Methods to manipulate KeySets. More...

Functions
KeySet *	ksNew (size_t alloc,...)
KeySet *	ksVNew (size_t alloc, va_list va)

KeySet *	ksDup (const KeySet *source)
int	ksCopy (KeySet dest, const KeySet source)

int	ksDel (KeySet *ks)
ssize_t	ksGetSize (const KeySet *ks)
ssize_t	ksAppendKey (KeySet ks, Key toAppend)

ssize_t	ksAppend (KeySet ks, const KeySet toAppend)

KeySet *	ksCut (KeySet ks, const Key cutpoint)
Key *	ksPop (KeySet *ks)
int	ksRewind (KeySet *ks)
Key *	ksNext (KeySet *ks)
Key *	ksCurrent (const KeySet *ks)
Key *	ksHead (const KeySet *ks)
Key *	ksTail (const KeySet *ks)
cursor_t	ksGetCursor (const KeySet *ks)
Key *	ksAtCursor (KeySet *ks, cursor_t pos)
	return key at given cursor position
int	ksSetCursor (KeySet *ks, cursor_t cursor)
Key *	ksLookup (KeySet ks, Key key, option_t options)
Key *	ksLookupByName (KeySet ks, const char name, option_t options)

Detailed Description

Methods to manipulate KeySets.

A KeySet is a sorted set of keys. So the correct name actually would be KeyMap.

With ksNew() you can create a new KeySet.

You can add keys with ksAppendKey() in the keyset. Using ksAppend() you can append a whole keyset.

Note: Because the key is not copied, also the pointer to the current metadata keyNextMeta() will be shared.

ksGetSize() tells you the current size of the keyset.

With ksRewind() and ksNext() you can navigate through the keyset. Don't expect any particular order, but it is assured that you will get every key of the set.

KeySets have an internal cursor . This is used for ksLookup() and kdbSet().

KeySet has a fundamental meaning inside elektra. It makes it possible to get and store many keys at once inside the database. In addition to that the class can be used as high level datastructure in applications. With ksLookupByName() it is possible to fetch easily specific keys out of the list of keys.

You can easily create and iterate keys:

#include <kdb.h>
// create a new keyset with 3 keys
// with a hint that about 20 keys will be inside
KeySet *myConfig = ksNew(20,
        keyNew ("user/name1", 0),
        keyNew ("user/name2", 0),
        keyNew ("user/name3", 0),
        KS_END);
// append a key in the keyset
ksAppendKey(myConfig, keyNew("user/name4", 0));
Key *current;
ksRewind(myConfig);
while ((current=ksNext(myConfig))!=0)
{
        printf("Key name is %s.\n", keyName (current));
}
ksDel (myConfig); // delete keyset and all keys appended

Function Documentation

ssize_t ksAppend	(	KeySet *	ks,
		const KeySet *	toAppend
	)

Append all toAppend contained keys to the end of the ks.

toAppend KeySet will be left unchanged.

If a key is both in toAppend and ks, the Key in ks will be overridden.

Note: Because the key is not copied, also the pointer to the current metadata keyNextMeta() will be shared.

Postcondition: Sorted KeySet ks with all keys it had before and additionally the keys from toAppend

Returns: the size of the KeySet after transfer; -1 on NULL pointers

Parameters

ks	the KeySet that will receive the keys
toAppend	the KeySet that provides the keys that will be transferred

See Also: ksAppendKey()

ssize_t ksAppendKey	(	KeySet *	ks,
		Key *	toAppend
	)

Appends a Key to the end of ks.

A pointer to the key will be stored, and not a private copy. So a future ksDel() on ks may keyDel() the toAppend object, see keyGetRef().

The reference counter of the key will be incremented, and thus toAppend is not const.

Note: Because the key is not copied, also the pointer to the current metadata keyNextMeta() will be shared.

If the keyname already existed, it will be replaced with the new key.

The KeySet internal cursor will be set to the new key.

It is save to use ksAppendKey(ks, keyNew(..)).

Returns: the size of the KeySet after insertion; -1 on NULL pointers; -1 if insertion failed, the key will be deleted then.

Parameters

ks	KeySet that will receive the key
toAppend	Key that will be appended to ks or deleted

See Also: ksAppend(), keyNew(), ksDel(); keyIncRef()

Key* ksAtCursor	(	KeySet *	ks,
		cursor_t	pos
	)

return key at given cursor position

Parameters

ks	the keyset to pop key from
pos	where to get

Returns: the key at the cursor position on success

Return values

NULL	on NULL pointer, negative cursor position or a position that does not lie within the keyset

int ksCopy	(	KeySet *	dest,
		const KeySet *	source
	)

Copy a keyset.

Most often you may want a duplicate of a keyset, see ksDup() or append keys, see ksAppend(). But in some situations you need to copy a keyset to a existing keyset, for that this function exists.

You can also use it to clear a keyset when you pass a NULL pointer as source.

Note that all keys in dest will be deleted. Afterwards the content of the source will be added to the destination and the ksCurrent() is set properly in dest.

A flat copy is made, so the keys will not be duplicated, but there reference counter is updated, so both keysets need to be ksDel().

Note: Because the key is not copied, also the pointer to the current metadata keyNextMeta() will be shared.

int f (KeySet *ks)
{
        KeySet *c = ksNew (20, ..., KS_END);
        // c receives keys
        ksCopy (ks, c); // pass the keyset to the caller
        ksDel (c);
}       // caller needs to ksDel (ks)

Parameters

source	has to be an initialized source KeySet or NULL
dest	has to be an initialized KeySet where to write the keys

Returns: 1 on success; 0 if dest was cleared successfully (source is NULL); -1 on NULL pointer

See Also: ksNew(), ksDel(), ksDup(); keyCopy() for copying keys

Key* ksCurrent ( const KeySet * ks )

Return the current Key.

The pointer is NULL if you reached the end or after ksRewind().

Note: You must not delete the key or change the key, use ksPop() if you want to delete it.

Parameters

ks	the keyset object to work with

Returns: pointer to the Key pointed by ks's cursor; 0 on NULL pointer

See Also: ksNext(), ksRewind()

KeySet* ksCut	(	KeySet *	ks,
		const Key *	cutpoint
	)

Cuts out a keyset at the cutpoint.

Searches for the cutpoint inside the KeySet ks. If found it cuts out everything which is below (see keyIsBelow()) this key. If not found an empty keyset is returned.

The cursor will stay at the same key as it was before. If the cursor was inside the region of cutted (moved) keys, the cursor will be set to the key before the cutpoint.

Returns: a new allocated KeySet which needs to deleted with ksDel(). The keyset consists of all keys (of the original keyset ks) below the cutpoint. If the key cutpoint exists, it will also be appended.

Return values

0	on null pointers, no key name or allocation problems

Parameters

ks	the keyset to cut. It will be modified by removing all keys below the cutpoint. The cutpoint itself will also be removed.
cutpoint	the point where to cut out the keyset

int ksDel ( KeySet * ks )

A destructor for KeySet objects.

Cleans all internal dynamic attributes, decrement all reference pointers to all keys and then keyDel() all contained Keys, and free()s the release the KeySet object memory (that was previously allocated by ksNew()).

Parameters

ks	the keyset object to work with

Returns: 0 when the keyset was freed; -1 on null pointer

See Also: ksNew()

KeySet* ksDup ( const KeySet * source )

Return a duplicate of a keyset.

Objects created with ksDup() must be destroyed with ksDel().

Memory will be allocated as needed for dynamic properties, so you need to ksDel() the returned pointer.

A flat copy is made, so the keys will not be duplicated, but there reference counter is updated, so both keysets need ksDel().

Parameters

source has to be an initializised source KeySet

Returns: a flat copy of source on success; 0 on NULL pointer

See Also: ksNew(), ksDel(); keyDup() for Key duplication

cursor_t ksGetCursor ( const KeySet * ks )

Get the KeySet internal cursor.

Use it to get the cursor of the actual position.

Warning: Cursors are getting invalid when the key was ksPop()ed or ksLookup() with KDB_O_POP was used.

Read ahead

With the cursors it is possible to read ahead in a keyset:

cursor_t jump;
ksRewind (ks);
while ((key = keyNextMeta (ks))!=0)
{
        // now mark this key
        jump = ksGetCursor(ks);
        //code..
        keyNextMeta (ks); // now browse on
        // use ksCurrent(ks) to check the keys
        //code..
        // jump back to the position marked before
        ksSetCursor(ks, jump);
}

Restoring state

It can also be used to restore the state of a keyset in a function

int f (KeySet *ks)
{
        cursor_t state = ksGetCursor(ks);
        // work with keyset
        // now bring the keyset to the state before
        ksSetCursor (ks, state);
}

It is of course possible to make the KeySet const and cast its const away to set the cursor. Another way to achieve the same is to ksDup() the keyset, but it is not as efficient.

An invalid cursor will be returned directly after ksRewind(). When you set an invalid cursor ksCurrent() is 0 and ksNext() == ksHead().

Note: Only use a cursor for the same keyset which it was made for.

Parameters

ks	the keyset object to work with

Returns: a valid cursor on success; an invalid cursor on NULL pointer or after ksRewind()

See Also: ksNext(), ksSetCursor()

ssize_t ksGetSize ( const KeySet * ks )

Return the number of keys that ks contains.

Parameters

ks	the keyset object to work with

Returns: the number of keys that ks contains.; -1 on NULL pointer

See Also: ksNew(0), ksDel()

Key* ksHead ( const KeySet * ks )

Return the first key in the KeySet.

The KeySets cursor will not be effected.

If ksCurrent()==ksHead() you know you are on the first key.

Parameters

ks	the keyset object to work with

Returns: the first Key of a keyset; 0 on NULL pointer or empty keyset

See Also: ksTail() for the last Key; ksRewind(), ksCurrent() and ksNext() for iterating over the KeySet

Key* ksLookup	(	KeySet *	ks,
		Key *	key,
		option_t	options
	)

Look for a Key contained in ks that matches the name of the key.

Introduction

ksLookup() is designed to let you work with entirely pre-loaded KeySets, so instead of kdbGetKey(), key by key, the idea is to fully kdbGet() for your application root key and process it all at once with ksLookup().

This function is very efficient by using binary search. Together with kdbGet() which can you load the whole configuration with only some communication to backends you can write very effective but short code for configuration.

Usage

If found, ks internal cursor will be positioned in the matched key (also accessible by ksCurrent()), and a pointer to the Key is returned. If not found, ks internal cursor will not move, and a NULL pointer is returned.

Cascading is done if the first character is a /. This leads to ignoring the prefix like system/ and user/.

if (kdbGet(handle, "user/myapp", myConfig, 0 ) == -1)
        errorHandler ("Could not get Keys");
if (kdbGet(handle, "system/myapp", myConfig, 0 ) == -1)
        errorHandler ("Could not get Keys");
if ((myKey = ksLookup(myConfig, key, 0)) == NULL)
        errorHandler ("Could not Lookup Key");

This is the way multi user Programs should get there configuration and search after the values. It is guaranteed that more namespaces can be added easily and that all values can be set by admin and user.

KDB_O_NOALL

When KDB_O_NOALL is set the keyset will be only searched from ksCurrent() to ksTail(). You need to ksRewind() the keyset yourself. ksCurrent() is always set properly after searching a key, so you can go on searching another key after the found key.

When KDB_O_NOALL is not set the cursor will stay untouched and all keys are considered. A much more efficient binary search will be used then.

KDB_O_POP

When KDB_O_POP is set the key which was found will be ksPop()ed. ksCurrent() will not be changed, only iff ksCurrent() is the searched key, then the keyset will be ksRewind()ed.

Note: Like in ksPop() the popped key always needs to be keyDel() afterwards, even if it is appended to another keyset.

Warning: All cursors on the keyset will be invalid iff you use KDB_O_POP, so don't use this if you rely on a cursor, see ksGetCursor().

You can solve this problem by using KDB_O_NOALL, risking you have to iterate n^2 instead of n.

The more elegant way is to separate the keyset you use for ksLookup() and ksAppendKey():

int f(KeySet *iterator, KeySet *lookup)
{
        KeySet *append = ksNew (ksGetSize(lookup), KS_END);
        Key *key;
        Key *current;
        ksRewind(iterator);
        while (current=ksNext(iterator))
        {
                key = ksLookup (lookup, current, KDB_O_POP);
                // do something...
                ksAppendKey(append, key); // now append it to append, not lookup!
                keyDel (key); // make sure to ALWAYS delete poped keys.
        }
        ksAppend(lookup, append);
        // now lookup needs to be sorted only once, append never
        ksDel (append);
}

Parameters

ks	where to look for
key	the key object you are looking for
options	some `KDB_O_*` option bits: `KDB_O_NOCASE` Lookup ignoring case. `KDB_O_WITHOWNER` Also consider correct owner. `KDB_O_NOALL` Only search from ksCurrent() to end of keyset, see above text. `KDB_O_POP` Pop the key which was found. `KDB_O_DEL` Delete the passed key.

Returns: pointer to the Key found, 0 otherwise; 0 on NULL pointers

See Also: ksLookupByName() to search by a name given by a string; ksCurrent(), ksRewind(), ksNext() for iterating over a KeySet

Key* ksLookupByName	(	KeySet *	ks,
		const char *	name,
		option_t	options
	)

Look for a Key contained in ks that matches name.

ksLookupByName() is designed to let you work with entirely pre-loaded KeySets, so instead of kdbGetKey(), key by key, the idea is to fully kdbGetByName() for your application root key and process it all at once with ksLookupByName().

This function is very efficient by using binary search. Together with kdbGetByName() which can you load the whole configuration with only some communication to backends you can write very effective but short code for configuration.

If found, ks internal cursor will be positioned in the matched key (also accessible by ksCurrent()), and a pointer to the Key is returned. If not found, ks internal cursor will not move, and a NULL pointer is returned. If requested to pop the key, the cursor will be rewinded.

Cascading

Cascading is done if the first character is a /. This leads to ignoring the prefix like system/ and user/.

if (kdbGet(handle, "user/sw/myapp/current", myConfig, parentKey ) == -1)
        errorHandler ("Could not get Keys", parentKey);
if (kdbGet(handle, "system/sw/myapp/current", myConfig, parentKey ) == -1)
        errorHandler ("Could not get Keys", parentKey);
if ((myKey = ksLookupByName (myConfig, "/myapp/current/key", 0)) == NULL)
        errorHandler ("Could not Lookup Key");

This is the way multi user Programs should get there configuration and search after the values. It is guaranteed that more namespaces can be added easily and that all values can be set by admin and user.

It is up to the application to implement a sophisticated cascading algorithm, for e.g. a list of profiles (specific, group and fallback):

if ((myKey = ksLookupByName (myConfig, "/myapp/current/specific/key", 0)) == NULL)
        if ((myKey = ksLookupByName (myConfig, "/myapp/current/group/key", 0)) == NULL)
                if ((myKey = ksLookupByName (myConfig, "/myapp/current/fallback/key", 0)) == NULL)
                        errorHandler ("All fallbacks failed to lookup key");

Note that for every profile both the user and the system key are searched. The first key found will be used.

Full Search

When KDB_O_NOALL is set the keyset will be only searched from ksCurrent() to ksTail(). You need to ksRewind() the keyset yourself. ksCurrent() is always set properly after searching a key, so you can go on searching another key after the found key.

When KDB_O_NOALL is not set the cursor will stay untouched and all keys are considered. A much more efficient binary search will be used then.

Parameters

ks	where to look for
name	key name you are looking for
options	some `KDB_O_*` option bits: `KDB_O_NOCASE` Lookup ignoring case. `KDB_O_WITHOWNER` Also consider correct owner. `KDB_O_NOALL` Only search from ksCurrent() to end of keyset, see above text. `KDB_O_POP` Pop the key which was found.

Currently no options supported.

Returns: pointer to the Key found, 0 otherwise; 0 on NULL pointers

See Also: keyCompare() for very powerful Key lookups in KeySets; ksCurrent(), ksRewind(), ksNext()

KeySet* ksNew	(	size_t	alloc,
			...
	)

Allocate, initialize and return a new KeySet object.

Objects created with ksNew() must be destroyed with ksDel().

You can use a various long list of parameters to preload the keyset with a list of keys. Either your first and only parameter is 0 or your last parameter must be KEY_END.

So, terminate with ksNew(0) or ksNew(20, ..., KS_END)

For most uses

KeySet *keys = ksNew(0);
// work with it
ksDel (keys);

goes ok, the alloc size will be 16, defined in kdbprivate.h. The alloc size will be doubled whenever size reaches alloc size, so it also performs out large keysets.

But if you have any clue how large your keyset may be you should read the next statements.

If you want a keyset with length 15 (because you know of your application that you normally need about 12 up to 15 keys), use:

KeySet * keys = ksNew (15,
        keyNew ("user/sw/app/fixedConfiguration/key01", KEY_SWITCH_VALUE, "value01", 0),
        keyNew ("user/sw/app/fixedConfiguration/key02", KEY_SWITCH_VALUE, "value02", 0),
        keyNew ("user/sw/app/fixedConfiguration/key03", KEY_SWITCH_VALUE, "value03", 0),
        // ...
        keyNew ("user/sw/app/fixedConfiguration/key15", KEY_SWITCH_VALUE, "value15", 0),
        KS_END);
// work with it
ksDel (keys);

If you start having 3 keys, and your application needs approximately 200-500 keys, you can use:

KeySet * config = ksNew (500,
        keyNew ("user/sw/app/fixedConfiguration/key1", KEY_SWITCH_VALUE, "value1", 0),
        keyNew ("user/sw/app/fixedConfiguration/key2", KEY_SWITCH_VALUE, "value2", 0),
        keyNew ("user/sw/app/fixedConfiguration/key3", KEY_SWITCH_VALUE, "value3", 0),
        KS_END); // don't forget the KS_END at the end!
// work with it
ksDel (config);

Alloc size is 500, the size of the keyset will be 3 after ksNew. This means the keyset will reallocate when appending more than 497 keys.

The main benefit of taking a list of variant length parameters is to be able to have one C-Statement for any possible KeySet.

Due to ABI compatibility, the KeySet structure is only declared in kdb.h, and not defined. So you can only declare pointers to KeySets in your program. See http://tldp.org/HOWTO/Program-Library-HOWTO/shared-libraries.html#AEN135

See Also: ksDel() to free the KeySet afterwards; ksDup() to duplicate an existing KeySet

Parameters

alloc gives a hint for the size how many Keys may be stored initially

Returns: a ready to use KeySet object; 0 on memory error

Key* ksNext ( KeySet * ks )

Returns the next Key in a KeySet.

KeySets have an internal cursor that can be reset with ksRewind(). Every time ksNext() is called the cursor is incremented and the new current Key is returned.

You'll get a NULL pointer if the key after the end of the KeySet was reached. On subsequent calls of ksNext() it will still return the NULL pointer.

The ks internal cursor will be changed, so it is not const.

Note: You must not delete or change the key, use ksPop() if you want to delete it.

Parameters

ks	the keyset object to work with

Returns: the new current Key; 0 when the end is reached; 0 on NULL pointer

See Also: ksRewind(), ksCurrent()

Key* ksPop ( KeySet * ks )

Remove and return the last key of ks.

The reference counter will be decremented by one.

The KeySets cursor will not be effected if it did not point to the popped key.

Note: You need to keyDel() the key afterwards, if you don't append it to another keyset. It has the same semantics like a key allocated with keyNew() or keyDup().

ks1=ksNew(0);
ks2=ksNew(0);
k1=keyNew("user/name", KEY_END); // ref counter 0
ksAppendKey(ks1, k1); // ref counter 1
ksAppendKey(ks2, k1); // ref counter 2
k1=ksPop (ks1); // ref counter 1
k1=ksPop (ks2); // ref counter 0, like after keyNew()
ksAppendKey(ks1, k1); // ref counter 1
ksDel (ks1); // key is deleted too
ksDel (ks2);
 *

Returns: the last key of ks; NULL if ks is empty or on NULL pointer

Parameters

ks	KeySet to work with

See Also: ksAppendKey(), ksAppend(); commandList() for an example

int ksRewind ( KeySet * ks )

Rewinds the KeySet internal cursor.

Use it to set the cursor to the beginning of the KeySet. ksCurrent() will then always return NULL afterwards. So you want to ksNext() first.

ksRewind (ks);

while ((key = ksNext (ks))!=0) {}

Parameters

ks	the keyset object to work with

Returns: 0 on success; -1 on NULL pointer

See Also: ksNext(), ksCurrent()

int ksSetCursor	(	KeySet *	ks,
		cursor_t	cursor
	)

Set the KeySet internal cursor.

Use it to set the cursor to a stored position. ksCurrent() will then be the position which you got with.

Warning: Cursors may get invalid when the key was ksPop()ed or ksLookup() was used together with KDB_O_POP.

cursor_t cursor;
..
// key now in any position here
cursor = ksGetCursor (ks);
while ((key = keyNextMeta (ks))!=0) {}
ksSetCursor (ks, cursor); // reset state
ksCurrent(ks); // in same position as before

An invalid cursor will set the keyset to its beginning like ksRewind(). When you set an invalid cursor ksCurrent() is 0 and ksNext() == ksHead().

Parameters

cursor	the cursor to use
ks	the keyset object to work with

Returns: 0 when the keyset is ksRewind()ed; 1 otherwise; -1 on NULL pointer

See Also: ksNext(), ksGetCursor()

Key* ksTail ( const KeySet * ks )

Return the last key in the KeySet.

The KeySets cursor will not be effected.

If ksCurrent()==ksTail() you know you are on the last key. ksNext() will return a NULL pointer afterwards.

Parameters

ks	the keyset object to work with

Returns: the last Key of a keyset; 0 on NULL pointer or empty keyset

See Also: ksHead() for the first Key; ksRewind(), ksCurrent() and ksNext() for iterating over the KeySet

KeySet* ksVNew	(	size_t	alloc,
		va_list	va
	)

Allocate, initialize and return a new KeySet object.Objects created with ksNew() must be destroyed with ksDel().You can use a various long list of parameters to preload the keyset with a list of keys. Either your first and only parameter is 0 or your last parameter must be KEY_END.So, terminate with ksNew(0) or ksNew(20, ..., KS_END)For most uses

KeySet *keys = ksNew(0);
// work with it
ksDel (keys);

goes ok, the alloc size will be 16, defined in kdbprivate.h. The alloc size will be doubled whenever size reaches alloc size, so it also performs out large keysets.But if you have any clue how large your keyset may be you should read the next statements.If you want a keyset with length 15 (because you know of your application that you normally need about 12 up to 15 keys), use:

KeySet * keys = ksNew (15,
        keyNew ("user/sw/app/fixedConfiguration/key01", KEY_SWITCH_VALUE, "value01", 0),
        keyNew ("user/sw/app/fixedConfiguration/key02", KEY_SWITCH_VALUE, "value02", 0),
        keyNew ("user/sw/app/fixedConfiguration/key03", KEY_SWITCH_VALUE, "value03", 0),
        // ...
        keyNew ("user/sw/app/fixedConfiguration/key15", KEY_SWITCH_VALUE, "value15", 0),
        KS_END);
// work with it
ksDel (keys);

If you start having 3 keys, and your application needs approximately 200-500 keys, you can use:

KeySet * config = ksNew (500,
        keyNew ("user/sw/app/fixedConfiguration/key1", KEY_SWITCH_VALUE, "value1", 0),
        keyNew ("user/sw/app/fixedConfiguration/key2", KEY_SWITCH_VALUE, "value2", 0),
        keyNew ("user/sw/app/fixedConfiguration/key3", KEY_SWITCH_VALUE, "value3", 0),
        KS_END); // don't forget the KS_END at the end!
// work with it
ksDel (config);

Alloc size is 500, the size of the keyset will be 3 after ksNew. This means the keyset will reallocate when appending more than 497 keys.The main benefit of taking a list of variant length parameters is to be able to have one C-Statement for any possible KeySet.Due to ABI compatibility, the KeySet structure is only declared in kdb.h, and not defined. So you can only declare pointers to KeySets in your program. See http://tldp.org/HOWTO/Program-Library-HOWTO/shared-libraries.html#AEN135

See Also: ksDel() to free the KeySet afterwards; ksDup() to duplicate an existing KeySet

Parameters

alloc gives a hint for the size how many Keys may be stored initially

Returns: a ready to use KeySet object; 0 on memory error

Precondition: caller must call va_start and va_end

va the list of arguments

Functions

Detailed Description

Function Documentation

Read ahead

Restoring state

Introduction

Usage

KDB_O_NOALL

KDB_O_POP

Cascading

Full Search