Here is a GetFileExtension procedure. It reads the extension of a filename including the dot separator. The only condition to use the procedure is to avoid NULL pointers.
.386
.model flat,stdcall
option casemap:none
.data
GFEtable db 1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
db 0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
db 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
db 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
db 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
db 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
db 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
db 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
.code
GetFileExtension PROC uses esi _file:DWORD
mov esi,_file
mov edx,OFFSET GFEtable
xor ecx,ecx
l0:
mov eax,esi ; The procedure returns the address of the filename
@@: ; if there is no extension
add esi,1
mov cl,BYTE PTR [esi]
cmp BYTE PTR [edx+ecx],0
je @b
cmp cl,'.'
je l0
ret
GetFileExtension ENDP
END
[attachment deleted by admin]
Vortex,
You must scan the path backward to find the extension.
Your function will fail on this path...
C:\TEST.TEST.TXT
And your function should be called GetExtension because the returned extension could be one of a folder.
:wink
Exactly, above code fails to work on the scenerio that jdoe.I had that problem once but managed to make it :bdg
regards
jdoe,
The extension of the file TEST.TEST.TXT located on the root of partition C
C:\TEST.TEST.TXT
is
.TXT
This is what the procedure returns : Extension = .TXT
Where is the problem? On my Win Xp Pro Sp2 machine, I can see that the extension of this file .txt
On my Windows 2000 SP4 system, for "C:\TEST.TEST.TXT" I get "Extension = .TXT". The logic looks correct to me and I get correct results for all of the filenames I tried.
Hi Michael,
Many thanks for testing the routine.
Sorry Vortex, it's my mistake. I read the code too fast.
:red
No problem Jdoe, everything is fine.
What about "Test..txt" ?
Quote from: Grincheux on July 21, 2007, 02:31:00 PM
What about "Test..txt" ?
The routine gives the correct result :
.txtThere is no problem with that file name as the algo scans for the last dot in the string.
I have no time to make test, but it seems ther is a problem with a file named "..txt" because the second table byte is 0.
I made a little thinking, and something came out :lol
I cannot say that this routine is faster (MichaelW's timer routines show - no, Agner's - yes),
at least it's creative one...
.data
GFEtable2 db 22,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
db 8,8,8,8,8,8,8,8,8,8,8,8,8,8,6,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
db 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
db 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
db 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
db 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
db 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
db 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8
.code
GetFileExtensionR PROC uses esi edi _file:DWORD
mov edx, OFFSET GFEtable2
mov esi, _file
mov edi, $
nop
; :
mov eax, esi
; :
add esi, 1
xor ecx, ecx
mov cl, BYTE PTR [esi]
mov cl, BYTE PTR [edx+ecx]
add ecx, edi
jmp ecx
ret
GetFileExtensionR ENDP
Is there an infinite loop ?
Quote from: Grincheux on July 21, 2007, 04:16:35 PM
I have no time to make test, but it seems ther is a problem with a file named "..txt" because the second table byte is 0.
Grincheux,
Please test the routine before making comments about the functionnality of the algo, OK?
There is no problem with the file name
..txtThe correct result by executing the procedure :
Extension = .txt
Quote from: Grincheux on July 21, 2007, 06:12:07 PM
Is there an infinite loop ?
No. There is no any infinite loop. Depending on the array member's value, ecx is adjusted to jump to
mov eax, esi ,
add esi, 1 or
pop edi You can verify this by examining the procedure with Ollydbg. Ramguru, thanks for your contribution. Nice desing.
Using Agner Fog's strlen routine, I modified the design of my procedure :
.386
.model flat,stdcall
option casemap:none
StrLen PROTO :DWORD
.code
GetFileExtension PROC _file:DWORD
invoke StrLen,_file
@@:
mov edx,_file
mov ecx,eax
xor eax,eax
sub ecx,1
@@:
cmp BYTE PTR [edx+ecx],'.'
je @f
sub ecx,1
jnz @b
ret
@@:
lea eax,[edx+ecx]
ret
GetFileExtension ENDP
END
[attachment deleted by admin]
Sorry Vortex, I don't want to bother you but what about a path like this...
C:\TEST.1\TEST
::)
I guess jdoe found the tricky path and extension (no extension) that will fail...but it's always hard to foresee every possible case
congrats Vortex...your last routine rocks the world: 13584 (your last) < 24732 (mine) < 32772 (your first) cycles | according to Agners tests
edit > I think in documentation there should be a note: "if file has no extension append dot (".") at the end anyway" that could solve the mentioned problem.
Quote from: ramguru on July 24, 2007, 07:16:59 PM
...but it's always hard to foresee every possible case
In case of paths, there is no infinite possibilities. An extra compare for a backslash and the problem is solved.
:wink
Jdoe,
There are also other possibilities :
C:\test1...test2..test3\test.4\test
I guess checking for the symbol \ can defeat those variations.
The simpliest way was to call InString :bdg :bdg