I got stuck on a task.
I have to write a program, which reads strings from file, sorts the last 10 strings by sixth column and then outputs sorted strings in a new file.
The first two columns contain string with [1; 20] characters, then 3-5 columns contain integers in the range [-100; 100] and 6th column contain real numbers in range [-9.99; 9.99]. Each section is separated with ';' symbol.
Input file:
48CBC0h ucbe5u F bc6; 6 laY xdU;-62;-29;11;-6.34
AJKCvwHUvmL CjWRl;WQc1 E2wzyTU;3;-20;24;8.26
9jicRCI8S b ; p9m R7iHqOj 9ig h;-93;19;-92;6.18
Xv ;IufLkIUp;-23;-94;76;2.63
A o8P8T 26;Zy J IVg6;-80;-58;-42;-5.96
hlwCw Z8ChU;KX0 w C3 N60 KZV;-94;61;-71;-3.31
Rtqn58 2 l BTWdVgl H;rd U ;9;70;10;8.66
M x91pVZ6UQ Nb;p;-5;-27;74;4.04
rq8s3Gc Bj x 2XG ;8E cTH a ZF VLp2E;-4;21;-89;7.9
ca yfUpVXuC7 ;sXvttLZs4 nqcv 5fTg;-34;-40;14;-5.19
j;yL7 G dG C vR B;70;-89;-87;-9.52
2n 5 O F MMc16; Awcsl2sI;-97;-82;34;1.01
EeLVLB;qR q4i D5q ;70;49;-11;-5.43
nsWW j9;AGBGVXO N;59;97;74;6.22
ou 7vCvBUc;yVW30Jwcv Qtj;18;-10;77;9.38
I tried to save the pointers of these lines, as well as the sixth column, and I didn't started to think of a sorting function.
Here is what I wrote so far:
%include 'yasmmac.inc'
org 100h
section .text
start:
mov si, 0x80
mov di, readFile
dec di
mov cl, byte [si]
cmp cl, 01
jg .name
.name:
inc si
mov al, byte [si]
mov byte [di], al
inc di
loop .name
macPutString 'Input name for writing file', crlf, '$'
mov al, 128
mov dx, writeFile
call procGetStr
macNewLine
mov dx, readFile
call procFOpenForReading
jnc .reading
macPutString 'Error occured', crlf, '$'
jmp .end
.reading:
mov [readingD], bx
mov di, pointers
xor dx, dx
dec dx
.whileNotEndOfFile:
push bx
mov bx, numberOfLines
inc word [bx]
pop bx
mov [di], dx
inc di
inc di
.tillLineEnds:
inc dx
call procFGetChar
cmp ax, 0x0
je .readingClose
cmp cl, 0x0a
jne .tillLineEnds
jmp .whileNotEndOfFile
.readingClose:
call procFClose
mov di, pointers
inc word[di]
mov dx, writeFile
call procFCreateOrTruncate
jnc .filter
macPutString 'Error occured', crlf, '$'
jmp .end
.filter:
mov [writingD], bx
push bx
mov bx, numberOfLines
dec word [bx]
cmp word [bx], 0x000a
jl .errorInfo
pop bx
jge .continue
.errorInfo:
pop bx
mov bx, [writingD]
mov cx, 0x0050
mov dx, error
call procFWrite
mov bx, [readingD]
call procFClose
mov bx, [writingD]
call procFClose
jmp .end
.continue:
mov dx, readingFile
call procFOpenForReading
jnc .secondReading
macPutString 'Error occured', crlf, '$'
jmp .pabaiga
.secondReading:
mov [readingD], bx
mov dx, numberOfLines
sub dx, 0x000a
xor cx, cx
mov cx, dx
xor ax, ax
cmp cx, 0x0
jg .skipLines
mov di, sixth
jmp .sixthColumn
.skipLines:
push cx
.skip:
call procFGetChar
cmp cl, 0x0a
jne .skip
pop cx
loop .skipLines
.sixthColumn:
xor si, si
.skipSign:
call procFGetChar
cmp cl, ';'
jne .skipSign
inc si
cmp si, 0x0005
jl .skipSign
.save:
call procFGetChar
cmp ax, 0x0
je .ifNotEnd
cmp cl, 0x0a
je .ifNotEnd
cmp cl, '.'
je .save
mov [di], cl
inc di
jmp .save
.ifNotEnd:
cmp ax, 0x0
je .endReading
jmp .sixthColumn
.endReading:
call procFClose
.end
exit
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
%include 'yasmlib.asm'
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
section .data
readingFile:
times 255 db 00
writingFile:
times 255 db 00
readingD:
dw 0000
writingD:
dw 0000
error:
db 'Not enough data', 0x0d, 0x0a, '$'
numberOfLines:
times 128 db 00
pointers:
times 3000 db 00
sixth:
times 20000 db 00
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
section .bss
Thank you in advance.
CodePudding user response：
Review of your work (the most important part)
mov si, 0x80 mov di, readFile dec di mov cl, byte [si] cmp cl, 01 jg .name .name: inc si mov al, byte [si] mov byte [di], al inc di loop .name
I see 3 problems in this code that fetches a filespec from the commandline:
Because of the decrement in
mov di, readFiledec di, the loop that follows will be writing in a byte that lies outside of the readFile buffer!You always run the .name loop
jg .name .name:This construct is non-sense in assembly programming. If the condition is greater then you jump to .name, but if the condition is not greater then you fall-through in .name. No matter, you will always run the code at .name.
The
loop .nameinstruction depends on the whole CX register, yet you only initialized its lowest 8 bits (mov cl, byte [si]).
In .reading you have next code to increment the numberOfLines variable:
push bx mov bx, numberOfLines inc word [bx] pop bx
Firstly, you can do this in 1 instruction inc word [numberOfLines], and secondly, why do you increment the line count before even knowing that a following line will be found? Later in the program you decrement the count unconditionally. Something similar happens to the first pointer that you store: you begin with assigning -1 to it, and then at some later point you unconditionally set it to 0.
At .continue, you re-open the sourcefile after having closed it, but you are using a different filespec now!
.continue: mov dx, readingFile call procFOpenForReading
Decide where it is. At readFile or at readingFile.
The alternative solution that I propose below, will not require reading the source file twice.
At .secondReading, you forgot the square brackets for dereferencing.
mov dx, numberOfLines sub dx, 0x000a xor cx, cx mov cx, dx xor ax, ax cmp cx, 0x0 jg .skipLines mov di, sixth jmp .sixthCollumn
Because of this, .skipLines that follows will run for way to many iterations! And because of where you burried that mov di, sixth instruction, the code at .sixthCollumn will write to wherever the DI register happens to point at!
numberOfLines: times 128 db 00 pointers: times 3000 db 00 sixth: times 20000 db 00
Why do you reserve this enormous amount of bytes just for processing 10 lines? More is definitely not better!
numberOfLines: dw 0
pointers: times 10 dw 0
sixth: times 10 dw 0
My solution (important if you study it carefully)
Hint: 20 1 20 1 4 1 4 1 4 1 5 2 = 64 (longest line)
xor si, si ; NumberOfLines
mov bx, [ReadingD] ; Handle
.reuseBuffers:
mov bp, Buffers ; 10 slots of 64 bytes each
.nextLine:
mov di, bp ; Begin of current slot
lea bp, [di 64] ; End of current slot
.nextByte:
call procFGetChar ; -> AX CL CF
jc .err ; A DOS error occured
test ax, ax
jz .eof ; Reached the end of the file
cmp di, bp
jnb .err ; Line is longer than expected
mov [di], cl
inc di
cmp cl, 10 ; Check for end of the line
jne .nextByte ; No
inc si ; Yes, NumberOfLines
cmp bp, Buffers 640 ; Check for the end of the buffer chain
jb .nextLine ; No
jmp .reuseBuffers ; Yes, go recycle
.err:
...
.eof:
cmp si, 10 ; NumberOfLines
jb .err
Buffers: times 10*64 db 0
The last 10 lines are now held in the 10 64-byte buffers. The order is not important since the task asks you to sort them anyway (using the 6th column).
I'm looking forward to seeing how you will be sorting the records. Personally I would convert the fixed point (*) numbers from the 6th column into integers in the range [-999,999] (ignoring the decimal point) and store them in a word-sized array ready for sorting. Instead of a secondary array containing pointers, I would use an array with indexes from 0 to 9.
The conditional swaps in the first array must be matched by unconditional swaps in the second array. There's a nice bubble sort to be found in the answer at segmentation fault in x86 trying to do bubble sort, but feel free to use any sorting algorithm of your liking...
(*) Beware of the line that reads
rq8s3Gc Bj x 2XG ;8E cTH a ZF VLp2E;-4;21;-89;7.9
From previous similar Q/A's, I was under the impression that the numbers in the 6th column would all have a 2-digit fraction. This line does not use "7.90". Either it's a typo you made or you will have to take this into account in your program.