success, and a non-zero integer on error. And the arguments are:

• cfg: the structure for all configuration information;
• argc: the number of command line arguments, obtained frommain;
• argv: the command line argument list, obtained frommain;
• priority: the priority of values retrieved from command line, must be positive;
• optidx: the index of the argument list at which the parser is terminated.

In particular,priority decides which value to take if the parameter is defined in multiple sources, say, both in a configuration file and command line options. For instance, if a variable has already been set in a configuration file with a lowerpriority value than the one passed tocfg_read_opts, then its value will be overwritten by the one obtained from command line.

The supported formats of command line options are listed below:

Note that the- and= symbols in the formats are customisable. They are actually defined asCFG_CMD_FLAG andCFG_CMD_ASSIGN inlibcfg.h, respectively.

All command line arguments satisfying the above formats are interpreted as options, otherwise they are treated as values. And values can only be omitted for boolean type variables — which impliestrue — or function calls. If the value contains space or special characters that are reserved by the environment, then it should be enclosed by pairs of single or double quotation marks. Besides, values that may be confused with options (such as-x) are recommended to be passed with the assignment symbol=.

Furthermore, if the-- option is found, then the option scanning will be terminated, and the current index of the argument list is reported asoptidx. Therefore, when calling the program, non-option parameters should always be passed after all the options. And whenoptidx is equal toargc, it means that all command line arguments are parsed.

_[TOC]

Plain text files can be parsed using the function

intcfg_read_file(cfg_t*cfg,constchar*filename,constintpriority);

The argumentcfg indicates the structure for storing all configuration information,filename denotes the name of the input file, andpriority defines the priority of values read from this file. This function returns0 on success, and a non-zero integer on error.

By default the format of a valid configuration file has to be

# This is a comment.name_variable = value                # inline commentname_array = [ element1, element2 ]  # entry for an array

Here,name_variable,name_array, andname_long_array indicate the registered name of configuration parameters (seeParameter registration), andvalue,element1, andelement2 are the values to be loaded into memory.

In particular, scalar type definitions can be parsed as arrays with a single element. And by default array type definitions with multiple elements have to be enclosed by a pair of brackets[]. In addition, multiple-line definitions are only allowed for arrays, and the line break symbol\ can only be placed after the array element separator,. These symbols, including[,],\,,, as well as the comment indicator#, are customisable inlibcfg.h. And if a value or an element of an array contains special characters, the full value or element has to be enclosed by a pair of single or double quotation marks.

_[TOC]

The functionscfg_read_opts andcfg_read_file extract the parameter value from command line options and configuration files respectively. The value is then converted to the given data type, and passed to the address of the variable specified at registration.

To verify whether a variable or an array is set correctly, one can use the function

boolcfg_is_set(constcfg_t*cfg,constvoid*var);

It returnstrue if the variable or array values are set by the functionscfg_read_opts orcfg_read_file, andfalse if they are untouched. Here,var has to be the variable/array address specified at registration.

Moreover, the number of array elements read by the parser can be reported by the function

intcfg_get_size(constcfg_t*cfg,constvoid*var);

It returns0 if the array is not set. So it may not be necessary to verify arrays usingcfg_is_set. Note that the array is allocated with precisely the number of elements reported by this function, so the indices for accessing array elements must be smaller than this number.

Once the variable or array is verified successfully, it can then be used directly in the rest parts of the program.

_[TOC]

Once all the variables and arrays are retrieved and verified, thecfg_t type structure for storing all the configuration information can be deconstructed by the function

voidcfg_destroy(cfg_t*cfg);

After calling this function, the values of the variables and arrays are still accessible, but the size of arrays cannot be obtained using thecfg_get_size function anymore.

In addition, since the memory of arrays and string variables are allocated by this library, it is the user's responsibility to free them using the standardfree function. In particular, since string arrays are represented two-dimensional character arrays, the pointers to both the string array and its first element have to be freed, e.g.

char**str;/* declaration of the string array *//* parameter registration and retriving */free(*str);/* free the first element of the array */free(str);/* free the array itself */

_[TOC]

Errors can be caught by checking the return values of some of the functions, such ascfg_init,cfg_set_params,cfg_read_opts, etc. And once thecfg_init is executed successfully, error messages can be printed using the function

voidcfg_perror(constcfg_t*cfg,FILE*stream,constchar*msg);

It outputs the string indicated bymsg, followed by a colon and a space, and then followed by the error message produced by this library, as well as a newline character\n. The results are written tostream, which is typicallystderr.

Unexpected issues that are not critical enough to stop the program are treated as warnings. They cannot be handled by the return values of functions, but one can check warning messages using the function

voidcfg_pwarn(cfg_t*cfg,FILE*stream,constchar*msg);

This function is similar tocfg_perror. Note that there can be multiple warning messages, and once a warning message is printed, it is automatically removed from the message pool.

_[TOC]

An example for the usage of this library is provided in theexample folder.

It registers variables and arrays for all the supported data types, as well as two functions to be called via command line. Command line options and the configuration fileinput.conf are then parsed. The variables and arrays are printed if they are set correctly.

_[TOC]