Chapter 10 - Records
Prerequisites for this material
In order to do a profitable study of this material, you will need a good understanding of all of the material in Part I. The material concerning the scalar type from chapter 10 is also needed.
We come to the grandaddy of all data structures in Modula-2, the RECORD. A record is composed of a number of variables any of which can be of any predefined data type, including other records. Rather than spend time trying to define a record in detail, lets go right to the first example program, SMALLREC.MOD. This is a program using nonsense data that will illustrate the use of a record.
A very simple record
There is only one entry in the TYPE declaration part of the program, namely the record identified by "Description". The record is composed of three fields, the "Year", "Model", and "Engine" variables. Notice that the three fields are each of a different type, indicating that the record can be of mixed types. You have a complete example of the way a record is defined before you. It is composed of the identifier "Description", the reserved word RECORD, the list of elements, and followed by END;. Notice that this only defines a TYPE, it does not define any variables. That is done in the VAR declaration where the variable "Cars" is defined as an array of 10 elements or variables of type "Description". The variable "Cars" has three components, "Year", "Model", and "Engine", and any or all of these components can be used to store data pertaining to "Cars".
In order to assign values to the various fields, the variable name is followed by the sub-field with a separating period. Keep in mind that "Cars" is a complete record containing three variables, and to assign or use one of the variables, you must designate which sub-field you are interested in. See the program where the three fields are assigned meaningless data for illustration. The "Year" field is assigned an integer number varying with the subscript, all "Model" fields are assigned the name "Duesenburg", and all "Engine" variables are assigned the value "V8". In order to further illustrate that there are actually 30 variables in use here, a few are changed at random in the next few statements, being very careful to maintain the required types as defined in the TYPE declaration part of the program. Finally, all ten composite variables, consisting of 30 actual variables in a logical grouping are printed out using the same "var.subfield" notation described above.
If the preceding description of a record is not clear in your mind, review it very carefully. It's a very important concept in Modula-2, and you won't have a hope of a chance of understanding the next example until this one is clear.
A super record
Examine the example file BIGREC.MOD for a very interesting record. First we have a constant defined. Ignore it for the moment, we will come back to it later. Within the TYPE declaration we have three records defined, and upon close examination, you will notice that the first two records are included as part of the definition of the third record. The record identified as "Person", actually contains 8 variable definitions, three within the "FullName" record, two of its own, and three within the "Date" record. This is a TYPE declaration and does not actually define any variables, that is done in the VAR part of the program.
The VAR part of the program defines some variables beginning with the array of "Friend" containing 50 (because of the constant definition in the CONST part) records of "Person". Since "Person" defines 8 fields, we have now defined 8 times 50 = 400 separate and distinct variables. Each of the 400 separate variables has its own type associated with it, and the compiler will generate an error if you try to assign any of those variables the wrong type of data. Since "Person" is a TYPE definition, it can be used to define more than one variable, and in fact it is used again to define three more records, "Self", "Mother", and "Father". These three records are each composed of 8 variables, so we have 24 more variables which we can manipulate within the program. Finally we have the variable "Index" defined as a simple CARDINAL type variable. Notice that if we desired, we could also define a variable of type "FullName" composed of 3 simple variables.
How to manipulate all of that data
In the program we begin by assigning data to all of the fields of "Self". Examining the first three statements of the main program, we see the construction we learned in the last example program being used, namely the period between descriptor fields. The main record is named "Self", and we are interested in the first part of it namely the "Name" part of the person record. Since the "Name" part of the person record is itself composed of three parts, we must designate which part of it we are interested in. The complete description "Self.Name.FirstName" is the complete description of the first name of "Self" and is the first assignment statement which is assigned the name of "Charley". The next two fields are handled in the same way and are self explanatory.
What is the WITH statement?
Continuing on to the fourth field, the "City", there are only two levels required because "City" is not another record definition. The fourth field is therefore completely defined by "Self.City". Notice the "WITH Self DO" statement. This is a shorthand notation used with record definitions to simplify coding. From the BEGIN at the next statement to the matching END; about 10 statements later, any variables within the "Self" record are used as though they had a "Self." in front of them. It greatly simplifies coding to be able to omit the leading identifier within the WITH section of code. You will see that "City", and "State", are easily assigned values without further reference to the "Self" variable. When we get to the "Day" part of the birthday, we are back to three levels and the complete definition is "Self.Birthday.Day" but once again, the "Self." part is taken care of automatically because we are still within the "WITH Self DO" area.
To illustrate the WITH statement further, another is introduced, "WITH Birthday DO", and an area is defined by the BEGIN END pair. Within this area both leading identifiers are handled automatically to simplify coding, and "Month" is equivalent to writing "Self.Birthday.Month" if both WITH statements were removed. You may be wondering how many levels of nesting are allowed in record definitions. There doesn't appear to be a limit according to the Modula-2 definition, but we do get a hint at how far it is possible to go. In most implementations of Modula-2, you are allowed to have WITH statements nested to nine levels, and it would be worthless to nest WITH statements deeper than the level of records. Any program requiring more levels than nine is probably far beyond the scope of your programming ability, and mine, for a long time.
After assigning a value to the year, the entire record of "Self" is defined, all eight variables.
The next statement, "Mother := Self;" is very interesting. Since both of these are records, both are the same type of record, and both therefore contain 8 variables, Modula-2 is smart enough to recognize that, and assign all eight values contained in "Self" to the corresponding variables of "Mother". So after one statement, "Mother" is completely defined. The next statement assigns the same values to the eight respective fields of "Father", and the next two lines assign all 50 "Friend" variables the same data. We have therefore generated 400 + 24 = 424 separate pieces of data so far in this program. We could print it all out, but since it is nonsense data, it would only waste time and paper. The next three lines write out three sample pieces of the data for your inspection.
What good is all of this
It should be obvious to you that what this program does, even though the data is nonsense, appears to be the beginning of a database management program, which indeed it is. It is a crude beginning, and has a long way to go to be useful, but you should see a seed for a useful program.
Now to go back to the CONST as promised. The number of friends was defined as 50 and used for the size of the array and in the assignment loop near the end of the program. You can now edit this number and see how big this database can become on your computer. Your compiler should be capable of storing about 1000 records even within the smallest model available on any compiler. If your compiler uses a larger memory model, you will be able to store significantly more records. See how big you can make the number of friends before you get the memory overflow message. Keep the number in mind because when we get to the chapter on Pointers and Dynamic Allocation, you should see a marked increase in allowable size, especially if you have a large amount of RAM installed in your computer. If your compiler uses a large memory model, you won't see an increase in size but it will be an interesting exercise anyway.
A variant record
If any part of this chapter is still unclear, it would be good for you to go back and review it at this time. The next example will really tax your mind to completely understand it, especially if the prior material is not clear.
Examine the program VARREC.MOD for an example of a program with a variant record definition. In this example, we first define a scalar type, namely "KindOfVehicle" for use within the record. Then we have a record defining "Vehicle", intended to define several different types of vehicles, each with different kinds of data. It would be possible to define all variables for all types of vehicles, but it would be a waste of storage space to define the number of tires for a boat, or the number of propeller blades used on a car or truck. The variant record lets us define the data precisely for each vehicle without wasting data storage space.
What is a tag-field?
In the record definition we have the usual RECORD header followed by three variables defined in the same manner as the records in the last two example programs. Then we come to the CASE statement. Following this statement, the record is different for each of the four types defined in the associated scalar definition. The variable "WhatKind" is called the tag-field and must be defined as a scalar type prior to the record definition. The tag-field is what selects the variant used, when the program uses one of the variables with this record type. The tag-field is followed by a colon and its type definition, then the reserved word OF. A list of the variants is then given, with each of the variants having the variables for its particular case defined. The list of variables for one variant is called the field list.
A few rules are in order at this point. The variants do not have to have the same number of variables in each field list, and in fact, one or more of the variants may have no variables at all in its variant part. If a variant has no variables, it must still be defined with a blank followed by a semi-colon. All variables in the entire variant part must have unique names. The three variables, "Wheels", "Tires", and "Tyres", all mean the same thing to the user, but they must be different for the compiler. You may use the same identifiers again in other records and for simple variables anywhere else in the program. The Modula-2 compiler can tell which variable you mean by its context. Using the same variable name should be discouraged as bad programming practice because it may confuse you or another person trying to understand your program at a later date.
Using the variant record
We properly define four variables with the record type "Vehicle" and go on to examine the program itself.
We begin by defining one of our variables of type "Vehicle", namely the variable named "Ford". The seven lines assigning values to "Ford" are similar to the prior examples with the exception of the fourth line. In the fourth line the tag-field which selects the particular variant used is set equal to the value "Truck", which is a scalar definition, not a variable. This means that the variables named "Motor", "Tires", and "Payload" are available for use with the record "Ford", but the variables named "Wheels", "Engine", "Tyres", etc. are not available in the record named "Ford".
Next, lets define the record "Sunfish" as a boat, and define all of its variables. All of sunfish's variables are defined but in a rather random order to illustrate that they need not be defined in a particular order. Recall the use of WITH from the last example program.
To go even further in randomly assigning the variables to a record, we redefine "Ford" as having an "Engine" which it can only have if it is a car. This is one of the fine points of the record. If you assign any of the variant variables, the record is changed to that variant, but it is the programmers responsibility to assign the correct tag- field to the record, not the Modula-2 compiler's. Good programming practice would be to assign the tag-field before assigning any of the variant variables. The remainder of the "Ford" variables are assigned to complete that record, the non-variant part remaining from the last assignment.
The variable "Mac" is now set equal to the variable "Sunfish". All variables within the record are copied to "Mac" including the tag-field, making "Mac" a boat.
Now to see what we have in the records
We have assigned "Ford" to be a car, and two boats exist, namely "Sunfish" and "Mac". Since "Schwinn" was never defined, it has no data in it, and is at this point useless. The "Ford" tag-field has been defined as a car, so it should be true in the IF statement, and the first message should print. The "Sunfish" is not a bicycle, so it will not print. The "Mac" has been defined as a boat in the single assignment statement, so it will print a message with an indication that all of the data in the record was transferred to its variables.
Even though we can make assignment statements with records, they cannot be used in any mathematical operations such as addition, or multiplication. They are simply used for data storage. It is true however, that the individual elements in a record can be used in any mathematical statements legal for their respective types.
One other point should be mentioned. The tag-field can be completely eliminated resulting in a "free union" variant record. This is possible because Modula-2, as you may remember from above, will automatically assign the variant required when you assign data to one of the variables within a variant. This is the reason that all variables within any of the variants must have unique names. The free union record should be avoided in your early programming efforts because you cannot test a record to see what variant it has been assigned to.
A note to Pascal programmers
A record with a free union variant is commonly used in Pascal to do type transfers, but this should be discouraged in Modula-2 since it has a complete set of carefully defined type transfer functions for that purpose. In addition, the method of data storage is not specified as a part of the language and a free union would not operate the same way with different compilers if used for the purpose of type transfer.