LIBLIST List Manipulation Library

CONTENTS

* INTRODUCTION

When programming in C and you intend to write any type of complex program then you often find yourself wanting to manipulate lists of data. You therefore find yourself implementing list handling routines in each such program. This is costly in time and prone to error. Additionally the main logic of your program is often obscured by the code which does the mechanics of the list handling.

This library simplifies this problem by providing a packaged series of list routines that are known to work. The details of the list manipulation are no longer your concern - all you need to do is to specify the actions that you require. All details of the internal control structures used to maintain such lists are hidden from you. Instead you are just provided with a series of properties and methods that you can use against any list type.

The library treats lists in an object oriented manner which will allow you to clearly separate out those bits of any program that are associated with list handling. This leads to better structured source code that is much easier to follow. A consequence is that your program is less prone to logic errors.

A variety of different types of list are supported. Each type has different benefits in terms of processing cost, memory overheads per list instance and/or node, etc. The same methods and properties are supported for each list type as far as is practical, even though for some of the types this can be relatively inefficient in processing terms. This means that you can pick the list type that is optimised for your purposes, but if later your requirements change you can change the list type used without major change to your program. The new list type will support the same properties and methods that you have used on the previous list type.

Once you have mastered using this library, then you will find that many problems that manipulate lists become much easier to solve. You no longer have to think of the complexities of how lists operate, but just think of what you want to use them for. This will improve your productivity and reduce the number of errors you need to track down.

* USING THIS LIBRARY

        LIBLIST.HTM     Documentation in HTML format
        LIBLIST.H       Header file for use with library
        LIBLIST.HDR     Annotated (unpacked) version of above header
        LIBLIST.A       Standard version of library
        LIBLISTD.A      Debug version of library
        LISTSRC.ZIP     Source of the LIBLIST library as a ZIP archive

      #include <liblist.h>

You also need to tell the linker that it needs to search the list handling library for the routines that you have used. The library is provided in a number of different formats and you can specify which you want to use:

Standard Version

This is a normal statically linked version provided in the file liblist.a. If you want to use this version then you need to provide parameters to the linker to get it searched for required routines. The exact format of the parameter can vary according to the linker used, but would typically be something like:

        -llist

This will result in a program that has a version of the library that is optimised for runtime efficiency, and has no dependency on other files being present at runtime for list handling support.

Debug Version

Programs that make extensive use of lists are often by their very nature relatively complex and thus more prone to having internal logic faults.

The debugging version of this library helps identify such problems by including extensive internal checks against you misusing the library and providing silly parameter values. This is extremely useful while in the development phase of a program. The penalty is that there is significantly more overhead in terms of program size, memory usage and processing overhead while using the debugging version of the library.

To use the debugging version of the library which is provided in the file liblistD.a you need to provide parameters to the linker to get it searched for required routines. The exact format of the parameter can vary according to the linker used, but would typically be something like:

        -llistd

to the LD linker.

While you are using the debugging version of the library, if any of the internal checks fail then your program will be stopped and an assert style message displayed indicating the point at which the error was detected. The text of the assert message will display the details of the internal check that failed. These messages are designed to be self-explanatory.

Once you have finished debugging your program, then by simply changing the parameter provided to the LD linker you can switch to a non-debugging version of the library.

RLL Version (QDOS based systems only)

The RLL (Runtime Link Library) version is provided to allow the required routines to be picked up dynamically at runtime from a separately loaded RLL library.

To use the RLL version of the library you need to provide parameters of the form

        -rlist -llist

to the LD linker. Note that both the above parameters are needed as there are still a few routines (at this release anyway) that still need to be picked up from the statically linked version of the library.

Note also that if you use this latter format it is necessary to ensure that the user has a copy of the RLL library file LIBLIST_RLL as this will be required at runtime. The user will also need to have the Runtime Link Manager (or RLM) present and running on his system at the time that your program that uses a RLL file is started.

* LIST CLASSES

The following are the list classes currently supported by this library (more may be added in future releases). Although the use of this library largely hides the details of such complexity from you, you should still be aware of the characteristics of each list class so that you can select the class that is optimised for the purpose that you have in mind.

Single Linked List

This one of the commonest types of list. It is typically used to store items that are unordered, although ordered operations are still supported at the expense of additional processing overheads.

The advantages of this type of list are that it has the minimal amount of memory overhead associated with storing each entry (or node) in the list.

The disadvantages of this type of list are that operations that involve add new entries towards the end of the list; searching for particular entries in the list; or attempting to move backwards through the list tend to be rather expensive in processing terms, and the overhead tends to grow linearly with list size.

Double Linked list

This is a convenient way of storing a small amount of data in either ordered or unordered form.

It is similar in many ways to the Single Linked list mentioned above, except that it also supports reasonably efficient movement backwards through the list. The penalty you pay is slightly more memory overhead for each node in the list.

Queues (FIFO)

A queue is a specialised use of a list where you always want to add items at the end and remove them from the front. This is why it is also commonly known as a FIFO or First-In-First-Out list.

A queue is optimised for inserting new nodes at the end of the list and removing old nodes from the front. It does not support the concept of an ordered list. In other respects it tends to behave rather like the Single Linked List mentioned earlier.

Stacks (LIFO)

A stack is used when you always want to add items at the end of the list and remove them from the front. This is why it is also commonly known as a LIFO or Last-In-First-Out list.

A stack is optimised for inserting and removing new nodes from the front of the list. It does not support the concept of an ordered ordered list. In all other respects it tends to behave rather like the Single Linked List mentioned earlier.

Binary Tree

A binary tree is commonly used when you want to store large numbers of items that need to be kept in some sort of ordered manner. A typical example might be a list of program symbols.

The advantage of the binary tree is that it is optimised for quick searching and insertion of new nodes at any point using methods where the processing overheads do not grow linearly with list size.

The disadvantage of the binary tree is that the storage overhead is slightly more than for any of the linked list types mentioned earlier. It is also a very inefficient way to store unordered data - for this it is better to use one of the linked list variants.

Balanced Binary Tree

A balanced binary tree is a specialisation of the standard binary tree type that tries to optimise the search operations even more than the simple binary tree type of list.

A standard binary tree type of list will become inefficient at fast searches if you insert data that is largely ordered in nature. The balanced binary tree type of list caters for this situation at the expense of more processing overhead being incurred while adding data using any of the insertion methods, and a memory overhead dues to having to store slightly more information about each node.

Hash Table

A hash table is used in situations in which the key requirement is very fast look-up of whether nodes already exist or not. It does not by its very nature support the idea of an ordered list. A typical use might be in a compiler for storing information about symbols as one would continually be checking to see if a particular symbol exists.

The idea of a hashtable is that is initialised with a specified number of nodes (known as buckets). This number is then fixed for the duration of any particular instance of the hash table type of list. A routine (called the hash algorithm) that does some sort of calculation on the supplied data is used to determine in which bucket this data should be stored. As this does not involve any searches, and as the calculations can be optimised for speed and all the required buckets are pre-created, this process can very fast. The main disadvantage of a hash table form of list is the memory overhead required to hold the fixed number of buckets.

In practice a realistic use of a hash table is almost always implemented as a 'list of lists'. The first level is the true hash table which uses the hashing algorithm to get down to a particular bucket within the hash table, and then attached to that bucket is some other sort of list to resolve the situation of there being more than one node that has the same hash value. This second level can be another level of hash table (using a different hashing algorithm to the first level) or some other type of list. In many cases the last level is a LIFO (or stack) type of list as this means that it can store any amount of data (memory permitting) and the most recently inserted node (which is the one you are often most likely to want) is found first in any search. This is in fact the default that will be assumed if you do not specify otherwise using a LIST_HashSetup method. As long as the size of the hash table is well chosen, and the hashing algorithm gives a good distribution of data, these attached lists will be small, thus still giving fast insertion and searching of them.

* ORDERED and UNORDERED LISTS

Ordered lists

In an ordered list one (or more) of the fields within a list node are designated as key fields. The position that an entry should occupy in the list is determined by comparing the key fields of the various entries. When using an ordered list you normally just tell the system to add a new entry to the list and let it work out the correct position that it should occupy by doing any needed comparisons on the designated key fields. A typical use of an ordered list might be for a list of names that you want maintained in alphabetical order.

If you want to use ordered lists, then you have to supply a Node Comparison function at the time you create the list that is used by the library to decide the correct order. This is described in more detail later in this document in the section on user defined functions. Note also that not all the list types can support the concept of an ordered list. The description of each list type will tell you whether ordered lists are supported or not.

unordered lists

In an unordered list on the other hand there is no explicit key field. Instead the order is simply determined by the locations at which you insert new entries. When building an unordered list you have to specify exactly where any new entry should be inserted. This mode of operation is particularly suited to lists where you will be inserting and/or deleting entries from either the front or the back of the list.

* EMBEDDED LISTS

The drawback of an embedded list is that there are additional overheads in both memory terms and processing terms every time you add or delete nodes. These overheads are less than maintaining the two lists completely separately, but more than maintaining a single list. The big advantage is that it avoids the user having to explicitly use two (or more) method calls when creating, inserting or deleting nodes from the lists making up an embedded list with the possibility of forgetting to do so (and thus getting the lists out of step with each other).

The way an embedded list works is that you first create an empty list of one type. You then embed this list in another list (using the LIST_Embed method) and are returned a handle to this second list. Once you have done this, then any method that would add or delete nodes from the list will actually add or delete the node from both lists in a single call. It does not matter in this case whether you use the list handle belonging to the original list, or the handle belonging to the new containing list. The LIBLIST library will automatically maintain the necessary links to add or delete nodes from both lists simultaneously.

When you use a property or method that does not add or delete nodes, but merely does something that is determined by the lists order, then the LIBLIST library will use the method that is appropriate to the list type associated with the particular list handle that you use.

An example of when you might want to use an embedded list would be the case in which you wished to build up a symbol list for which you wanted the ability to access the symbols in either address order or alphabetical name order. In this case you could first create a list that was specified to maintain the symbols in address order. You would do this by using the LIST_Create method where the parameter specifying the node compare function was for a function that did a comparison on the symbol address. You could then embed this list in another one using the LIST_Embed method, but this time the parameter specifying the node compare function would be for a function that does a comparison on the symbol name. In an extreme case one could then perhaps embed both of these lists inside another one that was a hash table optimised for maximum speed on looking up whether the symbol exists or not.

* DATA TYPES

list_t

This is a generic pointer data type returned when a list is created. It is effectively the handle to any particular list operation, and is passed as a parameter to all such operations. It is the use of this technique that allows this library to insulate the programmer from the implementation details of each of the list types.

listclass_t

This is a data type that is passed as a parameter when creating a list. It defines which of the classes of lists given earlier should be used when creating this new list instance. This will be the only use made of this data type at the user level, although it is widely used internally within the implementation of this library.

node_t

This is a generic 'pointer to node' type that is passed as a parameter to the various library routines, and in some cases returned as an address from these routines. The important thing about it is that it always points at the user data part of any node - the programmer does not see any associated control information.

Within user code one would normally cast this generic type to a pointer to a particular structure type. This will allow you to access the fields that are stored within the node.

* PROPERTIES

The LIBLIST library actually supports all properties for all list types. However each list type has a selection of the properties that are most frequently used with that type. Use of other properties is likely in many cases to be relatively expensive in processing terms.

The following table show the properties that you are most likely to use with each list type, and then for each property (in alphabetical order) a detailed description is given describing its syntax and use.

Properties	List Types
	Single Linked List	Double Linked List	FIFO List (Queue)	LIFO List (Stack)	Binary Tree	Balanced Binary Tree	Hash Table
LIST_Class	yes	yes	yes	yes	yes	yes	yes
LIST_Type	yes	yes	yes	yes	yes	yes	yes
LIST_Ordered	yes	yes	yes	yes	yes	yes	n/a
LIST_Embedded	yes	yes	yes	yes	yes	yes	yes
LIST_Count	yes	yes	yes	yes	yes	yes	yes
LIST_Index	yes	yes	yes	rarely	yes	yes	-
LIST_First	yes	yes	-	-	-	-	-
LIST_Last	yes	yes	-	-	-	-	-
LIST_Next	yes	yes	-	-	rarely	rarely	rarely
LIST_Previous	yes	yes	-	-	rarely	rarely	rarely

int LIST_Class(list_t list);

       LIST_CLASS_SINGLE           Single linked list
       LIST_CLASS_DOUBLE           Double linked list
       LIST_CLASS_QUEUE            Queue
       LIST_CLASS_FIFO             FIFO (another name for queue)
       LIST_CLASS_STACK            Stack
       LIST_CLASS_LIFO             LIFO (another name for stack)
       LIST_CLASS_BTREE            Binary tree
       LIST_CLASS_BALTREE          Balanced Binary Tree
       LIST_CLASS_HASH             Hash Table

long  LIST_Count(list_t list);

int  LIST_Embedded(list_t list);

node_t LIST_First(list_t list);

This property would normally only be used on unordered lists. On ordered lists you are more likely to be using the LIST_Find or LIST_Enumerate methods.

size_t  LIST_Index(list_t list, node_t node);

One point to watch though is the index value may not stay valid if you insert new data into a list. Whether it does or not will depend on the relative positions in the list at which the new data is inserted. You should always assume that inserting new data invalidates any current index property that you may have obtained earlier.

node_t LIST_Last(list_t list);

This property would normally only be used on unordered lists. On ordered lists you are more likely to be using the LIST_Find or LIST_Enumerate methods.

node_t  LIST_Next(list_t list, node_t node);

int   LIST_Ordered(list_t);

	LIST_ORDERED	List is ordered
	LIST_UNORDERED	List is unordered
	LIST_ORDER_NA	Order not applicable to this list type

In practice you would rarely need to use this property as you had to specify when you created the list whether it was ordered or not so you probably already know the answer.

node_t LIST_Previous(list_t list, node_t node);

int LIST_Type(list_t list);

       LIST_TYPE_SINGLE           Single linked list
       LIST_TYPE_DOUBLE           Double linked list
       LIST_TYPE_QUEUE            Queue
       LIST_TYPE_FIFO             FIFO (another name for queue)
       LIST_TYPE_STACK            Stack
       LIST_TYPE_LIFO             LIFO (another name for stack)
       LIST_TYPE_BTREE            Binary tree
       LIST_TYPE_BALTREE          Balanced Binary Tree
       LIST_TYPE_HASH             Hash Table

* METHODS

The methods are used to carry out some action on the list. In most cases they will actually change the contents of the list.

The LIBLIST library supports most methods for all list types. However each list type has a selection of methods that are most frequently used with that type. There are a few methods that are very specific to particular list types, and can return an error code if used on the wrong type. These are indicated in the table below.

The following table show the methods that you are most likely to use with each list type, and then for each method (in alphabetical order) a detailed description is given describing its syntax and use.

Methods	List Types
	Single Linked List	Double Linked List	FIFO List (Queue)	LIFO List (Stack)	Binary Tree	Balanced Binary Tree	Hash Table
LIST_Create	yes	yes	yes	yes	yes	yes	yes
LIST_Clone	yes	yes	yes	yes	yes	yes	yes
LIST_Embed	yes	yes	yes	yes	yes	yes	yes
LIST_HashSetup	error	error	error	error	error	error	yes
LIST_Destroy	yes	yes	yes	yes	yes	yes	yes
LIST_NewNode	yes	yes	yes	yes	yes	yes	yes
LIST_FreeNode	yes	yes	yes	yes	yes	yes	yes
LIST_Add	yes	yes	-	-	yes	yes	yes
LIST_NewAdd	yes	yes	-	-	yes	yes	yes
LIST_Insert	yes	yes	-	-	-	-	-
LIST_NewInsert	yes	yes	-	-	-	-	-
LIST_NewAppend	yes	yes	-	-	-	-	-
LIST_Append	yes	yes	-	-	-	-	-
LIST_Before	yes	yes	rarely	rarely	-	-	-
LIST_NewBefore	yes	yes	rarely	rarely	-	-	-
LIST_After	yes	yes	rarely	rarely	-	-	-
LIST_NewAfter	yes	yes	rarely	rarely	-	-	-
LIST_Remove	yes	yes	rarely	rarely	yes	yes	yes
LIST_Delete	yes	yes	rarely	rarely	yes	yes	yes
LIST_Delete	yes	yes	rarely	rarely	yes	yes	yes
LIST_Position	yes	yes	yes	rarely	rarely	rarely	rarely
LIST_Find	yes	yes	-	-	yes	yes	yes
LIST_Enqueue	-	-	yes	-	-	-	-
LIST_Dequeue	-	-	yes	-	-	-	-
LIST_NewEnqueue	-	-	yes	-	-	-	-
LIST_Push	-	-	-	yes	-	-	-
LIST_NewPush	-	-	-	yes	-	-	-
LIST_Pop	maybe	maybe	-	yes	-	-	-
LIST_Peek	maybe	maybe	yes	yes	-	-	-

int   LIST_Add(list_t list, node_t node);

It is normally expected to be used with ordered lists, and in this case the new node will be inserted at the correct point to maintain the order. The user supplied node comparison function will be used to help determine what is the correct insertion point.

If used with an unordered list, then the new node will be inserted at the most sensible point. This means that it is treated as if it were an LIST_Append method for all list types except for a Stack/LIFO list and the node added at the end of the list. In the special case of the Stack/LIFO list type, this method acts as if it were an LIST_Insert method call and will insert data at the start of the list.

int   LIST_After(list_t list, node_t newnode, node_t oldnode);

It is normally expected to be used with unordered lists. If used with ordered lists, then it an LIST_ERROR_SEQUENCE error code will be returned if inserting the specified node at the specified point would violate the ordering constraints.

If the oldnode parameter is NULL, then this method will act just like the LIST_Append method and insert the new node as the last node in the list.

int   LIST_Append(list_t list, node_t node);

It is normally expected to be used with unordered lists. If used with ordered lists, then it a LIST_ERROR_SEQUENCE error code will be returned if inserting the specified node at the end of the list would violate the ordering constraints.

int   LIST_Before(list_t list, node_t newnode, node_t oldnode);

It is normally expected to be used with unordered lists. If used with ordered lists, then it a LIST_ERROR_SEQUENCE error code will be returned if inserting the specified node at the specified point would violate the ordering constraints.

If the oldnode parameter is NULL, then this will act just like the LIST_Insert method and insert the new node as the first node in the list.

list_t LIST_Clone(list_t oldlist);

The value returned is a handle to the new list instance if the method was successful. If for any reason an error occurred, then NULL will be returned and the global error variable errno set to indicate the type of error that occurred. In practice as long the oldlist parameter was valid probably the only error that will ever occur is ENOMEM because you have run out of memory.

The fundamental difference between using the LIST_Create and LIST_Clone method calls is that with the LIST_Clone method both the new and the old list will share information about the list attributes such as its node size, and the user supplied functions for node initialisation, node destruction and node comparison. This means that there is significantly less memory overhead when the LIST_Clone method is used compared to when another LIST_Create method call is used. This is particularly important for situations in which you may have a very large number of occurrences of a particular list type.

An example of when this situation might arise is if you have a list, and then to each node of that list you are going to attach further lists. In this case it is much more efficient in memory terms to issue a single LIST_Create method call to create the first instance of the attached list, and then use. LIST_Clone method calls to create all the rest of the instances. In fact you may well issue a dummy LIST_Create method call as part of your program initialisation to create the first instance which you will never destroy, and then simply use LIST_Clone method calls as appropriate to create further instances of this type of list.

int  LIST_Compare(list_t list, node_t node1, node_t node2)

If this is an ordered list (the user supplied a Node Compare function when the list was created) it will return values as defined for the Node Compare function. If this is an unordered list then the comparison will be based on the physical order in the list. This means that the return values are interpreted as meaning:

    ordered:
        -ve     node1 is less than node 2
         0      node1 is equal to node 2
        +ve     node1 is greater than node 2.

    unordered:
        -ve     node1 is earlier than node2 in the list
                OR node2 is not on the list              
         0      node1 and node2 are the same node
                OR neither node1 or node2 is in the list
        +ve     node1 is later than node2 in the list
                OR node1 is not on the list

list_t LIST_Create (list_t type, size_t node_size,
                     int (*initnode)(node_t node, va_list params),
                     int (*killnode)(node_t),
                     int (*compnode)(node_t, node_t);

The type parameter can be any one of the following constants that are defined in the liblist.h header file:

       LIST_CLASS_SINGLE           Single linked list
       LIST_CLASS_DOUBLE           Double linked list
       LIST_CLASS_QUEUE            Queue
       LIST_CLASS_FIFO             FIFO (another name for queue)
       LIST_CLASS_STACK            Stack
       LIST_CLASS_LIFO             LIFO (another name for stack)
       LIST_CLASS_BTREE            Binary tree
       LIST_CLASS_BALTREE          Balanced Binary Tree
       LIST_CLASS_HASH             Hash Table

The user can optionally provide the addresses of three user supplied functions.

• The initnode node initialisation function that is to be called each time a new node is created. Can be set to NULL if not required. However, if you wish to use any of the methods that combine a LIST_NewNode method with an insertion method in a single method call then you must specify the node initialisation function.
• The killnode node destroy function that is to be called each time a node is destroyed. Can be NULL if not required.
• The compnode node compare function that is to be called to compare two nodes of a list to decide which is larger. Can be set to NULL if not required. A value other than NULL for this function means that a list is to be an ordered list, while its absence means that the list is unordered.

  It is an error to try and specify ordering for list types that by their very nature cannot be ordered. Examples of such lists are Queue/FIFO lists; Stack/LIFO lists; and Hash table lists which have no linked list.

More detail on these User Supplied Functions is given later in this document.

Note also that if you are setting up a Hash Table type of list and it later has another table linked in via a LIST_HashSetip method call, then if they are not all NULL the values of the initnode, killnode and compnode parameters specified here take precedence over any similar values used when creating the table that is linked in via the LIST_HashSetup method.

int  LIST_Delete(list_t list, node_t node);

It is effectively a combination of the LIST_Remove method to remove the node from the list, followed immediately by a LIST_FreeNode method on the same node.

node_t  LIST_Dequeue(list_t list);

It will remove the first node from the given list and return a pointer to node. If the list is empty, then NULL will be returned.

It is worth noting, however, that if you have specified that the list for which is this method is being used is a Queue/FIFO list type then the LIBLIST library will ensure that it is optimised to make removing from the end of the list efficient which might not be the case if you simply use one of the other list types and use this method on them.

The implementation of this method means that it is functionally equivalent to writing LIST_Remove(LIST_First(list))

int   LIST_Destroy(list_t list);

First each node in turn will be destroyed using the LIST_Delete method, and then when all the resources associated with the nodes have been freed, the list handle itself will be destroyed.

If this is the last instance of a cloned list, then the resources describing the information that is common between clones will also be released.

If this list is part of an embedded list then all handles associated with handling embedding list will be destroyed, including the other list handles associated with the embedded list . You must not in this case use LIST_Destroy on the other handle(s) associated with the embedded list as they will no longer be valid once a LIST_Destroy method call has succeeded on any of them.

list_t LIST_Embed (listclass_t type, list_t oldlist,
                     int (*compnode)(node_t,node_t);

The type parameter can be any one of the following constants that are defined in the liblist.h header file:

       LIST_CLASS_SINGLE           Single linked list
       LIST_CLASS_DOUBLE           Double linked list
       LIST_CLASS_QUEUE            Queue
       LIST_CLASS_FIFO             FIFO (another name for queue)
       LIST_CLASS_STACK            Stack
       LIST_CLASS_LIFO             LIFO (another name for stack)
       LIST_CLASS_BTREE            Binary tree
       LIST_CLASS_BALTREE          Balanced Binary Tree
       LIST_CLASS_HASH             Hash Table

The oldlist parameter can be any of the list types, including another embedded list, so lists can be embedded any number of levels. The oldlist must not yet contain any data or the method call will fail with errno set to the EEXIST error code.

The user can optionally provide the address of one user supplied function, or NULL if not required.

• The compnode node compare function that is to be called to compare two nodes of a list to decide which is larger. The presence of this function means that a list is to be an ordered list, while its absence means that the list is unordered. It would be very rare with an embedded list for a list to be unordered as if so why bother to embed the list at all?

The new list will use the same values for the node initialisation and the node destruction functions as were used for the list specified by the oldlist parameter. This makes sense as the actual user data part of any node is only held once even though the node is linked into two (or more) different lists simultaneously. More detail on User Supplied Functions is given later in this document.

int   LIST_Enqueue(list_t list, node_t node);

The implementation of this method means it will insert a given node at the end of the list, and it is thus functionally equivalent to the LIST_Append method.

It is worth noting, however, that if you have specified that the list for which is this method is being used is a Queue/FIFO list type then the LIBLIST library will ensure that it is optimised to make inserting at the end of the list efficient. This might not be the case if you simply use one of the other list types and use this method on them.

long  LIST_Enumerate(list_t list, (*func)(node_t, va_list), ...);

The enumeration will proceed according to whatever ordering criteria have been specified. A typical use of this function is when you want to perform an operation on every node in the list, such as for instance listing out its details.

More details on the func parameter is given in the section on user defined functions under the node enumeration functions topic.

If any additional parameters are supplied following the name of the user defined function, then these will be passed to the function every time it is invoked using the methods defined in the stdarg.h header file. However some care needs to be taken within the user defined function on the use of these parameters as they are not reset between invocations for each node. This means that they must not be changed in any way unless you want the next invocation of the function to inherit the changes.

Note that the return type of this method is long. This is so that the user supplied function can, if desired, pass back a pointer as a result code. The ISO C standard says that that the programmer is entitled to assume that the long data type can be freely cast to/from any pointer data type.

node_t  LIST_Find (list_t, int (*findfunc)(node_t,va_list), ...);

You need to provide the name of the findfunc function that is to be called for each node (to determine whether it is the one that you want or not. If any parameters follow the name of the function, then they will be passed to the function each time it is invoked using the methods defined in the stdarg.h header file. More detail on how this function should operate is given under the node find function description in the user supplied functions section later in this document.

int   LIST_FreeNode(list_t list, node_t node);

It is an error to call this function for a node which is still linked into the list. If you want to destroy a node that is still linked into a list you should use the LIST_Destroy method instead.

int   LIST_HashSetup(list_t list,
                     size_t, bucket_count,
                     size_t  (*NodeHash)(node_t, va_list),
                     size_t  (*FindHash(va_list),
                     list_t  LinkedList);

The parameters are used as follows:

• plist is a pointer to the list handle that was returned from the original LIST_Create or LIST_Clone method call used to create this particular hash table list.
  Note: the value passed is a pointer to the handle returned by LIST_Create, and that the LIST_HashSetup may change the this value.

• bucket_count is used to set the number of buckets that are to be set up for the hash table (and thus the initial amount of space that will be occupied). A value of 0 means do not change the current value. The default value set when the LIST_Create method was used to create the original instance of the list will be 127 buckets.

  This value can be anything that you like. However a point to note is that typically prime numbers are often found to give best results in avoiding hash values tending to cluster at specific points in the list.

• NodeHash is the name of a function that will calculate the hash value given a pointer to anode. A value of NULL means that the existing function should be retained. This function will be used whenever you ask for a node to be inserted into a hash table.
  Note: to ensure consistency, the NodeHash and FindHash parameters must either both be NULL, or neither one must be NULL.

  The default function supplied assumes that the first element of the user data is a pointer to a string that is the 'key' for this node.

  More detail is given in the User Defined functions section of this document under the Node Hash Function topic.

• FindHash is the name of a function that will calculate the hash value when trying to find an existing node using the LIST_Find method. A value of NULL means that the existing function should be retained.
  Note: to ensure consistency, the NodeHash and FindHash parameters must either both be NULL, or neither one must be NULL.

  It is passed the same user parameters as are passed to the LIST_Find method, so it is up to the user to ensure that any such parameters are consistent.

  The default function supplied assumes that the first user parameter to LIST_Find is a pointer to a string that is the 'key' for this node.

  More detail is given in the User Defined functions section of this document under the Node Hash Functions topic.

• LinkedList is the instance handle of a list type that is to be attached to each node. A value of NULL means do not change the current setting. A special value of LIST_HASHONLY can be used if you do not want any type of list to be linked to the hash table level. This has the implication that only one node can be stored in the bucket with a given hash value, so it would be very rare to use this value.

  A special point about treatment of the linked list is that any method call that is made on the owning hash table (such as for instance a LIST_Add that does not make sense at that level will actually be performed by first using the NodeHash function to determine what bucket should contain the node to be acted on, and then calling the method on the list_t instance that is attached to that bucket. Any relevant parameters will be passed through so that method finally called will not be aware of this pass down mechanism.

int   LIST_Insert(list_t list, node_t node);

It is normally expected to be used with unordered lists. If used with ordered lists, then it is an error if inserting the specified node at the start of the list would violate the ordering constraints.

node_t  LIST_NewNode(list_t list, ...);

This method can optionally have additional parameters, and if present they are passed to the users node initialisation function as defined in the section on User Supplied Functions.

On success a pointer to the new node is returned. If any error occurs, then NULL will be returned, and the global error variable errno set to indicate the error type.

node_t  LIST_NewAdd (list_t, ...);
node_t  LIST_NewInsert (list_t,  ...);
node_t  LIST_NewAppend (list_t, ...);
node_t  LIST_NewBefore (list_t, node_t oldnode, ...);
node_t  LIST_NewAfter (list_t, node_t oldnode, ...);
node_t  LIST_NewEnqueue (list_t, ...);
node_t  LIST_NewPush (list_t, ...);

These methods are effectively a combination of the creation of a node with the LIST_NewNode method and the corresponding list insertion method. They are just provided for convenience as you frequently want to create a new node and immediately insert it into the list.

On success these methods return a pointer to the new node. On failure they return NULL, and the errno global error variable will be set to indicate the error type. In fact the only common error is likely to be LIST_ERROR_NOMEMORY that would be returned if you run out of free memory.

All these methods can optionally take additional parameters that will (if present) be passed to the user supplied node initialisation routine in the same manner as is described under the LIST_NewNode method.

For more details read the descriptions given with the LIST_NewNode method and the various methods for inserting nodes into a list.

node_t  LIST_Peek(list_t list);

It is actually implemented as functionally equivalent to the LIST_First property, but is the term more commonly associated with handling of queues and stacks.

node_t  LIST_Pop(list_t list);

It will remove the first node from the given list and return a pointer to that node. If the list is empty, then NULL will be returned. This method is functionally equivalent to writing LIST_Remove (LIST_First(list))

node_t  LIST_Position(list_t list, size_t index);

Note that any time a node is inserted or deleted from a list, the index values associated with the remaining nodes can change, so this needs to be taken into account if using this method.

int   LIST_Push(list_t list, node_t node);

The way it is actually implemented is to simply insert the given node at the start of the list, so it is functionally equivalent to the LIST_Insert method.

node_t  LIST_Remove(list_t list, node_t node);

With unordered lists it can be used if you want to link the node back into the list at a different point.

* USER SUPPLIED FUNCTIONS

These are functions that are supplied by the user. They are called at specified points in the processing of a list. They are what is commonly known as Callback Functions in that they are calls back to user supplied routines from within the library methods. They are called to carry out a number of different purposes:

Node Initialisation Function

When a new node is created, any pointer fields are automatically set to NULL, and all other data fields set to zero. This function will be used to do any additional initialisation that is required. It is specified as a parameter to the LIST_Create method call. This function has the following prototype:

        int  initnode (node_t node, va_list);

On success this routine should return LIST_ERROR_NONE. Any other value will be treated as an error code, and passed back to the calling program.

The purpose of this function is to simply allow nodes to be initialised immediately they have been created. This could simply mean setting up parts of the user node to have given values. More advanced use of this function allows for setting up of complex lists types such as lists that themselves contain lists linked to each user node.

If the Node Initialisation function allocates any memory to be attached to the node, then you should provide a corresponding Node Destruction function to release this memory. The only exception to this rule would be if you never use the List_Delete method to destroy individual nodes or the LIST_Destroy method to destroy a list. In this case the resources will instead be automatically released when the program terminates.

Node Destruction Function

When a node is to be destroyed for any reason, this function (if supplied) is called for each node. It is specified as a parameter to the LIST_Create method call. As such it is basically a companion function to the Node Initialisation function.

The prototype of this function is:

      int   killnode (node_t node)

The normal thing that is done within this routine is the release of any resources that have been attached to this node. This might simply be malloc'ed memory that has been attached to this node, or it might be something much more significant such as destroying further lists that are attached to this node.

Node Compare Function

The Node Comparison function is only required if a list is to be ordered (i.e. maintained in some sort of specified order). The names of such functions are passed as parameters to the LIST_Create and the LIST_EMbed method calls. The fact that you have bothered to specify a Node Comparison routine (rather than passing a NULL value) means that the particular list instance being created is to be treated as ordered. The purpose of this function is to compare two nodes in the list to determine which one is to be considered the larger for ordering purposes.

This function has the following prototype:

        int  comparenode (node_t node1, node_t node2);

        -ve     node1 is less than node 2
         0      node1 is equal to node 2
        +ve     node1 is greater than node 2.

Node Enumeration Function(s)

Node Enumeration functions are used whenever you decide to enumerate a list (i.e. do something with every node in turn) using the LIST_Enumerate method call. It is likely that each different use of the LIST_Enumerate method will have a different associated node enumeration function that carries out a different action.

This function has the following prototype:

        long  enumnode (node_t node, va_list ap);

If the return value is 0, then the list handling code will simply move onto the next node in the list and call the Node Enumeration function again, repeating this until the end of the list is reached. If for any reason you return a non-zero value from this function, then the enumeration of the list will be terminated immediately, and the value returned as the result of the LIST_Enumerate method call.

Note that the return type of this function is long. This is so that the user supplied function can, if desired, pass back a pointer as a result code. The ISO C standard says that that the programmer is entitled to assume that the long data type can be freely cast to/from any pointer data type.

Node Find Function

This function is used when you wish to locate a particular node from within a list using the LIST_Find method call. The name is passed as one of the parameters to this method. The purpose of this type of function is to check whether a particular node is the one that you are trying to find.

Very often there would only be one instance of the Node Find function for each list instance as you are probably trying to find node according to the keys you used for ordering the list. There is nothing, however, that mandates this and stops you having different versions for different LIST_Find method calls.

This function has the following prototype:

        int  findnode (node_t node, va_list ap);

The return value is interpreted as follows:

            -ve      This node is not wanted, and (for ordered lists)
                     the node is less than the desired node.
             0       This is the node that is wanted
             +ve     This node is not wanted, and (for ordered lists)
                     the node is greater than the desired node.
 LIST_ORDER_NA       This node is not wanted, and nothing can be said about
                     its relationship to the desired node.  This constant
                     is defined in the liblist.h header file.

Note that if using the LIST_Find method with a hash table that the initial parameters passed to the node find function must correspond to those expected by the find hash user supplied function.

Node Hash Functions

These functions are only relevant to Hash Table types of list. They are passed as parameters to a LIST_HashSetup method call. Their purpose is to calculate the hash value that is used to determine which bucket of a hash table a node belongs to. The value returned by this function will be automatically normalised to ensure that it falls into a valid range for the number of nodes that are actually in the hash table, so it is not necessary for the hash function itself to do this.

There are two different variants of the hash function that have to be supplied:

• The node hash function used to calculate the hash value given a pointer to an already existing node. This function has the following prototype:

          size_t  hashnode (node_t node);
  

  although the name does not need to be hashnode, but can be any name that the user desires. The node parameter is a pointer to the existing node

  If the user does not supply a node hash function, then a default one will be used. This default function assumes that the definition of the user part of the node starts with a pointer to name along the lines of:

        struct {
             char * name;	/* The type is important, not the name */
             ....
             };
  

• The find hash function used to calculate the hash value when trying to find a node using the LIST_Find method. This function has the following prototype:

          size_t  hashfind (va_list ap);
  

  although the name does not need to be hashfind, but can be any name that the user desires. The ap parameter is a pointer to any additional parameters that might have been passed with the LIST_Find method call.

  If the user does not supply a node find function, then a default one will be used. This default function assumes that the first user defined parameter to the LIST_Find method is a char * parameter pointing to the null terminated string that is the name being used as a key for the node.

The other point to note is that the node hash and the find hash functions always need to be kept in step. Therefore a check will be made in the LIST_HashSetup method call that you have either specified both these routines or neither of them.

* ERROR CODES

The following is a consolidated list of the error codes that can be generated by the LIBLIST library. They will all be outside the range of normal system errors that can be returned by routines in the default C library. They also all follow standard C style by being negative constants.

For convenience of users who want to define their own error codes relative to those used by the LIBLIST library, the range of values of the error codes used by the LIBLIST library is defined by the LIST_ERROR_BASE and the LIST_ERROR_TOP constants defined in the liblist.h header file.

Error Code	Description
LIST_ERROR_NONE	The value that is returned when no error has occurred.
LIST_ERROR_BADCOMP	You have not supplied a node comparison function, and it is mandatory for the list type you are trying to create.
LIST_ERROR_BADHASH	You have not supplied consistent parameters to the LIST_HashSetup method for the user defined hash functions. You must either supply both of them or neither of them.
LIST_ERROR_BADINDEX	The value that you passed as the index value is outside the range of legal values for this list.
LIST_ERROR_BADINIT	You have not supplied a node initialisation function, and it is mandatory for the list type you are trying to create.
LIST_ERROR_BADLIST	The value that you passed as a the list handle is invalid.
LIST_ERROR_BADNODE	The value that you passed as a the address of a list node is invalid.
LIST_ERROR_BADSIZE	The value that you passed as the size of the user data in a node is invalid. This will be because the value passed is negative or zero.
LIST_ERROR_BADTYPE	The value that you passed as a the type of list is invalid. This should never occur if you stick to the constants that are defined in the liblist.h header file.
LIST_ERROR_NOMEMORY	There is not enough free memory to carry out the requested operation. In most programs this will be a fatal error condition.
LIST_ERROR_NOTEMPTY	The operation that you have requested in not allowed on a list that already contains data.
LIST_ERROR_NOTFOUND	The node that you have specified cannot be found in this list.
LIST_ERROR_NOTSETUP	You have tried to use a Hash Table type of list before you have used the LIST_HashSetup method to complete its initialisation. In practice you should rarely encounter this error code as if you have not called the LIST_HashSetup method by the time you first try to add entries into the table, the LIBLIST library will automatically try and initialise the Hash Table with a default set of values.
LIST_ERROR_ORDERED	The operation that you have requested in not allowed on ordered tables
LIST_ERROR_SEQUENCE	You have asked for a node to be inserted in an order table at a specified point, and the position that you have specified would break the ordering constraints.
LIST_ERROR_UNORDERED	The operation that you have requested in not allowed on unordered tables
LIST_ERROR_NOTREADY	You have asked for a facility that has not yet been implemented. This error code will only happen when a new facility has been added to the documentation and has not yet been implemented. Therefore it should only happen it test releases of this library. If it happens at any other time please advise the author as it may indicate a discrepancy between the documentation and the current implementation.

* EXAMPLE PROGRAM

The source to this library includes the file LISTTEST.C which is used to test the library. The source is however copiously annotated, so it is worth examining it as an example of how the various properties and methods in the library can be used.

In particular it is worth seeing how the User Supplied functions are used as this is the one area that many users initially find confusing. Once you have seen how it works, instead of being confused, you are likely to realise how it simplifies the logic of most programs to write them in this way.

* AUTHOR and COPYRIGHT

This library and associated documentation has been produced by:

      Dave Walker
       22 Kimptons Mead
        Potters Bar,
         Herts,
          EN6 3HZ
           United Kingdom

      Email:      d.j.walker@slh0101.wins.icl.co.uk
      CompuServe: 100141,2626
      MSN:        itimpi@msn.com

Every effort has been made to ensure that the LIBLIST software operates correctly as specified in this document. However no guarantee of any kind is given that the software is reliable or that the descriptions contained in this document is accurate. It is up to the user of the LIBLIST library to satisfy himself that it is suitable for the use to which it will be put.

Permission is given to freely distribute this library, including its documentation and source as long as the following terms are adhered to:

• No attempt is made to claim ownership of this material or any rights to the LIBLIST library software.
• You do not distribute the LIBLIST library as a commercial product, although reasonable charges may be made to cover your time, media costs and postage in the case of software libraries that distribute Public Domain or shareware software.
• When distributing the LIBLIST package, the complete package must be enclosed including this notice.
• The LIBLIST library may be linked into a commercial software package, but in this case no additional charge may be made for the LIBLIST library component.

* RELEASE HISTORY

May 1996: Release 1.0

First version of the library. Released via QDOS BBS network. Does not include the RLL version of the library as the RLL system has not been made generally available for QDOS based systems.

June 1996: Release 1.1

Support for hash tables and balanced binary trees completed. Embedded list feature specified in documentation, and code to support it added.