NAME¶

nval - the ksh name/value library

SYNOPSIS¶

#include	<nval.h>
libshell.a	-lshell

Namval_t;
Namfun_t;
Namarr_t;
Namdisc_t;

Namval_t 	*nv_open(const char *name, Dt_t *dict, int flags);
Namval_t	*nv_create(const char *name,  Dt_t *dict, int flags, Namfun_t *fp);
Namval_t	*nv_namptr(void *ptr, int indx); 
void	nv_close(Namval_t *np);
void	nv_delete(Namval_t *np, Dt_t *dict, int nofree);

char	*nv_getval(Namval_t *np);
Sfdouble_t	nv_getnum(Namval_t *np);
char	*nv_name(Namval_t *np);
void	nv_putval(Namval_t *np, const char *val, int flags);
void	nv_unset(Namval_t *np, int flags);
int	nv_clone(Namval_t *src, Namval_t *dest, int flags);

int	nv_isnull(Namval_t *np);
int	nv_setsize(Namval_t *np, int size);
int	nv_size(Namval_t *np);
unsigned nv_isattr(Namval_t *np, unsigned flags);
Namfun_t	*nv_isvtree(Namval_t *np);
unsigned nv_onattr(Namval_t *np, unsigned flags);
unsigned nv_offattr(Namval_t *np, unsigned flags);
void	nv_newattr(Namval_t *np, unsigned flags, int size);

unsigned nv_isarray(Namval_t *np);
Namarr_t	*nv_setarray(Namval_t *np,void*(*fun)(Namval_t*,const char*,int));
Namarr_t	*nv_arrayptr(Namval_t *np);
Namval_t	*nv_putsub(Namval_t *np, char *name, long mode);
Namval_t	*nv_opensub(Namval_t *np);
void	nv_setvec(Namval_t *np, int append, int argc, char *argv[]);
char	*nv_getsub(Namval_t *np);
int	nv_nextsub(Namval_t *np);
int	nv_aindex(Namval_t *np);

Namfun_t	*nv_disc(Namval_t *np, Namfun_t *fp, int flags);
Namfun_t	*nv_hasdisc(Namval_t *np, const Namdisc_t *dp);
char	*nv_getv(Namval_t *np, Namfun_t *fp);
Sfdouble_t	nv_getn(Namval_t *np, Namfun_t *fp);
void	nv_putv(Namval_t *np, const char *val, int flags, Namfun_t *fp);
char	*nv_setdisc(Namval_t *np, const char *a, Namval_t *f, Namfun_t *fp);
char	*nv_adddisc(Namval_t *np, const char **names);
const Namdisc_t	*nv_discfun(int which);

Namval_t	*nv_type(Namval_t  *np);
int	*nv_settype(Namval_t  *np, Namval_t *tp, int flags);
Namval_t	*nv_mkinttype(char *name, size_t sz, int us, const char *str, Namdisc_t *dp);
void	nv_addtype(Namval_t *np, const char *str, Optdisc_t* *op, size_t sz);

int	nv_scan(Dt_t *dict, void(*fn)(Namval_t*,void*), void *data, int mask, int flags);
Dt_t	*nv_dict(Namval_t *np);
void	nv_setvtree(Namval_t *np);
void	nv_setref(Namval_t *np, Dt_t *dp, int flags);
Namval_t	*nv_lastdict(void);

DESCRIPTION¶

Nval is a library of functions for interacting with name-value pairs as used in ksh. It is built on top the container dictionary type library facility in libcdt. (See cdt(3)). Each name-value pair is represented by a type named Namval_t. A Namval_t contains the name, value and attributes of a variable. Some attributes can have an associated number that represents the field width, arithmetic base, or precision. Additionally, each name-value pair can be associated with one or more processing disciplines that affect its behavior.

The function nv_open() returns a pointer to a name-value pair corresponding to the given name. It can also assign a value and give attributes to a name-value pair. The argument dict defines the dictionary to search. A NULL value causes the shell global variable dictionary to be searched.

The flags argument consists of the bitwise-or of zero or more of the attributes listed later and zero or more of the following:

NV_VARNAME: An invalid variable name causes an error.

NV_IDENTIFIER: A variable name that is not an identifier causes an error.

NV_ASSIGN: The name argument can contain an assignment.

NV_NOARRAY: The name argument cannot contain a subscript.

NV_NOREF: Do not follow references when finding the name-value pair.

NV_NOADD: The name-value pair will not be added if it doesn't exist. Instead, a NULL pointer will be returned.

NV_NOSCOPE: Only the top level scope is used.

NV_NOFAIL: Just return NULL when an error occurs. By default an error message is displayed and the current command is aborted.

If a name-value pair by this name does not already exist, it is created unless flags contains the NV_NOADD flag. If NV_VARNAME, NV_IDENTIFIER and NV_ASSIGN are all not specified, then no validity check is performed on the name argument and no further processing is performed. Otherwise, if NV_ASSIGN is specified, then the characters up to the first = or += are used to find the name-value pair, and the characters after the = are used to define the value that will be assigned to this name-value pair. If name does not contain an =, than no assignment will be made. If the first identifier in name is a reference and is not preceded by a ., it will be replaced by the value of the reference to find the name of a variable. Unless flags contains the NV_NOREF flag, if the name-value pair give by name has the NV_REF attribute, it will be replaced by the variable whose name is the value of this name-value pair. If NV_ASSIGN is set in the flags argument, the name variable can contain an = and a value that will be assigned to the name-value pair. Any attributes appearing in the flags argument will be applied to the name-value pair after any value is assigned.

It is possible for an application to create additional dictionaries with the cdt library and associate them with name-value pairs. The nv_dict() function returns the dictionary associated with the specified name-value pair, or if no dictionary was specified, NULL is returned. The nv_lastdict() function returns a pointer the the name-value pair that contains the last dictionary searched on the previous nv_open().

Name-value pairs can also be allocated without belonging to a dictionary. They will typically be looked up by a create discipline associated with a parent node. In this case the node size will be NV_MINSZ and n nodes can be allocated vial calloc(5NV_MINSZ,n)(3). The nv_namptr function can be used on the pointer returned by calloc along with the element number to return the corresponding node. Each of these nodes must be given the NV_MINIMAL attributes.

The nv_close() indicates that the pointer returned by nv_open() or nv_opensub() will not be referenced again. If the name-value pair is unset, and not referenced elsewhere, the name-value pair may be freed.

The nv_delete() function will remove the node np from the dictionary dict. Unless nofree is non-zero, the node np will also be freed.

The nv_name() function returns the name of the given name-value pair np. The nv_setsize() function returns the size of the field for justified variables, the arithmetic base for integer variables, and the precision or number of places after the decimal point for floating point variables. If size is greater than or equal to zero, the current size is changed to this value. The nv_size() function is equivalent to nv_setsize() with the second argument negative.

The nv_getval() function returns the value of the given name-value pair as a string. A NULL return value indicates that the name-value pair is unset. The nv_getnum() function returns the value of the given name-value pair as a double precision number using the Sfio library (See Sfio(3)) type Sfdouble_t. For name-value pairs without the NV_INTEGER attribute, the string value is evaluated as an arithmetic expression to arrive at a numerical value.

The nv_putval() function is used to assign a value to the name-value pair np. The flags argument consists zero or more of the bitwise-or of NV_LONG, NV_SHORT, NV_DOUBLE, NV_INTEGER, NV_RDONLY, NV_REF, NV_BINARY, and NV_NOFREE. The presence of NV_RDONLY allows the assignment to occur even if the name-value pair has the NV_RDONLY attribute. The presence of NV_INTEGER indicates that the value argument is actually a pointer to a numerical type. By default this type is long, but can be modified with NV_LONG, NV_SHORT, and NV_DOUBLE to represent long long, short, double, long double, and float. The presence of NV_REF indicates that the value argument is actually a pointer to a name-value pair and np should become a reference to this name-value pair. If NV_NOFREE is specified, value itself becomes the value of the name-value pair np. Otherwise, a copy of the value is stored as the value for np.

The nv_unset() function clears out the value and attributes of the given name-value function but does not free the name-value pair. If called from the putval discipline function, use the flags argument as the flags to nv_unset(). Otherwise, use 0.

The following attributes can be associated with a name-value pair:

NV_EXPORT: The export attribute.

NV_RDONLY: The readonly attribute.

NV_LTOU: Lower case characters are converted to upper case characters.

NV_UTOL: Upper case characters are converted to lower case characters.

NV_RJUST: Right justify and blank fill. This attribute has an associated size that defines the string length of the value.

NV_LJUST: Left justify and blank fill. This attribute has an associated size that defines the string length of the value.

NV_ZFILL: Without NV_LJUST, right justifies and fills with leading zeros. With NV_LJUST, left justify and strip leading zeros. Left justify and blank fill. This attribute has an associated size that defines the string length of the value.

NV_TAGGED: Indicates the tagged attribute.

NV_INTEGER: Causes value to be represented by a number. This attribute has an associated number that defines the arithmetic base to be used when the value is expanded as a string.

NV_DOUBLE: Used in conjunction with NV_INTEGER to cause value to be stored as a double precision floating point number. This attribute has an associated number that defines the number of places after the decimal point to be used when the value is expanded as a string.

NV_EXPNOTE: Used in conjunction with NV_INTEGER and NV_DOUBLE to cause the value to be represented in scientific notation when expanded as a string. This attribute has an associated number that defines the the precision of the mantissa.

NV_HEXFLOAT: Used in conjunction with NV_INTEGER and NV_DOUBLE to cause the value to be represented in C99 %a format when expanded as a string.

NV_BINARY: The name-value pair contains a buffer of binary data and nv_size() is the number of bytes for this data. By default the value will be represented by the base64 encoding of the buffer. The NV_LJUST flag may also be specified and causes the buffer size to be fixed and data either truncated or filled with 0 bytes.

NV_REF: The name-value pair is a name reference variable.

NV_MINIMAL: The name-value pair node is not embedded in a dictionary and is minimal size, NV_MINSZ.

NV_NODISC: All discipline functions are ignored when performing assignments and lookups.

The nv_isattr() function can test whether or not any of the attributes given by flags is set. The nv_onattr() and nv_offattr() functions turn attributes on or off respectively. Only attributes that do not affect the value can be set in this way. The nv_newattr() function can be used to change the attributes and size of the given name-value pair which may result in the value being changed to conform to the new attributes and size. The size argument is needed for attributes that require an additional argument such as justifies variables. Changing the attribute may require changing the value to agree with the new attributes. For an array variable, the values for each of the subscripts will be changed.

The nv_isvtree() function returns a pointer to the compound variable discipline if the node np is a compound variable or NULL otherwise.

The nv_isarray() function returns a non-zero value if the specified name-value pair is an array.

The nv_scan() function is used to walk through all name-value pairs in the dictionary given by dict. If the flags variable contains the NV_NOSCOPE flag, then only the top scope will be examined. The remaining flags will be used in conjunction with mask to further restrict the walk. If mask is non-zero, only the nodes for which nv_isattr(node,mask) is equal to flags will be visited. If fn is non-zero, then this function will be executed for each name-value pair in the walk. The arguments to fn will be a pointer to the name-value pair and the data pointer passed to nv_scan(). The number of elements visited will be returned.

The nv_clone() function is used make a copy of the contents of name-value pair src to another name-value pair dest.

Disciplines provide a way to intercept the lookup and assignment operations, to manage the creation of sub-variables, and to extend the operations permitted on a name-value pair. A discipline consists of a set of one or more functions and related data that are used to override and extend the operations on a name-value pair. A discipline is defined by the types Namfun_t and Namdisc_t. The Namdisc_t is not modified by any of these functions and can therefore be shared by several name-value pairs. It contains following public fields in the order listed:

      size_t	dsize;


      void	(*putval)(Namval_t*,const char*,int,Namfun_t*);


      char	*(*getval)(Namval_t*,Namfun_t*);


      double	(*getnum)(Namval_t*,Namfun_t*);


      char	*(*setdisc)(Namval_t*,const char*,Namval_t*,Namfun_t*);


      Namval_t	*(*createf)(Namval_t*,const char*,Namfun_t*);


      Namfun_t	*(*clonef)(Namval_t*,Namval_t*,int,Namfun_t*);


      char	*(*namef)(Namval_t*,Namfun_t*);


      Namval_t	*(*nextf)(Namval_t*,Dt_t*,Namfun_t*);


      Namval_t	*(*typef)(Namval_t*,Namfun_t*);

When a variable is referenced by calling the nv_getval() function, and the NV_NODISC attribute is not set, the getval() discipline function is called with a pointer to the name-value pair, np, and a pointer to the discipline, fp. Inside the getval() function, the nv_getv() function can be called to get the value of the name-value pair that would have resulted if the discipline were not used. The getnum() discipline is called whenever a numerical value is needed for the name-value pair np and the NV_NODISC attribute is not set, The nv_getn() function can be called from within the getnum() discipline to get the value that would have resulted if there were no getnum() discipline.

The putval() discipline function is used to override the assignment of values to a name-value pair. It is called whenever a value is assigned with nv_putval() and the NV_NODISC attribute is not set, or when a name-value pair is unset with nv_unset(). When a name-value pair is unset, putval() is called with value set to NULL. The nv_putv() function is used within the putval() to perform the assignment or unset that would have occurred if the discipline had not been installed.

The createf() discipline function is called from nv_open() or nv_create() when the name-value pair preceding a . is found. This function is passed the name-value pointer plus the remaining string and the current flags argument. The createf() discipline function must return the created name-value pair, otherwise the default action will be taken. If the name-value pair that is returned is the same as the one given, then the the behavior will be the same as if an invalid name had been given to nv_open(). The nv_create() function may be called within the createf() discipline function to perform the action that would have occurred by an earlier nv_open() function.

The setdisc() discipline function is used to extend the set of available shell level discipline functions associated with a name-value pair by allowing builtins or functions whose name is of the form varname.action to be defined. By default, each name-value pair can have a get, set, and unset discipline associated with it. Whenever a builtin or function whose name is of the form varname.action is defined or is unset, and action is not get, set, or unset, the setdisc() function is invoked with the same argument format as nv_setdisc(). The argument f points to the name-value pair associated with the function being defined, or NULL if the function is being unset. If the given action a is not known by this discipline, it should return the value returned by calling nv_setdisc(np,a,f,fp) so that it can be searched for in previously stacked disciplines. Otherwise, the setdisc() function should save the function name-value pair pointer, and return a non-NULL value. The name-value pointer to the function can be used to invoke the function at an application defined point. If the action a is NULL, then f points to an action name instead of a name-value pair pointer. The setdisc() must return the name of the action that follows the action name given by f. If f is also NULL, the name of the first action must be returned. This allows an application to get the list of valid discipline action names allowed by a given name-value pair.

The nv_adddisc() function is a higher level function that adds a setdisc discipline to the name-value pair that allows shell level disciplines to be created for each of the name specified in names.

The nv_discfun() function can be used to get a pointer to discipline functions that are provided by the library. Currently, the only one that is provided is the ones used to implement nv_adddisc() which can be returned with an argument of NV_DCADD.

The clonef() discipline function is called by nv_clone() when making a copy of the Namfun_t discipline to the new node. The first argument is the original node, the second argument is the new node, and the third argument is the flags that were passed down to nv_clone(). It must return a new instance of the Namfun_t* fp. If omitted, then memory whose size is determinated by the size field of fp, if non-zero, or fp->disc, will be allocated and copied from fp.

The namef() discipline function returns the name for this name-value pair.

The nextf() is used for walking through the list of sub-variables associated with this name-value pair. If the dictionary argument is NULL, it must return the first sub-variable. Otherwise, it must return the next sub-variable, or NULL if there are no more variables.

A discipline is installed or removed with the nv_disc() function. The following flags can be specified:

NV_FIRST: If fp is non-NULL, the discipline is moved to the top of the stack or pushed onto the top of the stack of disciplines associated with the given name-value pair np if not already present. Otherwise, the top of the discipline stack is returned.

NV_LAST: If fp is non-NULL, the discipline is moved to the bottom of the stack or pushed onto the bottom of the stack of disciplines associated with the given name-value pair np if not already present. Otherwise, the bottom of the discipline stack is returned.

NV_POP: If fp is non-NULL and it is on the stack, it is removed and fp is returned. If fp is non-NULL and is not on the stack, NULL is returned. Otherwise, the the top discipline is popped and returned.

NV_CLONE: If fp is non-NULL and it is on the stack, it is replace by a copy created by malloc(3). The nofree field is set to 0. The new discipline is returned. Otherwise, NULL is returned.

0: If fp is non-NULL then it is equivalent to NV_FIRST. Otherwise, it is equivalent to NV_POP.

The nv_hasdisc() function can be used to tell whether a discipline whose discipline functions are those defined in dp. A pointer to this discipline is returned.

The nv_aindex() function returns the current index for the indexed array given by the name-value pair pointer np. The return value is negative if np refers to an associative array.

The nv_setarray() function is used to create an associative array from a name-value pair node. The function fun defines the semantics of the associative array. Using fun equal to nv_associative() implements the default associative array semantics that are used with typeset -A. The function fun will be called with third argument as follows:

NV_AINIT: This will be called at initialization. The function you supply must return a pointer to a structure that contains the type Namarr_t as the first element. All other calls receive this value as an argument.

NV_AFREE: This will be called after all elements of the name-value pair have been deleted and the array is to be freed.

NV_ADELETE: The current element should be deleted.

NV_ANEXT: This means that the array subscript should be advanced to the next subscript. A NULL return indicates that there are no more subscripts.

NV_ANAME: The name of the current subscript must be returned.

NV_ACURRENT: Returns a pointer to a name-value pair corresponding to the current subscript, or NULL if this array type doesn't create represent each element as a name-value pair.

NV_ASETSUB: Set the current subscript to the name-value pair passed in as the second argument.

If np refers to an array, nv_arrayptr() returns a pointer to the array discipline structure Namarr_t. Otherwise nv_arrayptr() returns NULL.

If np refers to an array, the nv_getsub() returns a pointer to the name of the current subscript. Otherwise, nv_getsub() returns NULL.

The nv_opensub() function returns a pointer to the name-value pair corresponding to the current subscript in an associative array. Note that the nv_close() function should be called when the pointer is no longer needed.

The nv_putsub() function is used to set the subscript for the next reference to np. If the name argument is not NULL, it defines the value of the next subscript. The mode argument can contain one or more of the following flags:

ARRAY_ADD: Add the subscript if not found. Otherwise, nv_putsub() returns NULL if the given subscript is not found.

ARRAY_SCAN: Begin a walk through the subscripts starting at the subscript given by name. If name is NULL the walk is started from the beginning.

ARRAY_UNDEF: This causes any current scan to terminate and leaves the subscript in an undefined state.

If ARRAY_ADD is not given and the subscript does not exist, a NULL value is returned.

The nv_nextsub() function is used to advance to the next subscript. It returns 0 if there are no more subscripts or if called when not in a scan.

The nv_setref() function makes the name-value pair np into a reference to the variable whose name is given by the value of np. The nv_open() open function is called with this name, the dictionary given by dp, and the flags argument. A NULL value causes the shell global variable dictionary to be searched.

The nv_setvtree() function makes the name-value pair np into a tree structured variable so that nv_getval() will return a string containing all the names and values of children nodes in a format that can be used in a shell compound assignment.

The nv_type() function returns a name_value pair pointer that contains the type definition for the specified name-value pair. The nvname field contains the name for the type.

The nv_settype() function converts the name-value pair given by np into the type given by tp.

The nv_addtype() function adds the name of the type given by np to the list of declaration built-ins. The str argument contains the string used by optget(3) to generate the man page and process the options. The op argument specifies the callback discipline used by optget(3) and sz specifies the size of the callback information so that the discipline optget(3) can be extended with private data used by the callback function.

The nv_mkinttype() function creates named integer types of the specified name. The size parameter is the size in bytes of the integer variable and us is non-zero for unsigned integer types. If dp is specified then integer variables of this type will all use this discipline.

SEE ALSO¶

calloc(3) cdt(3) shell(3) optget(3)

AUTHOR¶

David G. Korn (dgk@research.att.com).