diff options
Diffstat (limited to 'wk5/pset/speller/dictionary.c')
-rw-r--r-- | wk5/pset/speller/dictionary.c | 110 |
1 files changed, 10 insertions, 100 deletions
diff --git a/wk5/pset/speller/dictionary.c b/wk5/pset/speller/dictionary.c index 7e5cc61..a20c467 100644 --- a/wk5/pset/speller/dictionary.c +++ b/wk5/pset/speller/dictionary.c @@ -2,9 +2,7 @@ #include <ctype.h> #include <stdbool.h> -#include <stdio.h> -#include <stdlib.h> -#include <string.h> + #include "dictionary.h" // Represents a node in a hash table @@ -20,125 +18,37 @@ const unsigned int N = 26; // Hash table node *table[N]; -// Size integer -int siz = 0; - // Returns true if word is in dictionary, else false bool check(const char *word) { - // hash word to find bucket - int val = hash(word); - node *current = table[val]; - - // check all nodes in the bucket - while (current != NULL) - { - // check if this is the word - if (strcmp(current->word, word) == 0) - { - return true; - } - current = current->next; - } - + // TODO return false; } // Hashes word to a number unsigned int hash(const char *word) { - // (sum of (letter - 'A') % 26) of a word to get a value of where to store it in the hash table - int val = 0; - for (int i = 0; word[i] != '\0'; i++) - { - val += toupper(word[i]) - 'A'; - } - return val %= 26; + // TODO: Improve this hash function + return toupper(word[0]) - 'A'; } // Loads dictionary into memory, returning true if successful, else false bool load(const char *dictionary) { - // initialize the table - for (int i = 0; i < N; i++) { - // malloc size of next - table[i] = NULL; - } - - // Open dictionary file - FILE *source = fopen(dictionary, "r"); - if (source == NULL) - { - return false; - } - - //read each word in dictionary - char word[LENGTH + 1]; - // use fscanf(file, "%s", word) to grab words - // check for ended file - while(fscanf(source, "%s", word) == 1) - { - // update size int - siz++; - // create new node - // use malloc - node *ptr = malloc(sizeof(node)); - // check if return is NULL - if (ptr == NULL) - { - fclose(source); - return false; - } - // copy word from fscanf into node using strcpy - strcpy(ptr->word, word); - // set the new ptr - ptr->next = NULL; - // hash the word to find the bucket it goes in - int val = hash(ptr->word); - // put new node at begining of bucket - if (table[val] == NULL) - { - // if empty put it there - table[val] = ptr; - } - else - { - // if not empty move the current first one down, - // then put the new one there - ptr->next = table[val]; - table[val] = ptr; - } - - } - - // Close the dictionary file - fclose(source); - return true; + // TODO + return false; } // Returns number of words in dictionary if loaded, else 0 if not yet loaded unsigned int size(void) { - return siz; + // TODO + return 0; } // Unloads dictionary from memory, returning true if successful, else false bool unload(void) { - // for every bucket - for (int i = 0; i < N; i++) - { - // while there's more in the bucket - node *current = table[i]; - while (current != NULL) - { - // record position of this node - node *this = current; - // record position of next node - current = current->next; - // free this node - free(this); - } - } - return true; + // TODO + return false; } |