float field validation error

[Packs]

Code: Pascal  [Select][+][-]

1. procedure Tdm_statement.pr_update_overall_cgpa_grade();
2. var
3.   Fcgpa: double = 0.00;
4.   Fmultiple: double = 0.00;
5.   FMinsgpa: double = 0.00;
6.   FMaxsgpa: double = 0.00;
7.   lCondition : boolean = false ;
8. begin
9.   if (buf_sem_main.RecordCount <= 0) then
10.   begin
11.     ShowMessage('Record count :');
12.     Exit;
13.   end;
14.  
15.   buf_marks.First;
16.  
17.   buf_sem_main.First;
18.   while not buf_sem_main.EOF do
19.   begin
20.     if buf_sem_main.FieldByName('aadhaar').AsLargeInt = 868819501724 then
21.     begin
22.       ShowMessage('Yes');
23.     end;
24.  
25.     if buf_marks.Locate('Aadhaar', buf_sem_main.FieldByName(
26.       'aadhaar').AsLargeInt, []) then
27.     begin
28.       if buf_sem_main.FieldByName('total_multiple').AsFloat > 0 then
29.       begin
30.  
31.         Fcgpa := 0;
32.         Fmultiple := 0;
33.  
34.         Fmultiple := buf_sem_main.FieldByName('total_multiple').AsFloat;
35.         Fcgpa := (Fmultiple / Overallcreditpoints);
36.         Fcgpa := Math.RoundTo(Fcgpa, -3);
37.  
38.  
39.         buf_slab.First;
40.         while not buf_slab.EOF do
41.         begin
42.           FMinsgpa := 0;
43.           FMaxsgpa := 0;
44.           FMinsgpa := buf_slab.FieldByName('min_gpa').AsFloat;
45.           FMaxsgpa := buf_slab.FieldByName('max_gpa').AsFloat;
46.  
47.  
48.           lCondition := ((Fcgpa >= FMinsgpa) and (Fcgpa <= FMaxsgpa)) ;
49.           if lCondition then
50.           begin
51.             buf_marks.Edit;
52.             buf_marks.FieldByName('sem_comma').AsString :=
53.               buf_sem_main.FieldByName('commadata').AsString;
54.  
55.             //buf_marks.FieldByName('overall_cgpa').AsString := 'NA';
56.             //buf_marks.FieldByName('overall_grade').AsString := 'NA';
57.  
58.             buf_marks.FieldByName('overall_cgpa').AsString := Format('%.2f', [Fcgpa]);
59.             buf_marks.FieldByName('overall_grade').AsString :=
60.               buf_slab.FieldByName('grade').AsString;
61.             buf_marks.Post;
62.  
63.             Break;
64.           end;
65.  
66.           buf_slab.Next;
67.         end;
68.  
69.       end;
70.       //buf_marks.Post;
71.     end;
72.  
73.     buf_sem_main.Next;
74.   end;
75.  
76. end;
77.  
78.  

in this above code what was the mistake.

when I am geeting value for following variables 

Fcgpa = 8.99
FMinsgpa = 8
FMaxsgpa = 8.99

my condition is not get staified .( lCondition := ((Fcgpa >= FMinsgpa) and (Fcgpa <= FMaxsgpa))  )

[Zvoni]

when I am geeting value for following variables 

Fcgpa = 8.99
FMinsgpa = 8
FMaxsgpa = 8.99

my condition is not get staified .( lCondition := ((Fcgpa >= FMinsgpa) and (Fcgpa <= FMaxsgpa))  )

Then check the Value of FMaxsgpa.
If your condition is not satisfied, that means that FMaxsgpa<Fcgpa.

At a guess: Rounding error.
Maybe try to use an Epsilon-Threshold

[Packs]

i Have check the watch window and share the value

[Packs]

is there any method to find exact value of the float vairable

[Thaddy]

is there any method to find exact value of the float vairable

No, because that is not always possible for float comparisons like you describe. But FreePascal has a function SameValue (in math), where you can define an acceptible margin to compare two float values that corrects for the imprecise nature of floating point values.

Code: Pascal  [Select][+][-]

1. program test;
2. {$mode objfpc}
3. uses math;
4. var
5.   x:single = 10.0;
6.   y:single = 3.0;
7.   Epsilon:Single  = 0.000000001;
8. begin
9.    writeln(SameValue((x/y) * 3, x)); // should be false
10.    writeln(SameValue((x/y)*3, x,Epsilon); // Should be true
11.    Readln;
12. end.

Untested.

[Zvoni]

Maybe add a "rounding" in Line 45 like you did in Line 36

Code: Pascal  [Select][+][-]

1. FMaxsgpa := Math.RoundTo(buf_slab.FieldByName('max_gpa').AsFloat, -3);

[dseligo]

As Thaddy told you, use SameValue.

But there is an alternative. Depends on precision you need and calculations you do, you could also use 'currency' type.

Currency is fixed to 4 decimal places, but you can compare directly values.
There is equivalent support in FPC for currency as for other float types. You have: StrToCurr, FormatCurr, FieldByName('some DB field').AsCurrency, and so on.

[Thaddy]

Note that in my example the smallest theoretical epsilon is actually 10.0 -((10/3)* 3)
which uses the error between the result of the calculation and the value to compare against. That is usually just the loss of one bit "precision" that can be expressed in a float value. Actually it is proof that comparisons may not be so precise when comparing float values.
Note that for my purposes I tend to prefer two bits "precision" lost.
Float value comparisons must be considered approximations

[Paolo]

Actually it is proof that comparisons may not be so precise when comparing float values

I don't think that is something wrong in the comparison, the comparison said the truth: the two numbers are different.

what something can create confusion is also the "default" float value shown in the "debug wathces", for example

Code: Pascal  [Select][+][-]

1. procedure TForm1.FormCreate(Sender: TObject);
2. var
3.   a, b, c, d : double;
4. begin
5.   a:=0.1;
6.   b:=0.2;
7.   c:=0.3;
8.   d:=a+b;
9.   if (c = d) then
10.     beep;
11. end;
12.  

this code will never do "beep", but the comparison said the truth, infact (see picture 1) the binary repesentation of C and D confirms they are different, but if you look at the "default" value it says something false, both are 0.3  (picture-2).
I understand that it can be annoying to see similar numbers in watches, and they can be much-much longer: (eg 64-bit double, if I am not wrong)
C=+0.299999999999999988897769753748434595763683319091796875
D=+0.3000000000000000444089209850062616169452667236328125
So as main advice look always at the binary content of "float" variable, do not rely on other representation for float variable and even on what the watches say when “Default” is selected.

is there any method to find exact value of the float variable

If you mean decimal representation: no, that I know of. You have to write code for this purpose (see the real value of C and D above).
If you want see the values during coding look at the Memory Dump representation in the watches as said above.

[Paolo]

sorry, the  pictures

[Packs]

Thanks zvoni.

It is working. I have used round for min and Max variable.

[Thaddy]

A bitwise comparison will also reveal that they are different. That can be done with a simple hardcast.

[Packs]

Please elaborate in detail.

[Thaddy]

What I mean is simply to compare the bit patterns, easy:

Code: Pascal  [Select][+][-]

1. program expfrac;
2. uses sysutils;
3. var a,b:single;
4. begin
5.   a:= 33 * 100/33;
6.   b:= 100.0;
7.   writeln('bit pattern of the mantissa''s')
8.   writeln(a.mantissa.tobinstring);
9.   writeln('comparison of the bit patterns of exp and frac');
10.   writeln ((a.frac = b.frac) and  (a.exp = b.exp));
11. end.

In some cases it may be enough to just compare the mantissa directly, because of the insignificant bit.
All three used functions are a bitwise representation of the values.
See the documentation for typehelpers in sysutils.