Lookup table with fixed array

Question

I implemented the following excercise:

Implement a lookup table with operations such as find(struct table*, const char*), insert(struct table*, const char*,int), and remove(struct table*, const char*). The representation of the table could be an array of a struct pair or a pair of arrays (const char*[] and int*); you choose, also choose return types for your functions.

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#define ARR_SIZE 10
struct Pair {
 const char* word;
 int val;
};
struct Table {
 struct Pair* pairs[ARR_SIZE];
 size_t sz;
};
struct Pair* make_pair(const char* word, int val)
{
 assert(word);
 struct Pair* pair = (struct Pair*)malloc(sizeof(struct Pair));
 pair->val = val;
 pair->word = word;
 return pair;
}
void table_empty(struct Table* tbl)
 {
 assert(tbl);
 size_t i = 0;
 for (i = 0; i < tbl->sz; ++i) {
 free(tbl->pairs[i]);
 tbl->pairs[i] = NULL;
 }
 tbl->sz = 0;
}
int search_word(struct Table* tbl, const char* word)
{
 assert(tbl);
 assert(word);
 size_t i = 0;
 for (i = 0; i < tbl->sz; ++i) {
 //printf("%s %i\n", tbl->pairs[i]->word,i);
 if (strcmp(tbl->pairs[i]->word, word) == 0) {
 return i;
 }
 }
 return -1; // error
}
void table_insert(struct Table* tbl, const char* word, int val)
{
 assert(tbl);
 assert(word);
 int i = search_word(tbl, word);
 if (i != -1) { // replace val
 tbl->pairs[i]->val = val; 
 }
 else { // add new pair
 struct Pair* pair = make_pair(word, val);
 tbl->pairs[tbl->sz] = pair; // add pair at the last position
 ++tbl->sz;
 }
}
int table_find(struct Table* tbl, const char* word, int* return_val)
{
 assert(tbl);
 assert(word);
 assert(return_val);
 int i = search_word(tbl, word);
 if (i != -1) {
 *return_val = tbl->pairs[i]->val;
 return 0;
 }
 return -1; // error not found
}
int table_remove(struct Table* tbl, const char* word)
{
 assert(tbl);
 assert(word);
 int i = search_word(tbl, word);
 if (i == -1) return -1;
 free(tbl->pairs[i]); // free value at current pos
 tbl->pairs[i] = tbl->pairs[tbl->sz - 1]; // put address of last word at the pos of the current
 --tbl->sz; // "erase" last word
 return 0;
}
void table_print(struct Table* tbl)
{
 assert(tbl);
 printf("\n");
 printf("table size = %i\n", tbl->sz);
 for (int i = 0; i < tbl->sz; ++i)
 printf("%s %i\n", tbl->pairs[i]->word, tbl->pairs[i]->val);
 fflush(stdout);
}
void print_search_result(struct Table* tbl, const char* word)
{
 assert(tbl);
 assert(word);
 int val = 0;
 if (table_find(tbl, word, &val) == 0) 
 printf("%s %i\n",word, val);
 else 
 printf("%s not found in table\n", word);
 printf("\n");
 fflush(stdout);
}
void print_remove_result(struct Table* tbl, const char* word)
{
 assert(tbl);
 assert(word);
 if (table_remove(tbl, word) == -1) 
 printf("%s not deleted\n", word);
 else 
 printf("%s deleted\n", word);
 printf("\n");
 fflush(stdout);
}
int main()
{
 struct Table table = { 0 };
 int val = 0;
 table_insert(&table, "Hello", 10);
 table_insert(&table, "Test", 15);
 table_insert(&table, "Hello", 18); // testing overrite val
 table_insert(&table, "What", 5);
 table_insert(&table, "is", 3);
 table_insert(&table, "going", 4);
 table_insert(&table, "on", 77);
 table_print(&table);
 print_search_result(&table, "Hello");
 print_search_result(&table, "Test");
 print_search_result(&table, "keyword");
 print_remove_result(&table, "Hello");
 print_remove_result(&table, "Hello"); // double delete == false
 print_remove_result(&table, "What");
 print_remove_result(&table, "going");
 print_remove_result(&table, "is");
 print_remove_result(&table, "on");
 print_remove_result(&table, "Test");
 table_print(&table); 
 table_insert(&table, "Hello", 10);
 table_insert(&table, "Test", 15);
 table_insert(&table, "Hello", 18); // testing overrite val
 table_insert(&table, "What", 5);
 table_insert(&table, "is", 3);
 table_insert(&table, "going", 4);
 table_insert(&table, "on", 77);
 table_print(&table);
 table_empty(&table);
 table_print(&table);
 getchar();
}

Feel free to comment on any improvements. Is there a better way to do this?

I have one specific question also: are my uses of assert appropriate?

• Jakob: As a stylistic nitpick, I'd recommend the * in a pointer type be adjacent to the variable name (e.g. const char *word). The C type syntax was designed to mimic usage, where for example you'd type *word to access a const char element. Note also that this makes it more clear what type b is in int *a, b;.

Answer 1

Using struct

Personally, I think you made the right choice by using a struct with a char* and int in it rather than 2 arrays of char* and int, respectively. The Pair is a conceptual unit of data in your app, so it makes sense to put them together. If you had 2 separate arrays, it would be easy for them to get out of synch and hard to debug why that happened. Nice work!

Prefer const and functions to macros

You've defined the array size using a macro. (削除) This puts you at a disadvantage while programming. You've removed the type information for the value. By making it:

const int ARR_SIZE = 10;

you get type safety. (削除ここまで) Edit: That's a C++-ism that doesn't work in C. My bad! But the rest of the advice in the next paragraph is correct as far as I know.

With macros that take parameters you run the risk of them being used in unexpected ways and causing hard to debug issues. Luckily, you haven't done that here. But in general, if you find yourself reaching for a macro, ask yourself if you would be better off with a constant or a function. You almost always will be (assuming you can use the constant in the desired way).

Errors

There are some errors in your code. In make_pair(), you don't check to see if malloc() succeeded. If it fails, no memory is allocated, and pair points to NULL. When you try to assign pair->val or pair->word you will crash.

If you did fix that, table_insert() uses the result of make_pair() without checking to see if it's NULL first. This won't crash immediately, because you're just assigning the array pointer in tbl->pairs[tbl->sz] to have the value of pair. What will happen is later when you try to search or print or insert another item, you'll get a crash when you iterate over that entry in the table and try to do anything with it.

You could make these errors not possible by not dynamically allocating the array entries. Simply make the array an array of Pair structs rather than a pointer to them.

Naming

A lot of your names are really good. Pair and Table are decent, readable names for this task. make_pair(), table_insert(), etc. are informative. Some could be improved, though. tbl does not save you much typing over table. Just use the whole word. Same with sz vs. size. i is acceptable as a loop variable name, but it would be even better if it was more descriptive. For example, entry_index or pair_index. It should describe what you're iterating over.

• Sandro4912: when i change the define to const int ARR_SIZE = 10; it gives me an error in struct Table { struct Pair* pairs[ARR_SIZE]; size_t sz; }; that ARR_SIZE is not const, So how to use it in this case then?
• user1118321: Sorry - that's a C++-ism. I thought this was a C++ question. My bad. I'll update my answer.
• Jakob: I'm a staunch supporter of descriptive variable names, but with the simplicity of the loops in this program I think the variable name i works just fine.
• Toby Speight: If malloc() fails, assigning to pair->val or pair->word might crash, if you are lucky. It might just continue and give wrong results, or it might do something totally unexpected. That's the joy of Undefined Behaviour!

Answer 2

• Do not cast the result of malloc.

  malloc returns void *, and in C it is valid to convert void * to any pointer. If you cast to suppress a warning about integer to pointer conversion, it means that a compiler doesn't have a malloc prototype, and defaults its to return an int (it is an ancient C convention). Now if int and pointer are of different size, the malloc return value would be truncated, with all nasty consequences.

• It is not recommended to sizeof(type). Prefer sizeof(expression). In your case, consider

   struct Pair * pair = malloc(sizeof(*pair));
  

• table_insert blindly inserts into a full table. It should test for tbl->sz < ARR_SIZE and return an error indication if it is not.

• The actual insertion

   struct Pair* pair = make_pair(word, val);
   tbl->pairs[tbl->sz] = pair; // add pair at the last position
  

  should be really a single line:

   tbl->pairs[tbl->sz] = make_pair(word, val);
  

• Sandro4912: why is it not a good idea to cast malloc? In c++ it even would get a compiler error?
• vnp: @Sandro4912 The question is tagged C. C++ is a different language, particularly in this respect. Now if the C compiler complains, it means that the malloc prototype is missing, which may lead to nasty problems.
• Sandro4912: yes i know is c. i was just wondered why here its not a good practice to indicate the type changes with a cast
• vnp: @Sandro4912 See edit.
• chux: @Sandro4912 pair = malloc(sizeof *pair) is easier to code correctly the first time, easier to review and maintain.

Answer 3

Enable all compiler warnings

With a well enabled compiler, for (int i = 0; i < tbl->sz; ++i) should have warned about mixing sign-ness of i and tbl->sz as well as range. Save time and enable all warnings and use for (size_t i = 0; i < tbl->sz; ++i).

In general, code is squishy on using int,size_t nearly interchangeably. I'd re-architect and use size_t only.

Mixed use of shallow and deep copy

make_pair(const char* word, int val) allocates and forms a whole new struct Pair (deep copy), yet does not copy what word points to.

Perhaps

// pair->word = word;
pair->word = strdup(word);

Use const

search_word() does not modify *tbl, so use const to convey that. Same fortable_find(),table_print(),print_search_result()`.

// int search_word(struct Table* tbl, const char* word)
int search_word(const struct Table* tbl, const char* word)

Naming

Code uses const char* word; yet it is not a "word", but a string as used in strcmp().

----- Additions

Contract violation?

Nothing in the requirements "Implement a lookup table with operations such as ..." indicate that const char* is a pointer to a string. So calling strcmp() is questionable unless unstated requirements exist. As a char *, code could use a simple compare

// if (strcmp(tbl->pairs[i]->word, word) == 0)
if (tbl->pairs[i]->word == word)

Use of assert()

are my uses of assert appropriate?

If the char * pointer to add/search is not specified to be a string, assert(word); is not proper as word == NULL is not known to be invalid.

The assert(return_val) in table_find(struct Table* tbl, const char* word, int* return_val) is OK, yet I would re-design to allow return_val == NULL

if (i != -1) {
 // add
 if (return_val) {
 *return_val = tbl->pairs[i]->val;
 } 
 return 0;
}