Should variable be initialized before call to function? - delphi

Question

When I call Functions to get a value, I usually initialize varible, in case function fails or doesn't return anything and I want to avoid dealing with uninitialized variable. I do the same for string, integer or any other type.
Example for integer variables:
vPropValue := 0;
vPropValue := GetPropValue(vObject,'Height');
IF vPropValue > 0 Then
...
this it the most common how I use it.
I know I could use:
If GetPropValue(vObject,'Height') > 0 Then
...
but with first example I avoid multiple calls to function, if I need result again later in the code.
Same for string (even though i know local strings are initialized to empty string, while integers are not an can hold any value)
vName := '';
vName := GetObjectName(vObject,'ObjectName');
IF Trim(vPropStrValue) <> '' Then
...
I know I could take steps to avoid duplicate value assignment,like making sure Function returns 0 if everything fails. But I have 100s of functions and I can't rely I never made a mistake how functions handle everything and I'm sure some don't return 0, if everything fails.
I'm trying to understand why this is not desirable practice and how to best avoid it.
EDIT
Here is example where function doesn't return proper value or 0:
function GetValue(vType:integer):integer;
begin
if vType=1 then
Result:=100
else if (vType>2) and (vType<=9) then
Result:=200;
end;
procedure TForm1.Button1Click(Sender: TObject);
var vValue:integer;
begin
vValue:=GetValue(11);
Button1.Caption:=IntToStr(vValue);
end;
In this case the value returned from function is some random number.
In this case the initialization appears to be valid approach. OR NOT?
EDIT 2:
As David pointed out in his answer, correct, there was a warning
[dcc32 Warning] Unit1.pas(33): W1035 Return value of function 'GetValue' might be undefined
but, I ignored it, for no reason, just didn't look there. As it let me compile it, I thought it was OK. So, I did look for the warning and I 'fixed' quite a few functions that had similar issue, because of all IFs Result might not have been defined.
EDIT 3 & CONCLUSION:
I hope it adds to the scope of question and explanation:
Maybe an example of another twist that I use in most of my Functions, would also explain why I thought my initialization of variable was needed, is that I wasn't sure my Functions would behave correctly all the times, especially in case of nested function. Mots of them still are set like this:
function GetProperty(vType:integer):integer;
begin
Try
if vType = 99 then
Result:=GetDifferentProperty(vType)// <-- Call to another Function, that could return whatever...
else
begin
if vType=1 then
Result:=100
else if (vType>2) and (vType<=9) then
Result:=200;
end;
except
end;
end;
Now I'm addressing these Try Except End; but some functions are 10 years old and to expect them to work 100%, based on my experience back then, is not something to rely on.
As the only developer on this project, I assume I should trust my functions (and he rest of my code), but I can't imagine in multiple developers environment that all function are set up properly.
So my conclusion: since I didn't take care of the basics - properly designed Functions, I need to have all these checks (variable initialization, Try Except lines..) and probably some other unneccessary stuff.

Answer 1

Assuming that vPropValue is a local variable then this code
vPropValue := 0;
vPropValue := GetPropValue(vObject,'Height');
is indistinguishable from
vPropValue := GetPropValue(vObject,'Height');
A simpler example might be like so:
i := 0;
i := 1;
What is the point of assigning 0 to i and then immediately assigning 1 to i? So, you would surely never write that. You would write:
i := 1;
In your code, at the top of this answer, you assign twice to the same variable. The value assigned in the first assignment is immediately replaced by the value assigned in the second assignment. Therefore, the first assignment is pointless and should be removed.
The second example is a little more complex. Assuming that your functions are written correctly, and always assign to their return value, and that vName is a local variable, then
vName := '';
vName := GetObjectName(vObject,'ObjectName');
is indistinguishable from
vName := GetObjectName(vObject,'ObjectName');
The reason why I have added an extra proviso relates to a quirk of the implementation of function return values, discussed below. The difference between this case and the case above is the return value type. Here it is a managed type, string, whereas in the first example the type is a simple Integer.
Again, given the proviso about the function always assigning to the return value, the first assignment is pointless because the value is immediately replaced. Remove that first assignment.
Regarding the function in your edit, the compiler will warn you of its erroneous implementation if you enable hints and warnings. The compiler will tell you that not all code paths return a value.
function GetValue(vType:integer):integer;
begin
if vType=1 then
Result:=100
else if (vType>2) and (vType<=9) then
Result:=200;
end;
If neither condition is met, then no value is assigned to the result variable. This function should be:
function GetValue(vType:integer):integer;
begin
if vType=1 then
Result:=100
else if (vType>2) and (vType<=9) then
Result:=200
else
Result:=0;
end;
I cannot stress how important it is that you always return a value from a function. In fact it is a terrible weakness that Delphi even allows your function to be compiled.
The reason that your double assignment sometimes appears useful to you is due to a quirk of of the implementation of function return values in Delphi. Unlike almost all other languages a Delphi function return value for certain more complex types is actually a var parameter. So this function
function foo: string;
is actually, semantically, the same as this:
procedure foo(var result: string);
This is a really odd decision made by the Delphi designers. In most other languages, like C, C++, C#, Java etc., a function return value is like a by-value parameter passed from callee to caller.
This means that you can, if you wish to be perverse, pass values into a function via its return value. For instance, consider this code:
// Note: this code is an example of very bad practice, do not write code like this
function foo: string;
begin
Writeln(Result);
end;
procedure main;
var
s: string;
begin
s := 'bar';
s := foo;
end;
When you call main, it will output bar. This is a rather strange implementation detail. You should not rely on it. Let me repeat myself. You should not rely on it. Do not get in the habit of initializing return values at the call site. That leads to unmaintainable code.
Instead follow the simple rule of ensuring that the function return value is always assigned by the function, and never read before it has been assigned.
More detail on the implementation of function return values is provided by the documentation, with my emphasis:
The following conventions are used for returning function result
values.
Ordinal results are returned, when possible, in a CPU register. Bytes are returned in AL, words are returned in AX, and double-words
are returned in EAX.
Real results are returned in the floating-point coprocessor's top-of-stack register (ST(0)). For function results of type Currency,
the value in ST(0) is scaled by 10000. For example, the Currency value
1.234 is returned in ST(0) as 12340.
For a string, dynamic array, method pointer, or variant result, the effects are the same as if the function result were declared as an
additional var parameter following the declared parameters. In other
words, the caller passes an additional 32-bit pointer that points to a
variable in which to return the function result.
Int64 is returned in EDX:EAX.
Pointer, class, class-reference, and procedure-pointer results are returned in EAX.
For static-array, record, and set results, if the value occupies one byte it is returned in AL; if the value occupies two bytes it is
returned in AX; and if the value occupies four bytes it is returned in
EAX. Otherwise, the result is returned in an additional var parameter
that is passed to the function after the declared parameters.

Answer 2

The following code (A)
vPropValue := 0;
vPropValue := GetPropValue(vObject,'Height');
is indistinguishable from (B)
vPropValue := GetPropValue(vObject,'Height');
The question of whether GetPropValue is "written correctly" is totally irrelevant.
Let's consider what happens even if you did write GetPropValue incorrectly E.g.
function GetPropValue(AObject: TObject; AStr: String): Integer;
begin
if AStr = 'Hello' then Result := 5;
end;
As you know when the input AStr is anything other than "Hello" the result of the function will be pretty much random. (For the sake of the discussion, lets assume it will return -42.)
Code block (A) will do the following:
Set vPropValue to 0
Then set vPropValue to - 42
Code block (B) will simply set vPropValue to - 42 immediately.
TIP: There is no point in writing a wasteful line of code just because you're worried you might have made a mistake in a function you call.
First, As David points out, you can avoid many mistakes simply by paying attention to your compiler hints and warnings.
Second, that sort of "paranoid" coding simply leads to more wasteful code because now you have to start considering invalid values as possible results.
This becomes worse when one day your "safe-value" is actually a valid value. E.g. how would you tell the difference between "default 0" and "correctly returned 0"?
Don't make programming artificially difficult by bloating code with unnecessary redundancies.
Side Note
There are a couple of special situations where the code can behave differently. But you should in any case avoid the designs that lead to those situations because they make it much more difficult to maintain the code.
I mention them purely for the sake of completeness, the advice above still stands.
1) if vPropValue is implemented as a property, the setter could have side-effects that cause different behaviour. While there's nothing wrong with properties, when they do unexpected things you have a serious problem on your hands.
2) if vPropValue is a field on the class (or worse a global variable), then (A) and (B) could behave differently but only if GetPropValue raises an exception. This is because the exception would prevent the result from being assigned. Note this is something to avoid except in special cases because it does make it more difficult to reason about what your code is doing.
And in fact, this is something that makes it so much more important to avoid the redundant initialisations. You want your special case code to look stand out from the rest.

Answer 3

Scavenging my advices from top-level comments in case Pastebin fails
function GetValueUpdate3(vType:integer):integer;
begin
// with SOME types like Integer you can use Pascal case-block instead of error-prone many-ifs-ladder
case vType of
99: Result:=GetDifferentProperty(vType);// <-- Call to another Function, that could return whatever...
1: Result:=100;
3..9: Result:=200;
else Result := 12345; // initialization with safe default value for illegal input like vType=2
end; // case-block
end;
function GetValueUpdate3e(vType:integer):integer;
begin
case vType of
99: Result:=GetDifferentProperty(vType);// <-- Call to another Function, that could return whatever...
1: Result:=100;
3..9: Result:=200;
else raise EInvalidArgument.Create('prohibited value - GetValue( ' + IntToStr(vType) +' )' );
// runtime eror when vType = 2 or any other illegal input
end;
end;
function GetValueUpdate1(vType:integer):integer;
begin
Result := 12345; // initialization with safe default value;
if vType=1 then Exit(100); // special value for special case #1
if (vType>2) and (vType<=9) then Exit(200); // special value for special case #2
// exit with default value
end;
procedure TForm1.Button1Click(Sender: TObject);
var vValue:integer;
begin
vValue:=GetValue(11);
Button1.Caption:=IntToStr(vValue);
end;
// http://stackoverflow.com/questions/33927750

Answer 4

You can use Assertions to validate if your input is correct.
Assertions help a lot in finding logical errors in your code.
Using assertions forces you to think about valid and invalid input data and helps writing better code.
Enabling and disabling assertions at compile time is done via the \$C or \$ASSERTIONS compiler (global switch)
In your example the function GetValue could use an assertion as follows:
function GetValue(vType:integer):integer;
begin
Assert((vType >= 1) and (vType <=9), 'Invalid input value in function GetValue');
if vType=1 then
Result:=100
else if (vType>2) and (vType<=9) then
Result:=200
else Result := 0; // always catch the last else!
end;
In addition every if-statement should catch the final else! (In my opinion ALWAYS!)