Descartes of Borg
November 1993

Chapter 3: Complex Data Types

3.1 Simple Data Types

In the textbook LPC Basics, you learned about the common, basic LPC data types: int, string, object, void. Most important you learned that many operations and functions behave differently based on the data type of the variables upon which they are operating. Some operators and functions will even give errors if you use them with the wrong data types. For example, "a"+"b" is handled much differently than 1+1. When you ass "a"+"b", you are adding "b" onto the end of "a" to get "ab". On the other hand, when you add 1+1, you do not get 11, you get 2 as you would expect.

I refer to these data types as simple data types, because they atomic in that they cannot be broken down into smaller component data types. The object data type is a sort of exception, but you really cannot refer individually to the components which make it up, so I refer to it as a simple data type.

This chapter introduces the concept of the complex data type, a data type which is made up of units of simple data types. LPC has two common complex data types, both kinds of arrays. First, there is the traditional array which stores values in consecutive elements accessed by a number representing which element they are stored in. Second is an associative array called a mapping. A mapping associates to values together to allow a more natural access to data.

3.2 The Values NULL and 0

Before getting fully into arrays, there first should be a full understanding of the concept of NULL versus the concept of 0. In LPC, a null value is represented by the integer 0. Although the integer 0 and NULL are often freely interchangeable, this interchangeability often leads to some great confusion when you get into the realm of complex data types. You may have even encountered such confusion while using strings.

0 represents a value which for integers means the value you add to another value yet still retain the value added. This for any addition operation on any data type, the ZERO value for that data type is the value that you can add to any other value and get the original value. Thus: A plus ZERO equals A where A is some value of a given data type and ZERO is the ZERO value for that data type. This is not any sort of official mathematical definition. There exists one, but I am not a mathematician, so I have no idea what the term is. Thus for integers, 0 is the ZERO value since 1 + 0 equals 1.

NULL, on the other hand, is the absence of any value or meaning. The LPC driver will interpret NULL as an integer 0 if it can make sense of it in that context. In any context besides integer addition, A plus NULL causes an error. NULL causes an error because adding valueless fields in other data types to those data types makes no sense.

Looking at this from another point of view, we can get the ZERO value for strings by knowing what added to "a" will give us "a" as a result. The answer is not 0, but instead "". With integers, interchanging NULL and 0 was acceptable since 0 represents no value with respect to the integer data type. This interchangeability is not true for other data types, since their ZERO values do not represent no value. Namely, "" represents a string of no length and is very different from 0.

When you first declare any variable of any type, it has no value. Any data type except integers therefore must be initialized somehow before you perform any operation on it. Generally, initialization is done in the create() function for global variables, or at the top of the local function for local variables by assigning them some value, often the ZERO value for that data type. For example, in the following code I want to build a string with random words:

    string build_nonsense() {
        string str;
        int i;

        str = ""; /* Here str is initialized to the string ZERO value */
        for(i=0; i<6; i++) {
            switch(random(3)+1) {
                case 1: str += "bing"; break;
                case 2: str += "borg"; break;
                case 3: str += "foo"; break;
            }
            if(i==5) str += ".\n";
            else str += " ";
        }
        return capitalize(str);
    }
If we had not initialized the variable str, an error would have resulted from trying to add a string to a NULL value. Instead, this code first initializes str to the ZERO value for strings, "". After that, it enters a loop which makes 6 cycles, each time randomly adding one of three possible words to the string. For all words except the last, an additional blank character is added. For the last word, a period and a return character are added. The function then exits the loop, capitalizes the nonsense string, then exits.

3.3 Arrays in LPC

An array is a powerful complex data type of LPC which allows you to access multiple values through a single variable. For instance, Nightmare has an indefinite number of currencies in which players may do business. Only five of those currencies, however, can be considered hard currencies. A hard currency for the sake of this example is a currency which is readily exchangeable for any other hard currency, whereas a soft currency may only be bought, but not sold. In the bank, there is a list of hard currencies to allow bank keepers to know which currencies are in fact hard currencies. With simple data types, we would have to perform the following nasty operation for every exchange transaction:

    int exchange(string str) {
        string from, to;
        int amt;

        if(!str) return 0;
        if(sscanf(str, "%d %s for %s", amt, from, to) != 3)
          return 0;
        if(from != "platinum" && from != "gold" && from !=
          "silver" &&
          from != "electrum" && from != "copper") {
            notify_fail("We do not buy soft currencies!\n");
            return 0;
        }
        ...
    }
With five hard currencies, we have a rather simple example. After all it took only two lines of code to represent the if statement which filtered out bad currencies. But what if you had to check against all the names which cannot be used to make characters in the game? There might be 100 of those; would you want to write a 100 part if statement? What if you wanted to add a currency to the list of hard currencies? That means you would have to change every check in the game for hard currencies to add one more part to the if clauses. Arrays allow you simple access to groups of related data so that you do not have to deal with each individual value every time you want to perform a group operation.

As a constant, an array might look like this:

    ({ "platinum", "gold", "silver", "electrum", "copper" })
which is an array of type string. Individual data values in arrays are called elements, or sometimes members. In code, just as constant strings are represented by surrounding them with "", constant arrays are represented by being surrounded by ({ }), with individual elements of the array being separated by a ,.

You may have arrays of any LPC data type, simple or complex. Arrays made up of mixes of values are called arrays of mixed type. In most LPC drivers, you declare an array using a throw-back to C language syntax for arrays. This syntax is often confusing for LPC coders because the syntax has a meaning in C that simply does not translate into LPC. Nevertheless, if we wanted an array of type string, we would declare it in the following manner:

    string *arr;
In other words, the data type of the elements it will contain followed by a space and an asterisk. Remember, however, that this newly declared string array has a NULL value in it at the time of declaration.

3.4 Using Arrays

You now should understand how to declare and recognize an array in code. In order to understand how they work in code, let's review the bank code, this time using arrays:
    string *hard_currencies;

    int exchange(string str) {
        string from, to;
        int amt;

        if(!str) return 0;
        if(sscanf(str, "%d %s for %s", amt, from, to) != 3)
    return 0;
        if(member_array(from, hard_currencies) == -1) {
            notify_fail("We do not buy soft currencies!\n");
            return 0;
        }
        ...
    }
This code assumes hard_currencies is a global variable and is initialized in create() as:
    hard_currencies = ({ "platinum", "gold", "electrum", "silver",
        "copper" });
Ideally, you would have hard currencies as a #define in a header file for all objects to use, but #define is a topic for a later chapter.

Once you know what the member_array() efun does, this method certainly is much easier to read as well as is much more efficient and easier to code. In fact, you can probably guess what the member_array() efun does: It tells you if a given value is a member of the array in question. Specifically here, we want to know if the currency the player is trying to sell is an element in the hard_curencies array. What might be confusing to you is, not only does member_array() tell us if the value is an element in the array, but it in fact tells us which element of the array the value is.

How does it tell you which element? It is easier to understand arrays if you think of the array variable as holding a number. In the value above, for the sake of argument, we will say that hard_currencies holds the value 179000. This value tells the driver where to look for the array hard_currencies represents. Thus, hard_currencies points to a place where the array values may be found. When someone is talking about the first element of the array, they want the element located at 179000. When the object needs the value of the second element of the array, it looks at 179000 + one value, then 179000 plus two values for the third, and so on. We can therefore access individual elements of an array by their index, which is the number of values beyond the starting point of the array we need to look to find the value. For the array hard_currencies array:

The efun member_array() thus returns the index of the element being tested if it is in the array, or -1 if it is not in the array. In order to reference an individual element in an array, you use its index number in the following manner:
    array_name[index_no]
Example:
    hard_currencies[3]
where hard_currencies[3] would refer to "silver".

So, you now should now several ways in which arrays appear either as a whole or as individual elements. As a whole, you refer to an array variable by its name and an array constant by enclosing the array in ({ }) and separating elements by ,. Individually, you refer to array variables by the array name followed by the element's index number enclosed in [], and to array constants in the same way you would refer to simple data types of the same type as the constant. Examples:

You can use these means of reference to do all the things you are used to doing with other data types. You can assign values, use the values in operations, pass the values as parameters to functions, and use the values as return types. It is important to remember that when you are treating an element alone as an individual, the individual element is not itself an array (unless you are dealing with an array of arrays). In the example above, the individual elements are strings. So that:
    str = arr[3] + " and " + arr[1];
will create str to equal "silver and gold". Although this seems simple enough, many people new to arrays start to run into trouble when trying to add elements to an array. When you are treating an array as a whole and you wish to add a new element to it, you must do it by adding another array.

Note the following example:

    string str1, str2;
    string *arr;

    str1 = "hi";
    str2 = "bye";
    /* str1 + str2 equals "hibye" */
    arr = ({ str1 }) + ({ str2 });
    /* arr is equal to ({ str1, str2 }) */
Before going any further, I have to note that this example gives an extremely horrible way of building an array. You should set it: arr = ({ str1, str2 }). The point of the example, however, is that you must add like types together. If you try adding an element to an array as the data type it is, you will get an error. Instead you have to treat it as an array of a single element.

3.5 Mappings

One of the major advances made in LPMuds since they were created is the mapping data type. People alternately refer to them as associative arrays. Practically speaking, a mapping allows you freedom from the association of a numerical index to a value which arrays require. Instead, mappings allow you to associate values with indices which actually have meaning to you, much like a relational database.

In an array of 5 elements, you access those values solely by their integer indices which cover the range 0 to 4. Imagine going back to the example of money again. Players have money of different amounts and different types. In the player object, you need a way to store the types of money that exist as well as relate them to the amount of that currency type the player has. The best way to do this with arrays would have been to store an array of strings representing money types and an array of integers representing values in the player object. This would result in CPU-eating ugly code like this:

    int query_money(string type) {
        int i;

        i = member_array(type, currencies);
        if(i>-1 && i < sizeof(amounts))
                /* sizeof efun returns # of elements */
            return amounts[i];
        else return 0;
    }
And that is a simple query function. Look at an add function:
    void add_money(string type, int amt) {
        string *tmp1;
        int * tmp2;
        int i, x, j, maxj;

        i = member_array(type, currencies);
        if(i >= sizeof(amounts))
                /*  corrupt data, we are in a bad way */
            return;
        else if(i== -1) {
            currencies += ({ type });
            amounts += ({ amt });
            return;
        }
        else {
            amounts[i] += amt;
            if(amounts[i] < 1) {
                tmp1 = allocate(sizeof(currencies)-1);
                tmp2 = allocate(sizeof(amounts)-1);
                for(j=0, x =0, maxj=sizeof(tmp1); j < maxj; j++) {
                    if(j==i) x = 1;
                    tmp1[j] = currencies[j+x];
                    tmp2[j] = amounts[j+x];
                }
                currencies = tmp1;
                amounts = tmp2;
            }
        }
    }
That is really some nasty code to perform the rather simple concept of adding some money. First, we figure out if the player has any of that kind of money, and if so, which element of the currencies array it is. After that, we have to check to see that the integrity of the currency data has been maintained. If the index of the type in the currencies array is greater than the highest index of the amounts array, then we have a problem since the indices are our only way of relating the two arrays. Once we know our data is in tact, if the currency type is not currently held by the player, we simply tack on the type as a new element to the currencies array and the amount as a new element to the amounts array. Finally, if it is a currency the player currently has, we just add the amount to the corresponding index in the amounts array. If the money gets below 1, meaning having no money of that type, we want to clear the currency out of memory.

Subtracting an element from an array is no simple matter. Take, for example, the result of the following:

    string *arr;

    arr = ({ "a", "b", "a" });
    arr -= ({ arr[2] });
What do you think the final value of arr is? Well, it is:
    ({ "b", "a" })
Subtracting arr[2] from the original array does not remove the third element from the array. Instead, it subtracts the value of the third element of the array from the array. And array subtraction removes the first instance of the value from the array. Since we do not want to be forced on counting on the elements of the array as being unique, we are forced to go through some somersaults to remove the correct element from both arrays in order to maintain the correspondence of the indices in the two arrays.

Mappings provide a better way. They allow you to directly associate the money type with its value. Some people think of mappings as arrays where you are not restricted to integers as indices. Truth is, mappings are an entirely different concept in storing aggregate information. Arrays force you to choose an index which is meaningful to the machine for locating the appropriate data. The indices tell the machine how many elements beyond the first value the value you desire can be found. With mappings, you choose indices which are meaningful to you without worrying about how that machine locates and stores it.

You may recognize mappings in the following forms:

So now my monetary functions would look like:
    int query_money(string type) { return money[type]; }

    void add_money(string type, int amt) {
        if(!money[type]) money[type] = amt;
        else money[type] += amt;
        if(money[type] < 1)
          map_delete(money, type);          /* this is for MudOS */
                ...OR... 
          money = m_delete(money, type)  /* for some LPMud 3.* varieties */
                ... OR... 
          m_delete(money, type);    /* for other LPMud 3.* varieties */
    }
Please notice first that the efuns for clearing a mapping element from the mapping vary from driver to driver. Check with your driver's documentation for the exact name an syntax of the relevant efun.

As you can see immediately, you do not need to check the integrity of your data since the values which interest you are inextricably bound to one another in the mapping. Secondly, getting rid of useless values is a simple efun call rather than a tricky, CPU-eating loop. Finally, the query function is made up solely of a return instruction.

You must declare and initialize any mapping before using it. Declarations look like:

    mapping map;
Whereas common initializations look like:
    map = ([]);
    map = allocate_mapping(10)   ...OR...   map = m_allocate(10);
    map = ([ "gold": 20, "silver": 15 ]);
As with other data types, there are rules defining how they work in common operations like addition and subtraction:
    ([ "gold":20, "silver":30 ]) + ([ "electrum":5 ]) 
gives:
    (["gold":20, "silver":30, "electrum":5])   
Although my demonstration shows a continuity of order, there is in fact no guarantee of the order in which elements of mappings will stored. Equivalence tests among mappings are therefore not a good thing.

3.6 Summary

Mappings and arrays can be built as complex as you need them to be. You can have an array of mappings of arrays. Such a thing would be declared like this:
    mapping *map_of_arrs;
which might look like:
    ({ ([ ind1: ({ valA1, valA2}), ind2: ({valB1, valB2}) ]), ([ indX:
    ({valX1,valX2}) ]) })
Mappings may use any data type as an index, including objects. Mapping indices are often referred to as keys as well, a term from databases. Always keep in mind that with any non-integer data type, you must first initialize a variable before making use of it in common operations such as addition and subtraction. In spite of the ease and dynamics added to LPC coding by mappings and arrays, errors caused by failing to initialize their values can be the most maddening experience for people new to these data types. I would venture that a very high percentage of all errors people experimenting with mappings and arrays for the first time encounter are one of three error messages:
  1. Indexing on illegal type.
  2. Illegal index.
  3. Bad argument 1 to (+ += - -=) /* insert your favourite operator */
Error messages 1 and 3 are darn near almost always caused by a failure to initialize the array or mapping in question. Error message 2 is caused generally when you are trying to use an index in an initialized array which does not exist. Also, for arrays, often people new to arrays will get error message 3 because they try to add a single element to an array by adding the initial array to the single element value instead of adding an array of the single element to the initial array. Remember, add only arrays to arrays.

At this point, you should feel comfortable enough with mappings and arrays to play with them. Expect to encounter the above error messages a lot when first playing with these. The key to success with mappings is in debugging all of these errors and seeing exactly what causes wholes in your programming which allow you to try to work with uninitialized mappings and arrays. Finally, go back through the basic room code and look at things like the set_exits() (or the equivalent on your mudlib) function. Chances are it makes use of mappings. In some instances, it will use arrays as well for compatibility with mudlib.n.

Chapter 4: The LPC Pre-Compiler


Copyright (c) George Reese 1993