Short Strings in a Variant Record? - string

Question

I'd like to be able to access sections of a short string as part of a record
Something like
TMyRecord = record
case Boolean of
True:
(
EntireString: String[20];
);
False
(
StringStart: String[8];
StringMiddle: String[4];
StringEnd: String[8];
);
end;
Is this possible or would I have to declare each char individually
TMyRecord = record
private
Chars: Array[1..20] of Char;
Function GetStringStart:String;
Procedure SetStringStart(Value: String);
public
Property StringStart: String read GetStringStart write SetStringStart; // Can I have properties on a record?
end;
Function GetStringStart: String;
begin
Result := Chars[1] + Char[2]....;
end;
Procedure SetStringStart(Value: String);
begin
for i := 1 to 8 do
begin
Chars[i] := Value[i];
end;
end;
Is this possible / worth the effort?

Answer 1

A Delphi short string contains more than just the string contents. The initial byte in the data structure contains the length of the string. This is why short strings are limited to 255 characters.
So, you can't use short strings in your variant array the way you propose.
What you could do is adapt your second approach based on getter and setter methods to be a bit more readable.
For example:
function TMyRecord.GetStringStart: string;
begin
SetString(Result, #Chars[1], 8);
end;
You might consider using a string rather than a char array, but it's a little hard to be 100% sure of that advice without knowing exactly what your underlying problem is.
As a final thought, why not turn the problem around? Store 3 strings: StartString, MiddleString and EndString. Then have a property backed with a getter and setter called EntireString. When you read EntireString it pieces it together from the 3 individual parts, and when you write to it it pulls the individual parts out. I suspect it would be easier that way around.

Answer 2

Your first sample doesn't consider the length byte. The memory layout looks like this:
case True:
L12345678901234567890
^....................
case False:
L12345678L1234L12345678
^........^....^........
(L = length byte).
Depending on your requirements (e.g.: Are the partial strings always 8, 4 and 8 Chars?) I'd try storing the partial strings and make EntireString the property, using System.Copy, StrUtils.LeftStr etc.

Answer 3

ShortString has an implied length, so your first example will map the length parts of the substrings on top of the main string.
Your second sample is the way to start, with these notes:
properties on records are possible
you should think of the length of each sub-string (or is it always a fixed array of 20 characters?)
Edit
It totally depend on the reason you want this, and mixing character arrays and strings will get you into trouble because strings can be shorter than the array length.
Small example:
program VariantRecordsWithCharactersAndStrings;
{$APPTYPE CONSOLE}
uses
SysUtils,
Math;
const
Size20 = 20;
Size8 = 8;
Size4 = 4;
type
TChar20 = array[0..Size20-1] of Char;
TChar8 = array[0..Size8-1] of Char;
TChar4 = array[0..Size4-1] of Char;
TMyRecord = record
class var FillCharValue: Byte;
function GetEntireString: string;
function GetStringStart: string;
function GetStringMiddle: string;
function GetStringEnd: string;
procedure SetEntireString(const Value: string);
procedure SetStringStart(const Value: string);
procedure SetStringMiddle(const Value: string);
procedure SetStringEnd(const Value: string);
property EntireString: string read GetEntireString write SetEntireString;
property StringStart: string read GetStringStart write SetStringStart;
property StringMiddle: string read GetStringMiddle write SetStringMiddle;
property StringEnd: string read GetStringEnd write SetStringEnd;
procedure SetCharArray(const CharArrayPointer: PChar; const CharArraySize: Integer; const Value: string);
case Boolean of
True:
(
CharFull: TChar20;
);
False:
(
CharStart: TChar8;
CharMiddle: TChar4;
CharEnd: TChar8;
);
end;
function TMyRecord.GetEntireString: string;
begin
Result := CharFull;
end;
function TMyRecord.GetStringStart: string;
begin
Result := CharStart;
end;
function TMyRecord.GetStringMiddle: string;
begin
Result := CharMiddle;
end;
function TMyRecord.GetStringEnd: string;
begin
Result := CharEnd;
end;
procedure TMyRecord.SetEntireString(const Value: string);
begin
SetCharArray(CharFull, SizeOf(CharFull), Value);
end;
procedure TMyRecord.SetCharArray(const CharArrayPointer: PChar; const CharArraySize: Integer; const Value: string);
begin
FillChar(CharArrayPointer^, CharArraySize, FillCharValue);
Move(Value[1], CharArrayPointer^, Min(CharArraySize, SizeOf(Char)*Length(Value)));
end;
procedure TMyRecord.SetStringStart(const Value: string);
begin
SetCharArray(CharStart, SizeOf(CharStart), Value);
end;
procedure TMyRecord.SetStringMiddle(const Value: string);
begin
SetCharArray(CharMiddle, SizeOf(CharMiddle), Value);
end;
procedure TMyRecord.SetStringEnd(const Value: string);
begin
SetCharArray(CharEnd, SizeOf(CharEnd), Value);
end;
var
MyRecord: TMyRecord;
procedure Dump();
begin
Writeln(MyRecord.EntireString);
Writeln(MyRecord.StringStart);
Writeln(MyRecord.StringMiddle);
Writeln(MyRecord.StringEnd);
end;
procedure TestWithFillCharValue(const FillCharValue: Byte);
begin
Writeln('Testing with FillCharValue ', FillCharValue);
TMyRecord.FillCharValue := FillCharValue;
MyRecord.EntireString := '123456789001234567890';
Dump();
MyRecord.StringStart := 'AAA';
MyRecord.StringMiddle := 'BBB';
MyRecord.StringEnd := 'CCC';
Dump();
end;
begin
try
TestWithFillCharValue(0); // this will truncated all the sub arrays when you pass strings that are too short
TestWithFillCharValue(20); // when using Unicode, this fails even more horribly
Write('Press <Enter>');
Readln;
except
on E: Exception do
Writeln(E.ClassName, ': ', E.Message);
end;
end.
This class does more or less what you want:
it has overlapping data structures
when you assign the arrays: no problem
when you assign the strings: be aware when strings get to short

Answer 4

As other stated, it won't work, because the variant-sized record will add some lengths for StringStart/StringMiddle/StringEnd in the middle of the EntireString type.
You are confusing the *char type of C with the pascal shortstring type. There is an hidden character at position [0] which is the shortstring length.
You could use regular string type, then split in on purpose:
procedure StringSplit(const EntireString: string; out StringStart, StringMiddle, StringEnd: string);
begin
if length(EntireString)<>20 then
exit;
StringStart := copy(EntireString,1,8);
StringMiddle := copy(EntireString,9,4);
StringEnd := copy(EntireString,13,8);
end;
Note that the out parameter type will set all output String* variables into '' before calling the function.
This version will expect entering entire string of 20 chars long.
You could use shortstrings, but with custom types of the exact length, if you want to avoid hidden copies from/to string[255] (which occur when you use a shortstring type and work with string[n] with n<255):
type
String20 = string[20];
String4 = string[4];
String8 = string[8];
procedure StringSplit(const EntireString: String20; out StringStart: String8;
out StringMiddle: String4; out StringEnd: String8);
begin
if length(EntireString)<>20 then
exit;
StringStart := copy(EntireString,1,8);
StringMiddle := copy(EntireString,9,4);
StringEnd := copy(EntireString,13,8);
end;