Board index » delphi » StrToInt - Embedded Zeroes Issue

2006-10-26 11:33:56 PM
delphi268
Hi
I made a new poll because the other one was closed to early. I had by
mistake set a very early closing date.
Is it a bug that StrToInt does not handle embedded zeroes in the input
string?
tech.groups.yahoo.com/group/fastcodeproject/surveys
Please help me get many votes in.
Best regards
Dennis Kjaer Christensen
----------------------------------------------------------------------------
----
Jeg beskyttes af den gratis SPAMfighter til privatbrugere.
Den har indtil videre sparet mig for at f?3478 spam-mails
Betalende brugere får ikke denne besked i deres e-mails.
Hent en gratis SPAMfighter her.
 

"Dennis" <XXXX@XXXXX.COM>wrote
characters with an ord value of zero and sometimes
referred to as null characters.
"Does not handle" is rather vague.
Does StrToInt report an exception?
Does StrToInt assume the end of the string's data?
Does StrToInt ignore the null characters?
Does StrToInt give erroneous output when encountering
the null characters?
HTH to clarify things, JohnH
 

The RTL StrToInt function is coded as:-
function StrToInt(const S: string): Integer;
var
E: Integer;
begin
Val(S, Result, E);
if E <>0 then ConvertErrorFmt(@SInvalidInteger, [S]);
end;
The Val function called (_ValLong in system.pas) is designed to work with
both AnsiStrings and PChars, and assumes the first #0 character found is the
end of the string.
Given and input string of '0'#0, The call to Val from StrToInt returns a
Result of 0 and an error code (E) of 0, and no exception is raised within
IntToStr.
--
regards,
John
The Fastcode Project:
www.fastcodeproject.org/
"John Herbster" <herb-sci1_at_sbcglobal.net>writes

"John O'Harrow" <XXXX@XXXXX.COM>wrote
 

John Herbster writes:
123 - OK
00123 - not OK.
If this is wrong, please clarify the question and start the poll again.
David
 

Hi David
My wording in this thread was bad. I assumed that you all had read the
previous thread.
Look at the post from John in this thread and also read this thread"
Fastcode StrToInt32 B&V 0.8.0"
Leading zeroes are handled just fine by the RTL function.
Embedded Zeroes = '1'+#0+'2'
This string will be converted into integer=1 by the RTL function, because it
terminates on the #0= zero terminator in PChars.
I think that '1'+#0+'2' is a valid AnsiString that cannot be converted into
an integer and StrToInt must raise a convert error exception.
There are two more QC reports like issues
AnsiPos does not handle embedded nulls (like 'Pos')
qc.borland.com/wc/qcmain.aspx
SysUtils.StringReplace - strings containing #0 character handled incorrectly
qc.borland.com/wc/qcmain.aspx
They are both open and the conclusion is that #0 is a valid char in an
AnsiString and treating it as a terminator is a bug.
Best regards
Dennis Kjaer Christensen
Best regards
Dennis Kjaer Christensen
----------------------------------------------------------------------------
----
Jeg beskyttes af den gratis SPAMfighter til privatbrugere.
Den har indtil videre sparet mig for at f?3478 spam-mails
Betalende brugere får ikke denne besked i deres e-mails.
Hent en gratis SPAMfighter her.
 

Hi
The poll result is that it is not an error that StrToInt treats a AnsiString
as a nul terminated PChar.
Then we cannot use Length(S) to get the length of the input string.
Should we make it a general rule that the length field of an AnsiString does
not represent the string lenght?
And make a QC report that Length is buggy - it has to call StrLen to search
for the zero terminator?
Best regards
Dennis Kjaer Christensen
----------------------------------------------------------------------------
----
Jeg beskyttes af den gratis SPAMfighter til privatbrugere.
Den har indtil videre sparet mig for at f?3619 spam-mails
Betalende brugere får ikke denne besked i deres e-mails.
Hent en gratis SPAMfighter her.
 

"Dennis" <XXXX@XXXXX.COM>writes
function is designed to work with PChars. AFAIK no other string functions
are coded like this.
--
regards,
John
The Fastcode Project:
www.fastcodeproject.org/
 

Hi
buggy.
different behaviours for AnsiString functions. Sometimes the length is
defined by the length field and sometimes by the zero terminator. Sometimes
#0 are valid chars for an AnsiString and sometimes they have a special
meaning. Very weird if you ask me !
Best regards
Dennis Kjaer Christensen
----------------------------------------------------------------------------
----
Jeg beskyttes af den gratis SPAMfighter til privatbrugere.
Den har indtil videre sparet mig for at f?3619 spam-mails
Betalende brugere får ikke denne besked i deres e-mails.
Hent en gratis SPAMfighter her.
 

Hi
The Pascal version of does _ValLong not treat #0 as a terminator as the BASM
version does.
function _ValLong(const s: String; var code: Integer): Longint;
{$IFDEF PUREPASCAL}
var
I: Integer;
Negative, Hex: Boolean;
begin
I := 1;
code := -1;
Result := 0;
Negative := False;
Hex := False;
while (I <= Length(s)) and (s[I] = ' ') do Inc(I);
if I>Length(s) then Exit;
case s[I] of
'$',
'x',
'X': begin
Hex := True;
Inc(I);
end;
'0': begin
Hex := (Length(s)>I) and (UpCase(s[I+1]) = 'X');
if Hex then Inc(I,2);
end;
'-': begin
Negative := True;
Inc(I);
end;
'+': Inc(I);
end;
if Hex then
while I <= Length(s) do
begin
if Result>(High(Result) div 16) then
begin
code := I;
Exit;
end;
case s[I] of
'0'..'9': Result := Result * 16 + Ord(s[I]) - Ord('0');
'a'..'f': Result := Result * 16 + Ord(s[I]) - Ord('a') + 10;
'A'..'F': Result := Result * 16 + Ord(s[I]) - Ord('A') + 10;
else
code := I;
Exit;
end;
end
else
while I <= Length(s) do
begin
if Result>(High(Result) div 10) then
begin
code := I;
Exit;
end;
Result := Result * 10 + Ord(s[I]) - Ord('0');
Inc(I);
end;
if Negative then
Result := -Result;
code := 0;
end;
Best regards
Dennis Kjaer Christensen
----------------------------------------------------------------------------
----
Jeg beskyttes af den gratis SPAMfighter til privatbrugere.
Den har indtil videre sparet mig for at f?3690 spam-mails
Betalende brugere får ikke denne besked i deres e-mails.
Hent en gratis SPAMfighter her.
 

"Dennis" <XXXX@XXXXX.COM>writes
not get fixed, because it could break backwards compatibility.
--
regards,
John
The Fastcode Project:
www.fastcodeproject.org/
 

"Dennis" <XXXX@XXXXX.COM>writes
--
regards,
John
The Fastcode Project:
www.fastcodeproject.org/
 

Hi John
be a bad idea to fix the bug.
But the poll ended with 8-3 in favor of StrToInt not being buggy.
tech.groups.yahoo.com/group/fastcodeproject/surveys
Best regards
Dennis Kjaer Christensen
----------------------------------------------------------------------------
----
Jeg beskyttes af den gratis SPAMfighter til privatbrugere.
Den har indtil videre sparet mig for at f?3696 spam-mails
Betalende brugere får ikke denne besked i deres e-mails.
Hent en gratis SPAMfighter her.
 

Hi John
Best regards
Dennis Kjaer Christensen
----------------------------------------------------------------------------
----
Jeg beskyttes af den gratis SPAMfighter til privatbrugere.
Den har indtil videre sparet mig for at f?3696 spam-mails
Betalende brugere får ikke denne besked i deres e-mails.
Hent en gratis SPAMfighter her.
 

"Dennis" <XXXX@XXXXX.COM>writes
have changed functionality which has existed for 10+ years.
Lets just accept the poll results, treat StrToInt as a special case, and
move on.
--
regards,
John
The Fastcode Project:
www.fastcodeproject.org/