RTR logo

R. T. RUSSELL

BBC BASIC (86) Manual



Indexed Data Files

Deficiencies of Random Access Files

As you will see if you dump a random file, a lot of space is wasted. This is because all the records must be allocated the same amount of space, otherwise you could not calculate where the record started. For large data files, over 50% of the space can be wasted. Under these circumstances it is possible to save space by using two files, one as an index to the other. In order for this to work efficiently, you must have complete control over the file pointer. Not many version of BASIC allow this control, but it is quite simple with BBC BASIC.

The Address Book Program

The final program is an example of an indexed file. It is a computer implementation of the address book discussed way back at the beginning of these notes. Two files are used, one as an index to the other. Both are serial and no space is wasted between records.

File Organisation

The files are organised as shown below:

NAME.NDX (index file)
maxrec
5 bytes
length
5 bytes
index$(1)
1 to 31 bytes
index(1)
5 bytes
index$(2)
1 to 31 bytes
index(2)
5 bytes
etc.Right

Where index(n) points to a record in the data file as follows:

ADDRESS.DTA (data file)
Phone Num
1 to 31 bytes
Address 1
1 to 31 bytes
Address 2
1 to 31 bytes
Address 3
1 to 31 bytes
Address 4
1 to 31 bytes
Post Code
1 to 31 bytes

maxrec Is the maximum number of records permitted in this file. The practical limit is governed by the amount of memory available for the index arrays which are held in memory. If you want to write a disk access and sort program for the index - the best of luck. And please can I have a copy?
length Is the number of entries in the index.
index(n) Is the value of PTR#datanum just prior to the first byte of the data for this entry being written to it. In the Random File examples this value was calculated and it increased by a constant amount for every record.

Program Organisation

The example looks horribly long and complicated. However the actual file handling bits are quite simple. The rest is, as usual, required for tidy input and output of data. The meat of the program is in the procedures and functions for putting and deleting index entries and finding the right place in the index. The latter uses a routine called a 'binary chop' (you could get arrested for that). This looks simple, and it is - when it works. If you are interested there is a flow chart and a brief explanation of how it works at the end of these notes. For the faithful, just use it. It takes considerably less time than any other method to search an ordered list.

The Index

The index is read into memory at the start and written back at the end. In memory, it consists of two arrays called index$() and index(). Oh that we could have mixed type arrays!

Ex 12 (the LAST)

 10 REM F-INDEX
 20 REM EXAMPLE OF AN INDEXED FILE
 30 :
 40 REM Written by Doug Mounter - Feb 1982
 50 REM Modified for BBCBASIC(86) - Dec 1985
 60 :
 70 REM This is a simple address book filing
 80 REM system.  It will accept names, telephone
 90 REM numbers and addresses and store them in a
100 REM file called ADDRESS.DTA.  The index is in
110 REM name order and is kept in a file called
120 REM NAME.NDX.  All the fields are character
122 REM and the maximum length  of any field
124 REM is 30.
130 :
140 PROC_initialise
150 PROC_open_files
160 ON ERROR IF ERR<>17 PRINT:REPORT:PRINT" At line ";ERL:END
170 REPEAT
180   CLS
190   PRINT TAB(5,3);"If you want to:-" '
200   PRINT TAB(10);"End This Session";TAB(55);"Type 0"
210   PRINT TAB(10);"Enter Data";TAB(55);"Type 1"
220   PRINT TAB(10);"Search For/Delete an Entry";TAB(55);"Type 2"
230   PRINT TAB(10);"List in Alphabetical Order";TAB(55);"Type 3"
240   PRINT TAB(10);"Reorg data File and Index";TAB(55);"Type 4";
250   REPEAT
260     PRINT TAB(5,11);
270     PRINT "Please enter the appropriate number (0 to 4) ";
280     function$=GET$
290     PRINT return$;:PROC_cteol
300   UNTIL function$>"/" AND function$<"5"
310   function=VAL(function$)
320   PRINT TAB(54,function+5);"<====<<";
330   ON function PROC_enter,PROC_search,PROC_list,PROC_reorg,ELSE
340 UNTIL function=0
350 CLS
360 PROC_close_files
370 *ESC ON
380 PRINT "Address Book Files Closed"''
390 END
400 :
410 :
420 REM ENTER DATA
430 :
440 DEF PROC_enter
450 flag=TRUE
460 temp$=""
470 i=1
480 REPEAT
490   REPEAT
500     IF temp$="N" PROC_message("Data NOT Accepted")
510     PROC_get_data
520     IF length=maxrec OR data$(1)="" flag=FALSE:GOTO 590
530     IF data$(1)="+" OR data$(1)="-" PROC_message("Bad Data"):GOTO 590
540     i=FN_find_place(0,data$(1))
550     IF i>0 PROC_message("Duplicate Record")
560     PRINT '"Is this data correct ? ";
570     temp$=FN_yesno
580     :
590   UNTIL NOT flag OR temp$<>"N"
600   PROC_cteos
610   IF NOT flag THEN 670
620   PROC_put_index(i,data$(1),PTR#datanum)
630   FOR i=2 TO 7
640     PRINT#datanum,data$(i)
650   NEXT
660   :
670 UNTIL NOT flag
680 ENDPROC
690 :
700 :
710 REM SEARCH FOR AN ENTRY
720 :
730 DEF PROC_search
740 i=0
750 REPEAT
760   PRINT TAB(0,11);:PROC_cteol
770   INPUT "What name do you want to look for ",name$
780   IF name$="" THEN 800
790   IF name$<>""IF name$="DELETE" PROC_delete(i) ELSE i=FN_display(i,name$)
800 UNTIL name$=""
810 ENDPROC
820 :
830 :
840 REM LIST IN ALPHABETICAL ORDER
850 :
860 DEF PROC_list
870 entry=1
880 REPEAT
890   CLS
900   line_count=0
910   REPEAT
920     PRINT TAB(0,line_count);
930     PROC_read_data(entry)
940     PROC_print_data
950     entry=entry+1
960     line_count=line_count+8
970     temp$=INKEY$(0)
980   UNTIL entry>length OR line_count>16 OR temp$<>""
990   PROC_message("Push any key to continue or E to end ")
1000 UNTIL entry>length OR GET$="E"
1010 ENDPROC
1020 :
1030 :
1040 REM REORGANISE THE DATA FILE AND INDEX
1050 :
1060 DEF PROC_reorg
1070 entry=1
1080 PRINT TAB(0,13);"Reorganising the Data File" '
1090 newdata=OPENOUT"ADDRESS.BAK"
1100 REPEAT
1110   PROC_read_data(entry)
1120   index(entry)=PTR#newdata
1130   FOR i=2 TO 7
1140     PRINT#newdata,data$(i)
1150   NEXT
1160   entry=entry+1
1170 UNTIL entry>length
1180 CLOSE#newdata
The time taken to rename a file can be considerable.
1190 PRINT "Re-naming the Data File" '
1200 *REN ADDRESS.$$$=ADDRESS.BAK
1210 PRINT "*";
1220 *REN ADDRESS.BAK=ADDRESS.DTA
1230 PRINT "*";
1240 *REN ADDRESS.DTA=ADDRESS.$$$
1250 PRINT "*";
1260 datanum=OPENUP "ADDRESS.DTA"
1270 ENDPROC
1280 :
1290 :
1300 REM INITIALISE VARIABLES AND ARRAYS
1310 :
1320 DEF PROC_initialise
1330 MODE 7
1340 *ESC OFF
1350 esc$=CHR$(27)
1360 bell$=CHR$(7)
1370 return$=CHR$(13)
1380 maxrec=100
1390 :
1400 REM The maximum record number, maxrec, is
1402 REM read in
1410 REM PROC_read_index if the file already exists.
1420 :
1430 DIM message$(7)
1440 FOR i=1 TO 7
1450   READ message$(i)
1460 NEXT
1470 DATA Name,Phone Number,Address,-- " --,-- "--,-- " --,Post Code
1480 :
1490 DIM data$(7)
1500 FOR i=1 TO 7
1510   data$(i)=STRING$(30," ")
1520 NEXT
1530 temp$=STRING$(255," ")
1540 temp$=""
1550 :
1560 DIM code 3
1570 PROC_assemble
1580 ENDPROC
1590 :
1600 :
1610 REM OPEN THE FILES
1620 :
1630 DEF PROC_open_files
1640 indexnum=OPENUP"NAME.NDX"
1650 datanum=OPENUP"ADDRESS.DTA"
1660 IF indexnum=0 OR datanum=0 PROC_setup ELSE PROC_read_index
1670 PTR#datanum=EXT#datanum
1680 ENDPROC
1690 :
1700 :
1710 REM SET UP NEW INDEX AND DATA FILES
1720 :
1730 DEF PROC_setup
1740 CLS
1750 PRINT TAB(0,13);"Setting Up Address Book"
1760 indexnum=OPENOUT"NAME.NDX"
1770 datanum=OPENOUT"ADDRESS.DTA"
1780 length=0
1790 PRINT#indexnum,maxrec,length
1800 CLOSE#indexnum
1810 DIM index$(maxrec+1),index(maxrec+1)
1820 index$(0)=""
1830 index(0)=0
1840 index$(1)=CHR$(&FF)
1850 index(1)=0
1860 ENDPROC
1870 :
1880 :
1890 REM READ INDEX AND LENGTH OF DATA FILE
1900 :
1910 DEF PROC_read_index
1920 CLS
1930 INPUT#indexnum,maxrec,length
1940 DIM index$(maxrec+1), index(maxrec+1)
1950 index$(0)=""
1960 index(0)=0
1970 FOR i=1 TO length
1980   INPUT#indexnum,index$(i),index(i)
1990 NEXT
2000 CLOSE#indexnum
2010 index$(length+1)=CHR$(&FF)
2020 index(length+1)=0
2030 ENDPROC
2040 :
2050 :
2060 REM WRITE INDEX AND CLOSE FILES
2070 :
2080 DEF PROC_close_files
2090 indexnum=OPENOUT"NAME.NDX"
2100 PRINT#indexnum,maxrec,length
2110 FOR i=1 TO length
2120   PRINT#indexnum,index$(i),index(i)
2130 NEXT
2140 CLOSE#0
2150 ENDPROC
2160 :
2170 :
2180 REM WRITE A MESSAGE AT LINE 23
2190 :
2200 DEF PROC_message(line$)
2210 LOCAL x,y
2220 x=POS
2230 y=VPOS
2240 PRINT TAB(0,23);:PROC_cteol:PRINT bell$;line$;
2250 PRINT TAB(x,y);
2260 ENDPROC
2270 :
2280 :
2290 REM GET A Y/N ANSWER
2300 :
2310 DEF FN_yesno
2320 LOCAL temp$
2330 temp$=GET$
2340 IF temp$="y" OR temp$="Y" ="Y"
2350 IF temp$="n" OR temp$="N" ="N"
2360 =""
2370 :
2380 :
2390 REM CLEAR 9 LINES FROM PRESENT POSITION
This procedure makes use of the machine code routine at the end of the program. It works in a similar fashion to the clear-to-end-of-line and clear-to-end-of-screen procedures defined towards the end of the program.
2400 :
2410 DEF PROC_clear9
2420 LOCAL x,y,i
2430 PRINT return$;
2440 A%=&A20:B%=0:C%=720:D%=0
2450 CALL int10
2460 ENDPROC
2470 :
2480 :
2490 REM GET INPUT DATA - LIMIT TO 30 CHAR
2500 :
2510 DEF PROC_get_data
2520 LOCAL i
2530 PRINT TAB(0,13);
2540 PROC_clear9
2550 IF length=maxrec PROC_message("Add Book Full")
2560 FOR i=1 TO 7
2570   PRINT TAB(10);message$(i);TAB(25);
2580   INPUT temp$
2590   data$(i)=LEFT$(temp$,30)
2600   IF data$(1)="" i=7
2610 NEXT
2620 ENDPROC
2630 :
2640 :
2650 REM FIND AND DISPLAY THE REQUESTED DATA
2660 :
2670 DEF FN_display(i,name$)
2680 PRINT TAB(0,12);:PROC_cteos
2690 i=FN_find_place(i,name$)
2700 IF i<0 PROC_message("Name Not Known - Next Highest Given")
2710 PROC_read_data(i)
2720 PRINT
2730 PROC_print_data
2740 =i
2750 :
2760 :
2770 REM DELETE THE ENTRY FROM THE INDEX
2780 :
Move everything below the entry you want deleted up one and subtract 1 from the length
2790 DEF PROC_delete(i)
2800 INPUT "Are you SURE ",temp$
2810 PRINT TAB(0,VPOS-1);:PROC_cteos
2820 IF temp$<>"YES" ENDPROC
2830 IF i<0 i=-i
2840 FOR i=i TO length
2850   index$(i)=index$(i+1)
2860   index(i)=index(i+1)
2870 NEXT
2880 length=length-1
2890 ENDPROC
2900 :
2910 :
2920 REM READ DATA FOR ENTRY i
Get the start of the position of the start of the data record for entry 'i' in the index and read it into the buffer array data$(). Save the current value of the data file pointer on entry and restore it before leaving.
2930 :
2940 DEF PROC_read_data(i)
2950 PTRdata=PTR#datanum
2960 IF i<0 i=-i
2970 PTR#datanum=index(i)
2980 data$(1)=index$(i)
2990 FOR i=2 TO 7
3000   INPUT#datanum,data$(i)
3010 NEXT
3020 PTR#datanum=PTRdata
3030 ENDPROC
3040 :
3050 :
3060 REM PRINT data$() ON VDU
3070 :
3080 DEF PROC_print_data
3090 LOCAL i
3100 FOR i=1 TO 7
3110   IF data$(i)<>"" PRINT TAB(10);message$(i);TAB(25);data$(i)
3120   IF data$(1)=CHR$(&FF) i=7
3130 NEXT
3140 ENDPROC
3150 :
3160 :
3170 REM PUT A NEW ENTRY IN INDEX AT POSITION i
Move all the directory entries from position i onwards down the index. (In fact you have to start at the end and work back.) Slot the new entry in in the gap made at position i and add 1 to the length.
3180 :
3190 DEF PROC_put_index(i,entry$,ptr)
3200 LOCAL j
3210 IF i<0 i=-i
3220 FOR j=length+1 TO i STEP -1
3230   index$(j+1)=index$(j)
3240   index(j+1)=index(j)
3250 NEXT
3260 index$(i)=entry$
3270 index(i)=ptr
3280 length=length+1
3290 ENDPROC
3300 :
3310 :
3320 REM FIND ENTRY IN INDEX OR PLACE TO PUT IT
This function looks in the index for the string entry$. If it finds it it returns with i set to its position in the index. If not, i is set to minus the position of the next highest string. (In other words, the position you wish to put the a new entry.) Thus if a part of the index looked like:
(34)   BERT
(35)FRED
(36)JOHN
and you entered with FRED, it would return 35. However if you entered with GEORGE, it would return -36.

The function consists of two parts. The first looks at the entry$ to see if it should just up or down the entry number by 1, taking account of wrap-around at the start and end of the index. The second part is the binary chop advertised with such telling wit in the introduction to indexed files. Since we enter this function with the entry pointer i set to its previous value, we must cater for a negative value.

3330 :
3340 DEF FN_find_place(i,entry$)
3350 LOCAL top,bottom
3360 IF i<0 i=-i
3370 IF entry$="+" AND i<length =i+1
3380 IF entry$="+" AND i=length =1
3390 IF entry$="-" AND i>1 =i-1 
3400 IF entry$="-" AND i<2 =length
Here, at last, T H E  B I N A R Y  C H O P
3410 top=length+1
3420 bottom=0
3430 i=(top+1) DIV 2
3440 IF entry$<>index$(i) i=FN_search(entry$)
3450 REPEAT
3460   IF entry$=index$(i-1) i=i-1
This bit moves the pointer up the index to the first of any duplicate entries.
3470 UNTIL entry$<>index$(i-1)
3480 IF entry$=index$(i) =i ELSE =-i
3490 :
3500 :
3510 REM DO THE SEARCHING FOR FN_find_place
3520 :
3530 DEF FN_search(entry$)
3540 REPEAT
3550   IF entry$>index$(i) bottom=i ELSE top=i
3560   i=(top+bottom+1) DIV 2: REM round
3570 UNTIL entry$=index$(i) OR top=bottom+1
3580 =i
3590 :
3600 :
The two following procedures rely on mode 3 or 7 being selected. They will not work properly if a graphics mode has been selected or if some characters on the screen have attributes set.
3610 REM There are no 'native' clear to end of
3620 REM line/screen vdu procedures.  The
3630 REM following two procedures clear to the end
3640 REM of the line/screen using interrupt 10.
3650 DEF PROC_cteol
3660 LOCAL x
3670 x=POS
3680 A%=&A20:B%=0:C%=80-x:D%=0
3690 CALL int10
3700 ENDPROC
3710 :
3720 :
3730 DEF PROC_cteos
3740 LOCAL I,x,y
3750 x=POS:y=VPOS
3760 A%=&A20:B%=0:C%=80-x+(24-y)*80:D%=0
3770 CALL int10
3780 ENDPROC
3790 :
3800 :
3810 DEF PROC_assemble
3820 P%=code
3830 [OPT 2
3832 ;no forward references - only 1 pass needed
3840 .int10 INT &10
3850 RETF:]
3860 ENDPROC
Well, that's it. Apart from the following notes on the binary chop you have read it all.


The Binary Chop

Explanation

The quickest way to find an entry in an ORDERED list is not to search through it from start to end, but to continue splitting the list in two until you reach the entry you are looking for. You begin by setting one pointer to the bottom of the list, another to the top, and a third to mid-way between bottom and top. Then you compare the entry pointed to by this third pointer with the number you are searching for. If your number is bigger you make the bottom equal the pointer, if not make the top equal to it. Then you repeat the process.

Let's try searching the list of numbers below for the number 14.

bottom>  (1)3    Set bottom to the lowest position in the list, and top to the highest. Set the pointer to (top+bottom)/2. Is that entry 14? No it's more, so set top to the current value of pointer and repeat the process.
  (2)6   
  (3)8   
  (4)14   
pointer> (5)19   
  (6)23   
  (7)34   
  (8)45   
top> (9) 61   
bottom>  (1)3    Set the pointer to (top+bottom)/2. Is that entry 14? No it's less, so set bottom to the current value of pointer and try again.
  (2)6   
pointer> (3)8   
  (4)14   
top> (5)19   
  (6)23   
  (7)34   
  (8)45   
  (9) 61   
  (1)3    Set the pointer to (top+bottom)/2. Is that entry 14? Yes, so exit with the pointer set to the position in the list of the number you are looking for.
  (2)6   
bottom>  (3)8   
pointer> (4)14   
top> (5)19   
  (6)23   
  (7)34   
  (8)45   
  (9) 61   
As you can imagine, things are not always as simple as this carefully chosen example. You have to cater for the number not being there, and for the list being empty. There are a number of ways of doing this, but the easiest is to add two numbers of your choice to the list. Make the first entry the most negative number the computer can hold, and the last entry the most positive. This will prevent you ever trying to search outside the list. Preventing a perpetual loop when the number you want is not in the list is quite simple, just exit when 'top' is equal to 'bottom'+1. If you have not found the number by then, it's not in the list.

You can use this routine to add numbers to the list in order. If you can't find the number, you exit with the position it should go in the list. Just move all the numbers under it down one slot and put the new number in. This works just as well when the list is empty except for your two 'end markers'.

Have a look at the flow chart below and work through a couple of dry runs with a short list of numbers. You may think that it's not worth doing it this way and that a 'linear search' would be as quick. Try it with a list of 100 numbers. It should take you no more than 7 goes to find the number. The AVERAGE number of comparisons required for a linear search would be 50.

Flow chart

Left CONTENTS

CONTINUE Right


Best viewed with Any Browser Valid HTML 3.2!
© Doug Mounter and Richard Russell 1999