關于數據精确問題(20分)

zhangjensy · 2006-07-26

在DELPHI代碼中計算A+B=C+D 我們自己計算時A+B和C+D的結果相等,但不知為什麼當我把它放在一個條件語句中時,如 if A+B<>C+D then messagebox('不相等!'); 由于A,B和C,D都是字段值,但不知為什麼有時就相等,有時就不相等!但前提是我們看到和手工計算都相等! 再例如:A=89.45,B=23.34,C=78.12,D=34.67,其實這里A+B和C+D相等,但如果 if A+B<>C+D then messagebox('不相等!');就會執行messagebox('不相等!');! 如果A,B,C,D換成其它數據時,可能又不會執行這條語句messagebox('不相等!');! 不知有什麼辦法能使兩邊精度一樣!

zqw0117 · 2006-07-26

你的A,B,C三个变量是什么类型的?浮点数是相对精确的 ,只有整形才是绝对精确,对于浮点不能简单的用=/<>来比对,需要用CompareValue函数.

zhangjensy · 2006-07-26

A,B,C是DOUBLE,具體如何操作? 這樣好象不是很精确!

zqw0117 · 2006-07-26

Double当然是不精确的.需要用CompareValue Returns the relationship between two numeric values. Unit Math Category Comparison routines Delphi syntax: function CompareValue(const A, B: Integer): TValueRelationship; overload; function CompareValue(const A, B: Int64): TValueRelationship; overload; function CompareValue(const A, B: Single; Epsilon: Single = 0): TValueRelationship; overload; function CompareValue(const A, B: Double; Epsilon: Double = 0): TValueRelationship; overload; function CompareValue(const A, B: Extended; Epsilon: Extended = 0): TValueRelationship; overload; C++ syntax: extern PACKAGE TValueRelationship __fastcall CompareValue(const int A, const int B); extern PACKAGE TValueRelationship __fastcall CompareValue(const __int64 A, const __int64 B); extern PACKAGE TValueRelationship __fastcall CompareValue(const float A, const float B, float Epsilon = 0); extern PACKAGE TValueRelationship __fastcall CompareValue(const double A, const double B, double Epsilon = 0); extern PACKAGE TValueRelationship __fastcall CompareValue(const Extended A, const Extended B, Extended Epsilon =0); Description Call CompareValue to determine the relationship between two numeric values. When comparing floating-point values, CompareValue lets you specify a maximum difference to use when comparing values, so that they are considered the same if they are within that amount. A and B are the values to compare. Epsilon is the maximum amount by which A and B can differ and still be considered the same value. CompareValue returns LessThanValue if A is less than B (by more than Epsilon if A and B are floating-point numbers). EqualsValue if A is equivalent to B (the same, or within Epsilon if A and B are floating-point numbers). GreaterThanValue if A is larger than B (by more than Epsilon if A and B are floating-point numbers).

zhangjensy · 2006-07-26

我的數據只要數确2位小數就可,Epsilon這個如何定義?

zqw0117 · 2006-07-26

Example code : Comparing floating point numbers var  A : Single;  B : Single;  C : Single; begin  A := 23.0;  B := 23.0;  C := 23.1;  // Compare 2 equal floats  case CompareValue(A, B) of    LessThanValue    : ShowMessage('A < B');    EqualsValue      : ShowMessage('A = B');    GreaterThanValue : ShowMessage('A > B');  end;  // Compare 2 unequal floats  case CompareValue(A, C) of    LessThanValue    : ShowMessage('A < C');    EqualsValue      : ShowMessage('A = C');    GreaterThanValue : ShowMessage('A > C');  end;  // Compare 2 unequal floats - but allow for a difference of up to +/- 0.2  case CompareValue(A, C, 0.2) of    LessThanValue    : ShowMessage('A < C');    EqualsValue      : ShowMessage('A = C');    GreaterThanValue : ShowMessage('A > C');  end; end;

zqw0117 · 2006-07-26

如果编译的时候提示LessThanValue/EqualsValue/GreaterThanValue标识符不存在的话,你需要引用types单元.

djrj · 2006-07-26

你用numeric字段类型就不会有这个问题了,如两位小数可以这样numeric(18,2)

zhangjensy · 2006-07-26

djrj: 你說的情況不行! 我的程序全是numeric(18,2) 有更好的辦法?

zqw0117 · 2006-07-26

我的答案不能满足要求？还是楼主看不懂那段英文？

weiliu · 2006-07-26

浮点数比较时，可以比较双方的差异到一个足够小的小数，即可认为相等。

zhangjensy · 2006-07-26

if ((qsettlementdetail.FieldByName('invamount').AsFloat+qsettlementdetail.FieldByName('applyamount').AsFloat)<>qpublicsave.FieldByName('invamount').AsFloat) or ((qsettlementdetail.FieldByName('fbalance').AsFloat+qsettlementdetail.FieldByName('applyfamount').AsFloat)<>qpublicsave.FieldByName('fbalance').AsFloat) then        begin        MessageBox(GetActiveWindow(), PChar(qsettlementdetail.FieldByName('glcode').AsString+' Settlement Have Changed!'), PChar('Tungfai Group Ltd---Account System'), MB_ICONERROR + MB_OK + MB_SYSTEMMODAL);        abort;        end; 其中 (qsettlementdetail.FieldByName('invamount').AsFloat+qsettlementdetail.FieldByName('applyamount').AsFloat)<>qpublicsave.FieldByName('invamount').AsFloat) 和 (qsettlementdetail.FieldByName('fbalance').AsFloat+qsettlementdetail.FieldByName('applyfamount').AsFloat)<>qpublicsave.FieldByName('fbalance').AsFloat 在我們人工計算是相等,但不知為什麼有時明明相等的東西卻提示不等

kaida · 2006-07-26

为什么不用 zqw0117 的方法呢？ if (CompareValue(qsettlementdetail.FieldByName('invamount').AsFloat + qsettlementdetail.FieldByName('applyamount').AsFloat, qpublicsave.FieldByName('invamount').AsFloat)<>EqualsValue) or     (CompareValue(qsettlementdetail.FieldByName('fbalance').AsFloat + qsettlementdetail.FieldByName('applyfamount').AsFloat, qpublicsave.FieldByName('fbalance').AsFloat)<>EqualsValue)  then  begin    .......  end;

zhangjensy · 2006-07-27

我的程序中有C,B兩個變量. 當電腦計算C-B=A之后,我再放入語句中 if C=A+B then messagebox('ok'); 但當C和B換成不同的變量時,有時顯示不相等,有顯示相等! 我想問一下這樣只用CompareValue(C,A+B)這樣行不行?要不要加參數?

HunterTeam · 2006-07-27

限制一下精度,如小数点后两位: uses math; if Roundto(A+B,-2)<>RoundTo(C+D,-2) then  messagebox('不相等!');

SandWater · 2006-07-31

浮点数的比较不应该用精确的=来判断 可以比较两者的差值，看是否小于一个指定精度就好了 如Abs(A-B)<0.0001

關于數據精确問題(20分)

zhangjensy

Unregistered / Unconfirmed

zqw0117

Unregistered / Unconfirmed

zhangjensy

Unregistered / Unconfirmed

zqw0117

Unregistered / Unconfirmed

zhangjensy

Unregistered / Unconfirmed

zqw0117

Unregistered / Unconfirmed

zqw0117

Unregistered / Unconfirmed

djrj

Unregistered / Unconfirmed

zhangjensy

Unregistered / Unconfirmed

zqw0117

Unregistered / Unconfirmed

weiliu

Unregistered / Unconfirmed

zhangjensy

Unregistered / Unconfirmed

kaida

Unregistered / Unconfirmed

zhangjensy

Unregistered / Unconfirmed

HunterTeam

Unregistered / Unconfirmed

SandWater

Unregistered / Unconfirmed

Similar threads