In this file organization, the records of the file are stored one after another in the order they are added to the file.
In contrast to RELATIVE files, records of a INDEXED
SEQUENTIAL file can be accessed by specifying an ALPHANUMERIC
key in the READ statement (the KEY) .
It is the programmer's responsibility to take care
of the record sizes in files. You must be careful when declaring
record structures for files. Any mistake you make in record sizes
will cause an error and abnormal termination.
Please note that you must provide the record structure
to have a special field allocated to contain the KEY.
INDEXED files can have RANDOM or SEQUENTIAL ACCESS
MODEs. If the ACCESS MODE is declared to be RANDOM in
the corresponding SELECT statement, the program can read the record
with key value "ABC" and then the record with key="A12",
then record with key "D2X" etc etc. In short, one can
access any record of an indexed file, in any order, provided that
the KEY value is specified before the READ statement is executed.
If the ACCESS MODE is SEQUENTIAL, that means the
records of the file will be accesses in their physical sequential
order (just like SEQUENTIAL and LINE SEQUENTIAL files) and no
specific KEY value be given for the READ statements; but instead,
the NEXT clause will appear in READ statements meaning "Go
get the record with the next consecutive key value.
SELECT MYFILE ASSIGN TO DISK "C:\DATADIR\MYFILE.DAT"
ORGANIZATION IS INDEXED
ACCESS MODE IS RANDOM
RECORD KEY IS M-IDNO.
SELECT MYFILE-2 ASSIGN TO DISK "C:\DATADIR\MYFILE2.TXT"
ORGANIZATION IS INDEXED
ACCESS MODE IS SEQUENTIAL
RECORD KEY IS M-IDNO.
In the FILE-SECTION, you must provide fields for
the KEY, the M-IDNO in this example, in the FD block of the INDEXED
file;
FD MYFILE.
01 MYFILE-REC.
02 M-IDNO PIC XXXX.
02 M-NAME PIC X(16).
02 M-SURNAME PIC X(16).
02 M-BIRTHDATE.
03 M-BD-YEAR PIC 9999.
03 M-BD-MONTH PIC 99.
03 M-BD-DAY PIC 99.
Note : Since the
keys are not supposed to be numerical, you can use the X picture
for key fields.
Like all files, Indexed files must be OPENed
before they can be processes and CLOSEd when they are not
needed anymore.
Indexed files can be opened for :
Opening an indexed file for OUTPUT means that
the program will only issue WRITE statements to a NON-EXISTING
and JUST CREATED file. Therefore, when you open a file for OUTPUT,
COBOL assumes that the file does not exist and try to
create a new one. IF THE FILE EXISTS, ITS CONTENTS WILL BE CLEARED
WHEN OPENED AND ASSUMED TO BE A BRAND NEW FILE INTO WHICH
RECORDS WILL BE ADDED. You should be very careful when opening
a file for OUTPUT. One small mistake and all your valuable
records are lost forever.
Once you open an indexed file in Output mode, you
are expected to write records in the INCREASING order of keys.
That is, before writing the record with key A98 you should first
write all the records with key values alphanumerically less than
"A98". Please note that indexed fikle keys are case
sensitive; that is the key value "a89" is NOT EQUAL
TO "A89"; in fact it is alphanumerically greater than
"A89" because the ASCII value of "a" (97)
is greater than the ASCII value of "A" (65).
Opening an indexed file for INPUT means that
the program will only issue READ statements to an EXISTING file.
Therefore, when you open a file for INPUT, COBOL assumes that
the file exists and try to access it. IF THE FILE DOES
NOT EXIST, AN ERROR MESSAGE WILL BE ISSUED INDICATING THAT THE
MENTIONED FILE COULD NOT BE FOUND.
If you have declared the ACCESS MODE to be RANDOM,
before each READ statement, you are supposed to move a valid KEY
value to the record field variable that is declared as the RECORD
KEY.
For instance;
OPEN INPUT MYFILE.
....
MOVE "X23" TO M-IDNO.
READ MYFILE.
Good programmers should take precautions in their
program to avoid error messages and subsequent abnormal terminations
is an INVALID value for the record key specified before the READ
statement.
The INVALID KEY clause handles this in COBOL.
MOVE "2300" TO M-IDNO.
READ MYFILE INVALID KEY PERFORM OUT-OF-RANGE.
The INVALID KEY condition raises if there is no record
in the file with key value equal to the specified key value.
If you have declared the ACCESS MODE as SEQUENTIAL,
you should use the NEXT clause in the READ statement. Like
READ MYFILE NEXT AT END PEFORM NO-MORE-RECORDS.
When you are finished with a file you must CLOSE
it.
CLOSE NEW-FILE.
CLOSE OLD-FILE NEW-FILE.
You can close more than one files with a single
CLOSE statement. When a COBOL program terminates, all files mentioned
in the program are automatically closed; therefore you will not
get an error message for those files you forget to close or do
not close at all. Please note that relying on COBOL to close your
files is not a proper programming style. Although not absolutely
necessary, you should close your files when you are done with
them.
REWINDING an indexed file
with ACCESS MODE RANDOM is not meaninful. If the ACCESS MODE is
SEQUENTIAL, you can CLOSE and than OPEN INPUT again to rewind
a sequentialy accessed indexed file.
Once you open an indexed file, you can use READ or
WRITE statements to read or add records to this file. You cannot
use a READ statement for a file opened for OUTPUT and similarly
you cannot use a WRITE statement for a file you have opened for
INPUT.
A typical READ statement for RANDOM access mode looks like :
MOVE "1X3" TO M-IDNO.
READ MY-FILE.
A good programmer must check whether the READ statement
could find a record to read successfully. That means; the programmer
must check whether there were any records with the indicated key
value when the statement was executed. The situation can be checked
with the following construct :
READ MY-FILE INVALID KEY PERFORM NO-RECORDS-FOUND.
The construct tells the COBOL compiler to execute
the paragraph labeled NO-RECORDS-FOUND when the READ statement
cannot find any record with the key value stored in the key field
variable.
Once your program EXECUTES a successful READ statement,
the information in the data record that was just brought into
memory will be available in the corresponding variables mentioned
in your record description declaration (the FD block).
When you want to create an indexed file, you have
to open it for OUTPUT. The COBOL statement that is used to put
records into a new file is the WRITE statement. The important
pointa that you should be careful with are that,
Sometimes you need to make some small changes in
the contents of a file. Of course it possible to create a brand
new file with the new, modified contents but this is not practical.
COBOL provides you with an OPEN I-O mode with which you can modify
only the required record in a file.
In other words, there is another file opening mode;
the I-O mode; and another special REWRITE statement. Suppose
that you want to change the surname field of a record in an indexed
file (originally "AYFER") into "AYFEROGLU".
The program that you can write could read like
SELECT MYFILE ASSIGN TO DISK "MYFILE.DAT"
ORGANIZATION IS INDEXED
ACCESS MODE IS RANDOM
RECORD KEY IS M-IDNO.
....
FD MYFILE.
01 MYFILE-REC.
02 M-IDNO PIC XXXX.
02 M-NAME PIC X(16).
02 M-SURNAME PIC X(16).
02 M-BIRTHDATE.
03 M-BD-YEAR PIC 9999.
03 M-BD-MONTH PIC 99.
03 M-BD-DAY PIC 99.
...
PROCEDURE DIVISION.
....
OPEN I-O MYFILE.
....
MOVE "X20" TO M-IDNO.
READ MYFILE INVALID KEY PERFORM KEY-ERROR.
*
* TO MAKE SURE THAT WE ARE ON THE CORRECT RECORD
*
IF M-SURNAME NOT = "AYFER" DISPLAY "WRONG RECORD"
DISPLAY"RECORD NOT UPDATED"
CLOSE MYFILE
STOP RUN.
MOVE "AYFEROGLU" TO M-SURNAME.
REWRITE MYFILE-REC.
....
CLOSE MYFILE.
.....
Important : The key field of a record in an indexed file CANNOT be updated. If the programmer has to update the key field, the only solution is to delete the record with the old key value and a add the same record to the file; this time with the new key value. |
When you need to add records to an indexed file, you should open
the file in I-O mode and just write new records.
Note : Keys in an indexed file must be unique. That is
to say, no two records can have tha same key. There is,
however, a technique to have duplicate keys in an indexed file,
but this will be covered later.
If you do not want a specific record to be kept in an indexed
file any more, you can use the DELETE statement to make the record
inaccessible.
The DELETE statement is used more or less like the WRITE statement.
You must move a coorect key value to the record key field and
iisue the DELETE statement.
MOVE "X12" TO M-IDNO.
DELETE MYFILE-REC.
Please note that, when you delete a record, the physical
record is NOT deleted. Only the record is rendered inaccessible.
Suppose, for instance that an indexed file is 20 Mbytes in total
length and you delete half of the records in this file. After
all the deletions are completed, you shall see that the file is
still 20 Mbytes in length. The only way you can recover the disk
space used by deleted records is REORGANIZING the indexed file.
(See section "REORGANIZING INDEXED FILES").
Reorganizing Indexed Files
: This operation is usually done by a utiliy program suppled by
the maunfacturer of the COBOL compiler that you use. Refer to
the manuals of your compiler for details and instructions.
Sometimes the programmers need to assign more than
one key fields in their records. An example could be a database
where one wants to keep records for vehicles registered to a certain
country. The Turkish vehicle plate numbering system is NOT suitable
for using relative files because it contains alphabetic parts
( eg. 06 AHT 08) . As you might guess, this type of key
values is very very suitable to be organized as an INDEXED file.
Having declared the "PLATE" field as the
key field, now, given a plate number, a program which has declared
an indexed file with RANDOM access mode, can find the corresponding
record very quickly (if the does exist, this condition will also
be detected very quickly).
Knowing that no two vehicles have the same plate
number (unique), the plate number is certainly a very good choice
for the key.
However, suppose that the program you have to write
is also expected to find records for vehicles of a certain brand
very quickly. Since the key is the plate number, the only way
you can find "OPEL" brand vehicles by going through
all the records (declare the access mode to be sequential and
READ NEXT all records till the AT END condition raises) testing
the BRAND field against the value given by the ýser ("OPEL"
for instance).
With indexed files, there is a better solution :
You can declare the BRAND field also to be a key; but AN ALTERNATE
key!. The PRIMARY key will still be the plate numbers. Since there
are many vehicles manufactured by OPEL in your database, you know
that the ALTERNATE KEY values will not be uniqe. THIS IS ALLOWED
IN COBOL. ONLY THE PRIMARY KEY HAS TO BE UNIQE. ALTERNATE KEYS
CAN HAVE DUPLICATE VALUES.
You can declare as many alternate keys your program
requires. More alternate keys you have, more extra data structures
will be required and slower be your program. Therefore certain
attention should be paid before assigning alternate keys. You
should avoid unnecessary alternate keys.
When you have alternate keys, you declare them in
the SELECT statement together with the PRIMARY key.
SELECT VEHICLES ASSIGN TO DISK "VEHICLES.DAT"
ORGANIZATION IS INDEXED
ACCESS MODE IS RANDOM
RECORD KEY IS PLATE
ALTERNATE KEY IS BRAND.
....
FD VEHICLES.
01 VEHICLE-REC.
02 PLATE PIC XXXXXXXXXX.
02 BRAND PIC X(20).
02 OWNER PIC X(32).
02 DATE-REGISTERED PIC 99999999.
02 ENGINE-NO PIC X(20).
02 CHASSIS-NO PIC X(20).
...
When you need to access the file with the primary
key, you just move the key value to the primary key field of the
record and issue a READ statement.
MOVE "06 AHT 08" TO PLATE.
READ VEHICLES INVALID KEY PERFORM NOT-FOUND.
When you need to access the file using an alternate
key, you move an appropriate value to the alternate key you want
to use and issue a READ statement in which you specify which key
you want to use
MOVE "CHEVROLET" TO BRAND.
READ VEHICLES KEY BRAND INVALID KEY PERFORM NOT-FOUND.
The START Statement
Sometimes, it is necessary to find a set of records
for which you want to specify a criterion. For example, in our
"vehicles" example, you might want to find the records
of "TOYOTA" brand vehicles. If the criterion is related
to one of the keys, you will not have to go through the whole
file, testing the fields against a certan value. Instead, you
can use the START statement to find the first vehicle with "TOYOTA"
recorded as the brand.
An example is
IMPORTANT : Please note that when you use the START statement, the file is accessed as a RANDOM file; whereas the subsequent READ NEXT statements acces the file in SEQUENTIAL mode. This dilemna is solved by the ACCESS MODE IS DYNAMIC phrase in the SELECT statement. When you declare the access mode to be DYNAMIC, you can access the file both sequentially ( the NEXT option of the READ statement) and randomly by specifying a certain key value. When you want to use the START statement you must open an index file either as INPUT or I-O. The START statement is not allowed and in fact not meaningful when the file is opened for OUTPUT.
START VEHICLES KEY BRAND IS EQUAL TO "TOYOTA"
INVALID KEY NO-SUCH-BRAND.
LOOP1.
READ VEHICLES NEXT AT END PERFORM END-OF-BRAND.
...
GO LOOP1.
...