learn-c-the-hard-waych33.txt

   Learn C The Hard Way A Learn Code The Hard Way Book
     * Book
     * Comments
     * Video Courses
     * Related Books

   [next] [prev] [prev-tail] [tail] [up]

Chapter 33
Exercise 32: Double Linked Lists

   The purpose of this book is to teach you how your computer really
   works, and included in that is how various data structures and
   algorithms function. Computers by themselves don't do a lot of useful
   processing. To make them do useful things you need to structure the
   data and then organize processing on these structures. Other
   programming languages either include libraries that implement all of
   these structures, or they have direct syntax for them. C makes you
   implement all the data structures you need yourself, which makes it the
   perfect language to learn how they actually work.

   My goal in teaching you these data structures and these algorithms is
   to help you do three things:
    1. Understand what is really going on in Python, Ruby, or JavaScript
       code like: data = {"name": "Zed"}
    2. Get even better at C code by applying what you know to a set of
       solved problems using the data structures.
    3. Learn a core set of data structures and algorithms so that you are
       better informed about what ones work best in certain situations.

33.1 What Are Data Structures

   The name "data structure" is self-explanatory. It is an organization of
   data that fits a certain model. Maybe the model is designed to allow
   processing the data in a new way. Maybe it's just organized to store it
   on disk efficiently. In this book I'll follow a simple pattern for
   making data structures that works reliably:

    1. Define a struct for the main "outer structure".
    2. Define a struct for the contents, usually nodes with links between
       them.
    3. Create functions that operate on these two.

   There's other styles of data structures in C, but this pattern works
   well and is consistent for most data structures you'll make.

33.2 Making The Library

   For the rest of this book you'll be creating a library that you can use
   when you're done with this book. This library will have the following
   elements:

    1. Header (.h) files for each data structure.
    2. Implementation (.c) files for the algorithms.
    3. Unit tests that test all of them to make sure they keep working.
    4. Documentation we'll autogenerate from the header files.

   You already have the c-skeleton so use it to create a liblcthw project:
     __________________________________________________________________

   Source 93: ex32.sh-session
   1  $ cp -r c-skeleton liblcthw
   2  $ cd liblcthw/
   3  $ ls
   4  LICENSE Makefile README.md bin build src tests
   5  $ vim Makefile
   6  $ ls src/
   7  dbg.h libex29.c libex29.o
   8  $ mkdir src/lcthw
   9  $ mv src/dbg.h src/lcthw
   10  $ vim tests/minunit.h
   11  $ rm src/libex29.* tests/libex29*
   12  $ make clean
   13  rm -rf build  tests/libex29_tests
   14  rm -f tests/tests.log
   15  find . -name "*.gc*" -exec rm {} \;
   16  rm -rf `find . -name "*.dSYM" -print`
   17  $ ls tests/
   18  minunit.h runtests.sh
   19  $
     __________________________________________________________________

   In this session I'm doing the following:
    1. Copy the c-skeleton over.
    2. Edit the Makefile to change libYOUR_LIBRARY.a to liblcthw.a as the
       new TARGET.
    3. Make the src/lcthw directory where we'll put our code.
    4. Move the src/dbg.h into this new directory.
    5. Edit tests/minunit.h so that it uses #include <lcthw/dbg.h> as the
       include.
    6. Get rid of the source and test files we don't need for libex29.*.
    7. Clean up everything that's left over.

   With that you're ready to start building the library, and the first
   data structure I'll build is the Double Linked List.

33.3 Double Linked Lists

   The first data structure we'll add to liblcthw is a double linked list.
   This is the simplest data structure you can make, and it has useful
   properties for certain operations. A linked list works by nodes having
   pointers to their next or previous element. A "double linked list"
   contains pointers to both, while a "single linked list" only points at
   the next element.

   Because each node has pointers to the next and previous, and because
   you keep track of the first and last element of the list, you can do
   some operations very quickly. Anything that involves inserting or
   deleting an element will be very fast. They are also easy to implement
   by most people.

   The main disadvantage of a linked list is that traversing it involves
   processing every single pointer along the way. This means that
   searching, most sorting, or iterating over the elements will be slow.
   It also means that you can't really jump to random parts of the list.
   If you had an array of elements you could just index right into the
   middle of the list, but a linked list uses a stream of pointers. That
   means if you want the 10th element, you have to go through elements
   1-9.

33.3.1 Definition

   As I said in the introduction to this exercise, the process to follow
   is to first write a header file with the right C struct statements in
   it.
     __________________________________________________________________

   Source 94: src/lcthw/list.h
   1  #ifndef lcthw_List_h
   2  #define lcthw_List_h
   3
   4  #include <stdlib.h>
   5
   6  struct ListNode;
   7
   8  typedef struct ListNode {
   9      struct ListNode *next;
   10      struct ListNode *prev;
   11      void *value;
   12  } ListNode;
   13
   14  typedef struct List {
   15      int count;
   16      ListNode *first;
   17      ListNode *last;
   18  } List;
   19
   20  List *List_create();
   21  void List_destroy(List *list);
   22  void List_clear(List *list);
   23  void List_clear_destroy(List *list);
   24
   25  #define List_count(A) ((A)->count)
   26  #define List_first(A) ((A)->first != NULL ? (A)->first->value : NUL
   L)
   27  #define List_last(A) ((A)->last != NULL ? (A)->last->value : NULL)
   28
   29  void List_push(List *list, void *value);
   30  void *List_pop(List *list);
   31
   32  void List_shift(List *list, void *value);
   33  void *List_unshift(List *list);
   34
   35  void *List_remove(List *list, ListNode *node);
   36
   37  #define LIST_FOREACH(L, S, M, V) ListNode *_node = NULL;\
   38      ListNode *V = NULL;\
   39      for(V = _node = L->S; _node != NULL; V = _node = _node->M)
   40
   41  #endif
     __________________________________________________________________

   The first thing I do is create two structs for the ListNode and the
   List that will contain those nodes. This creates the data structure
   I'll use in the functions and macros I define after that. If you read
   through these functions they seem rather simple. I'll be explaining
   them when I cover the implementation, but hopefully you can guess what
   they do.

   How the data structure works is each ListNode has three components:
    1. A value, which is a pointer to anything and stores the thing we
       want to put in the list.
    2. A ListNode *next pointer which points at another ListNode that
       holds the next element in the list.
    3. A ListNode *prev that holds the previous element. Complex right?
       Calling the previous thing "previous". I could have used "anterior"
       and "posterior" but only a jerk would do that.

   The List struct is then nothing more than a container for these
   ListNode structs that have been linked together in a chain. It keeps
   track of the count, first and last element of the list.

   Finally, take a look at src/lcthw/list.h:37 where I define the
   LIST_FOREACH macro. This is a common idiom where you make a macro that
   generates iteration code so people can't mess it up. Getting this kind
   of processing right can be difficult with data structures, so writing
   macros helps people out. You'll see how I use this when I talk about
   the implementation.

33.3.2 Implementation

   Once you understand that, you mostly understand how a double linked
   list works. It is nothing more than nodes with two pointers to the next
   and previous element of the list. You can then write the
   src/lcthw/list.c code to see how each operation is implemented.
     __________________________________________________________________

   Source 95: src/lcthw/list.c
   1  #include <lcthw/list.h>
   2  #include <lcthw/dbg.h>
   3
   4  List *List_create()
   5  {
   6      return calloc(1, sizeof(List));
   7  }
   8
   9  void List_destroy(List *list)
   10  {
   11      LIST_FOREACH(list, first, next, cur) {
   12          if(cur->prev) {
   13              free(cur->prev);
   14          }
   15      }
   16
   17      free(list->last);
   18      free(list);
   19  }
   20
   21
   22  void List_clear(List *list)
   23  {
   24      LIST_FOREACH(list, first, next, cur) {
   25          free(cur->value);
   26      }
   27  }
   28
   29
   30  void List_clear_destroy(List *list)
   31  {
   32      List_clear(list);
   33      List_destroy(list);
   34  }
   35
   36
   37  void List_push(List *list, void *value)
   38  {
   39      ListNode *node = calloc(1, sizeof(ListNode));
   40      check_mem(node);
   41
   42      node->value = value;
   43
   44      if(list->last == NULL) {
   45          list->first = node;
   46          list->last = node;
   47      } else {
   48          list->last->next = node;
   49          node->prev = list->last;
   50          list->last = node;
   51      }
   52
   53      list->count++;
   54
   55  error:
   56      return;
   57  }
   58
   59  void *List_pop(List *list)
   60  {
   61      ListNode *node = list->last;
   62      return node != NULL ? List_remove(list, node) : NULL;
   63  }
   64
   65  void List_shift(List *list, void *value)
   66  {
   67      ListNode *node = calloc(1, sizeof(ListNode));
   68      check_mem(node);
   69
   70      node->value = value;
   71
   72      if(list->first == NULL) {
   73          list->first = node;
   74          list->last = node;
   75      } else {
   76          node->next = list->first;
   77          list->first->prev = node;
   78          list->first = node;
   79      }
   80
   81      list->count++;
   82
   83  error:
   84      return;
   85  }
   86
   87  void *List_unshift(List *list)
   88  {
   89      ListNode *node = list->first;
   90      return node != NULL ? List_remove(list, node) : NULL;
   91  }
   92
   93  void *List_remove(List *list, ListNode *node)
   94  {
   95      void *result = NULL;
   96
   97      check(list->first && list->last, "List is empty.");
   98      check(node, "node can't be NULL");
   99
   100      if(node == list->first && node == list->last) {
   101          list->first = NULL;
   102          list->last = NULL;
   103      } else if(node == list->first) {
   104          list->first = node->next;
   105          check(list->first != NULL, "Invalid list, somehow got a fi
   rst that is NULL.");
   106          list->first->prev = NULL;
   107      } else if (node == list->last) {
   108          list->last = node->prev;
   109          check(list->last != NULL, "Invalid list, somehow got a nex
   t that is NULL.");
   110          list->last->next = NULL;
   111      } else {
   112          ListNode *after = node->next;
   113          ListNode *before = node->prev;
   114          after->prev = before;
   115          before->next = after;
   116      }
   117
   118      list->count--;
   119      result = node->value;
   120      free(node);
   121
   122  error:
   123      return result;
   124  }
     __________________________________________________________________

   I then implement all of the operations on a double linked list that
   can't be done with simple macros. Rather than cover every tiny little
   line of this file, I'm going to give high-level overview of every
   operation in both the list.h and list.c file, then leave you to read
   the code.

   list.h:List_count
          Returns the number of elements in the list, which is maintained
          as elements are added and removed.

   list.h:List_first
          Returns the first element of the list, but does not remove it.

   list.h:List_last
          Returns the last element of the list, but does not remove it.

   list.h:LIST_FOREACH
          Iterates over the elements in the list.

   list.c:List_create
          Simply creates the main List struct.

   list.c:List_destroy
          Destroys a List and any elements it might have.

   list.c:List_clear
          Convenience function for freeing the values in each node, not
          the nodes.

   list.c:List_clear_destroy
          Clears and destroys a list. It's not very efficient since it
          loops through them twice.

   list.c:List_push
          The first operation that demonstrates the advantage of a linked
          list. It adds a new element to the end of the list, and because
          that's just a couple of pointer assignments, does it very fast.

   list.c:List_pop
          The inverse of List_push, this takes the last element off and
          returns it.

   list.c:List_shift
          The other thing you can easily do to a linked list is add
          elements to the front of the list very fast. In this case I call
          that List_shift for lack of a better term.

   list.c:List_unshift
          Just like List_pop, this removes the first element and returns
          it.

   list.c:List_remove
          This is actually doing all of the removal when you do List_pop
          or List_unshift. Something that seems to always be difficult in
          data structures is removing things, and this function is no
          different. It has to handle quite a few conditions depending on
          if the element being removed is at the front; the end; both
          front and end; or middle.

   Most of these functions are nothing special, and you should be able to
   easily digest this and understand it from just the code. You should
   definitely focus on how the LIST_FOREACH macro is used in List_destroy
   so you can understand how much it simplifies this common operation.

33.4 Tests

   After you have those compiling it's time to create the test that makes
   sure they operate correctly.
     __________________________________________________________________

   Source 96: tests/list_tests.c
   1  #include "minunit.h"
   2  #include <lcthw/list.h>
   3  #include <assert.h>
   4
   5  static List *list = NULL;
   6  char *test1 = "test1 data";
   7  char *test2 = "test2 data";
   8  char *test3 = "test3 data";
   9
   10
   11  char *test_create()
   12  {
   13      list = List_create();
   14      mu_assert(list != NULL, "Failed to create list.");
   15
   16      return NULL;
   17  }
   18
   19
   20  char *test_destroy()
   21  {
   22      List_clear_destroy(list);
   23
   24      return NULL;
   25
   26  }
   27
   28
   29  char *test_push_pop()
   30  {
   31      List_push(list, test1);
   32      mu_assert(List_last(list) == test1, "Wrong last value.");
   33
   34      List_push(list, test2);
   35      mu_assert(List_last(list) == test2, "Wrong last value");
   36
   37      List_push(list, test3);
   38      mu_assert(List_last(list) == test3, "Wrong last value.");
   39      mu_assert(List_count(list) == 3, "Wrong count on push.");
   40
   41      char *val = List_pop(list);
   42      mu_assert(val == test3, "Wrong value on pop.");
   43
   44      val = List_pop(list);
   45      mu_assert(val == test2, "Wrong value on pop.");
   46
   47      val = List_pop(list);
   48      mu_assert(val == test1, "Wrong value on pop.");
   49      mu_assert(List_count(list) == 0, "Wrong count after pop.");
   50
   51      return NULL;
   52  }
   53
   54  char *test_shift()
   55  {
   56      List_shift(list, test1);
   57      mu_assert(List_first(list) == test1, "Wrong first value.");
   58
   59      List_shift(list, test2);
   60      mu_assert(List_first(list) == test2, "Wrong first value");
   61
   62      List_shift(list, test3);
   63      mu_assert(List_first(list) == test3, "Wrong last value.");
   64      mu_assert(List_count(list) == 3, "Wrong count on shift.");
   65
   66      return NULL;
   67  }
   68
   69  char *test_remove()
   70  {
   71      // we only need to test the middle remove case since push/shift

   72      // already tests the other cases
   73
   74      char *val = List_remove(list, list->first->next);
   75      mu_assert(val == test2, "Wrong removed element.");
   76      mu_assert(List_count(list) == 2, "Wrong count after remove.");
   77      mu_assert(List_first(list) == test3, "Wrong first after remove.
   ");
   78      mu_assert(List_last(list) == test1, "Wrong last after remove.")
   ;
   79
   80      return NULL;
   81  }
   82
   83
   84  char *test_unshift()
   85  {
   86      char *val = List_unshift(list);
   87      mu_assert(val == test3, "Wrong value on unshift.");
   88
   89      val = List_unshift(list);
   90      mu_assert(val == test1, "Wrong value on unshift.");
   91      mu_assert(List_count(list) == 0, "Wrong count after unshift.");
   92
   93      return NULL;
   94  }
   95
   96
   97
   98  char *all_tests() {
   99      mu_suite_start();
   100
   101      mu_run_test(test_create);
   102      mu_run_test(test_push_pop);
   103      mu_run_test(test_shift);
   104      mu_run_test(test_remove);
   105      mu_run_test(test_unshift);
   106      mu_run_test(test_destroy);
   107
   108      return NULL;
   109  }
   110
   111  RUN_TESTS(all_tests);
     __________________________________________________________________

   This test simply goes through every operation and makes sure it works.
   I use a simplification in the test where I create just one List *list
   for the whole program, then have the tests work on it. This saves the
   trouble of building a List for every test, but it could mean that some
   tests only pass because of how the previous test ran. In this case I
   try to make each test keep the list clear or actually use the previous
   test's results.

33.5 What You Should See

   If you did everything right, then when you do a build and run the unit
   tests it should look like this:
     __________________________________________________________________

   Source 97: Ex32 Session
   1  $ make
   2  cc -g -O2 -Wall -Wextra -Isrc -rdynamic -DNDEBUG  -fPIC   -c -o src/
   lcthw/list.o src/lcthw/list.c
   3  ar rcs build/liblcthw.a src/lcthw/list.o
   4  ranlib build/liblcthw.a
   5  cc -shared -o build/liblcthw.so src/lcthw/list.o
   6  cc -g -O2 -Wall -Wextra -Isrc -rdynamic -DNDEBUG  build/liblcthw.a
     tests/list_tests.c   -o tests/list_tests
   7  sh ./tests/runtests.sh
   8  Running unit tests:
   9  ----
   10  RUNNING: ./tests/list_tests
   11  ALL TESTS PASSED
   12  Tests run: 6
   13  tests/list_tests PASS
   14  $
     __________________________________________________________________

   Make sure 6 tests ran, that it builds without warnings or errors, and
   that it's making the build/liblcthw.a and build/liblcthw.so files.

33.6 How To Improve It

   Instead of breaking this, I'm going to tell you how to improve the
   code:

    1. You can make List_clear_destroy more efficient by using
       LIST_FOREACH and doing both free calls inside one loop.
    2. You can add asserts for preconditions that it isn't given a NULL
       value for the List *list parameters.
    3. You can add invariants that check the list's contents are always
       correct, such as count is never < 0, and if count > 0 then first
       isn't NULL.
    4. You can add documentation to the header file in the form of
       comments before each struct, function, and macro that describes
       what it does.

   These amount to going through the defensive programming practices I
   talked about and "hardening" this code against flaws or improving
   usability. Go ahead and do these things, then find as many other ways
   to improve the code.

33.7 Extra Credit

    1. Research double vs. single linked lists and when one is preferred
       over the other.
    2. Research the limitations of a double linked list. For example,
       while they are efficient for inserting and deleting elements, they
       are very slow for iterating over them all.
    3. What operations are missing that you can imagine needing? Some
       examples are copying, joining, splitting. Implement these
       operations and write the unit tests for them.

   [next] [prev] [prev-tail] [front] [up]
     __________________________________________________________________

   Please enable JavaScript to view the comments powered by Disqus.

Take An Online Video Course

   You can sign up for a video course at:

   http://www.udemy.com/learn-c-the-hard-way/

   This course is currently being built at the same time that the book is
   being built, but if you sign up now then you get early access to both
   the videos and PDF of the book.

Related Books

   You might want to check out these other books in the series:
    1. Learn Ruby The Hard Way
    2. Learn Regex The Hard Way
    3. Learn SQL The Hard Way
    4. Learn C The Hard Way
    5. Learn Python The Hard Way

   I'll be referencing other books shortly.

   Copyright 2011 Zed A. Shaw. All Rights Reserved.