Range check error

[Actor]

Originally Posted by TonyMLake

Well, the first problem I MIGHT see is that the file is only looking for COMMA, but not looking out for double quotes.

If that were the problem then val() would return a value of err that is greater than 0. Err would be the position of the double quote, an illegal character, in the string. The function would than return 0 which is not out of range.

Originally Posted by TonyMLake

Remember, the comma character does not ALWAYS indicate a new field... when a comma is inside double quotes it doesn't indicate a new field, and thus your field count would be off on the times that there was a sentence like "First time Lasix, under 121 add 2" (any text with a comma in it) preceding your target field because the comma inside the quotation marks... HOWEVER, this may not be a problem. It depends on how you're calling get_int.

Please post the line that calls "get_int".


Thanks

846 .. 855 : past[i - 845].speed := abs(get_int(f)) ;	// the offending line

[DLigett]

So if a comma or two is missing from the file, get_int() might return 999999 (longint) and try to stuff it into a simple integer?

I don't know the size of a longint in Free Pascal.

In get_int() you could test for size 32767 and squawk. Try on a known bad file.

[vegasone]

One of the tracks, either EMD or EVD (I think) keeps coming up with a funky speed rating something like 1781 which may be causing your problem. I would have to run into it again to know for sure but it is consistently one track.

[TonyMLake]

Originally Posted by Actor

If that were the problem then val() would return a value of err that is greater than 0. Err would be the position of the double quote, an illegal character, in the string. The function would than return 0 which is not out of range.

846 .. 855 : past[i - 845].speed := abs(get_int(f)) ;	// the offending line

What does this display?

until ch = COMMA ;
        ShowMessage(str);
	val(str, i, err) ;

[DLigett]

Originally Posted by TonyMLake

What does this display?

until ch = COMMA ;
        ShowMessage(str);
	val(str, i, err) ;

I vote for...

if i > 32767 then ShowMessage(str + ' will break our caller');

[TonyMLake]

Originally Posted by DLigett

I vote for...

if i > 32767 then ShowMessage(str + ' will break our caller');

Good point, and if that parser throws something "weird" based on it being a garbage input, that could happen.

THEREFORE, call the "ShowMessage" right before the val(str, i, err) ; statement, like the last snippet, up above.

[headhawg]

Originally Posted by ebcorde

I've never seen comma within a string yet, else my program would not be working, it works 99.9% of the time I wrote it 4 years ago. Too lazy to track Bris's mess ups.

Bris had in embedded commas in the field that had foreign race comments. (I can't recall the field number.) They would be in some of the Cali track files. I don't know if that's been fixed as it's been years since I looked at it. I think I have also seen them in owner's name or the race name fields as well (can't remember which). It drove me nuts. Wrote a routine to count the number of commas so I knew which files had extras.

[TonyMLake]

Originally Posted by headhawg

Bris had in embedded commas in the field that had foreign race comments. (I can't recall the field number.) They would be in some of the Cali track files. I don't know if that's been fixed as it's been years since I looked at it. I think I have also seen them in owner's name or the race name fields as well (can't remember which). It drove me nuts. Wrote a routine to count the number of commas so I knew which files had extras.

That's not actually an error (although you're right that Brisnet has a few errors but not as many as Formulator). Commas are legitimate input in .csv files, which is why you need a secondary text delimiter, hence text is enclosed in double quotes to tell applications (or application developers) to ignore the commas within the first and second double quote.

Of course, it doesn't have to be commas, and it doesn't have to be double quotes, but those are common delimiters.

This is common:

32, 45, T, F, "All American Race", 45, 92.3, 0,,,,"First Race Ever, Tired Old Cows", "Scott Stephens", n,,"", *,2

Note the comma in the text field between the double quotes. That's the only way to tell the app it's not a new field, so, it was designed that way.

[TonyMLake]

Whatever became of this, Actor?

[Actor]

Originally Posted by TonyMLake

Whatever became of this, Actor?

Sorry if I seem to have dropped out of sight.

I've been forced to put it on hold for a while. The tenant of one of my rental houses has moved out and left the place a mess. You would not believe how much of a mess he left. I've been working trying to clean up the place enough so that I can go to work with a mop and broom, plus repair some damage.

I do appreciate everyone's input.

[DJofSD]

Here's some code excerpts from an old, old program I wrote years ago using Delphi V1. It reads single file BRIS files. It treats all data tokens as strings and anything that is converted to numeric values is done elsewhere in the app. It has never failed to correctly parse a BRIS data file.

var
  Form1: TForm1;
  SrcFile: System.Text;   {regular ascii text file}
  F, revF: String;
  s0: string; {initial input string}
  ss: array[1..1436] of String; {input string broken out into substrings}
.....

const
  NULL: String = ''; // match what BASIC produces
  MAXTOKENS: Integer = 1435;
  MAXBRISPPRACES: Integer = 10; // BRIS has 10 pace lines per horse
  DOT: String = '.';
  BLANK: String = ' ';
.....

procedure TForm1.sparser2;
// destructive - modifies string s0
var
  sstoken, fchar, endquote, nextcomma: Integer;
  ls, lt: Integer;
const
  QUOTE: String = #34; { a double quotation mark '"' -- deer track}
  COMMA: String = #44;
  TERM: String = #34#44; { an end of a string '",' }
begin
  fchar := 1;
  for sstoken := 1 to MAXTOKENS do begin
    ls := Length(s0);
    if Copy(s0, 1, 1) = QUOTE then {this token is a BASIC string}
      begin
      endquote := Pos(TERM, s0); {find matching '",'}
      if sstoken = MAXTOKENS then {endquote should be zero}
        begin
        ss[sstoken] := Copy(s0, 2, ls - 2);
        Inc(currentlinetokens);
        break; {leave the for loop, clean up and return}
        end
      else
        begin
        if endquote - fchar > 1 then {more than just '"",', i.e. not a null}
          begin
          ss[sstoken] := Copy(s0, 2, endquote - 2);
          Inc(currentlinetokens);
          {remove token from s0 from '"' to '",' inclusive}
          Delete(s0, 1, endquote + 1);
          end
        else {nothing inbetween, just '"",', assign null }
          begin
          ss[sstoken] := NULL; // 'NULL STRING ' + IntToStr(sstoken);
          Inc(currentlinetokens);
          Delete(s0, 1, 3);
          end
        end
     end
    else {this token not a BASIC string (no '"' as 1st char);
           could be numeric (keep as characters) or null string,
           i.e. just a comma}
      begin
        nextcomma := Pos(COMMA, s0);
        lt := nextcomma - 1;
        if lt > 0 then
          ss[sstoken] := Copy(s0, 1, nextcomma - 1)
        else
          ss[sstoken] := NULL; // 'NULL TOKEN ' + IntToStr(sstoken);
        Inc(currentlinetokens);
        Delete(s0, 1, nextcomma);
      end
  end;
end; // end proc sparser2

procedure TForm1.getallBRISdata;
// read input .DRF file, parse records and insert into StringGrid1 (ALL DATA)
var
  icol, j: Integer;
begin
  TotalRaces := 0; // before any lines read, no races
  while not eof(SrcFile) do begin
    Inc(passes); { count the number of times through the loop }
    Readln(SrcFile, s0); {read a line}
    Inc(TotalBytes, Length(s0)); { count total number of bytes read }
    currentlinetokens := 0;
    sparser2;
.....

[Red Knave]

Originally Posted by Actor

So I changed the variable to a signed integer. That should have cured the problem. It did not.

Not knowing much about this language (although I'm knocked out that it is free with all the bells and whistles it has) is "signed integer" the same as "longint"?
It seems pretty obvious that it is the assignment of the result of the get_int function that is throwing the error so are you sure you changed the type of right var?

[Actor]

Originally Posted by Red Knave

Not knowing much about this language (although I'm knocked out that it is free with all the bells and whistles it has) is "signed integer" the same as "longint"?

• Signed integer is 16 bits denoting the range -32768 .. 32767. The language uses integer.
• Unsigned integer is 16 bits denoting the range 0 .. 65535. The language uses word.
• longint is 32 bits denoting the range -2147483648 .. 2147483647

Originally Posted by Red Knave

It seems pretty obvious that it is the assignment of the result of the get_int function that is throwing the error so are you sure you changed the type of right var?

Yes.

[DLigett]

Originally Posted by Actor

• Signed integer is 16 bits denoting the range -32768 .. 32767. The language uses integer.
• Unsigned integer is 16 bits denoting the range 0 .. 65535. The language uses word.
• longint is 32 bits denoting the range -2147483648 .. 2147483647

Yes.

Right, the function returns a longint, and stuffs it into the integer '.speed' which will fail once in a while.

[Red Knave]

Originally Posted by Actor

• longint is 32 bits denoting the range -2147483648 .. 2147483647

Can you change the .speed type to longint, turn on range checking, retry with a 'bad' data file and display the results so you can see the offending value?