I have done some really serious refactoring of my text editor. Now there is much less code, and it is much easier to extend the component. I made rather heavy use of OO design, such as abstract classes and interfaces. However, I have noticed a few losses when it comes to performance. The issue is about reading a very large array of records. It is fast when everything happens inside the same object, but slow when done via an interface. I have made the tinyest program to illustrate the details:
unit Unit3;
interface
uses
Windows, Messages, SysUtils, Variants, Classes, Graphics, Controls, Forms,
Dialogs;
const
N = 10000000;
type
TRecord = record
Val1, Val2, Val3, Val4: integer;
end;
TArrayOfRecord = array of TRecord;
IMyInterface = interface
['{C0070757-2376-4A5B-AA4D-CA7EB058501A}']
function GetArray: TArrayOfRecord;
property Arr: TArrayOfRecord read GetArray;
end;
TMyObject = class(TComponent, IMyInterface)
protected
FArr: TArrayOfRecord;
public
procedure InitArr;
function GetArray: TArrayOfRecord;
end;
TForm3 = class(TForm)
procedure FormCreate(Sender: TObject);
private
{ Private declarations }
public
{ Public declarations }
end;
var
Form3: TForm3;
MyObject: TMyObject;
implementation
{$R *.dfm}
procedure TForm3.FormCreate(Sender: TObject);
var
i: Integer;
v1, v2, f: Int64;
MyInterface: IMyInterface;
begin
MyObject := TMyObject.Create(Self);
try
MyObject.InitArr;
if not MyObject.GetInterface(IMyInterface, MyInterface) then
raise Exception.Create('Note to self: Typo in the code');
QueryPerformanceCounter(v1);
// APPROACH 1: NO INTERFACE (FAST!)
// for i := 0 to high(MyObject.FArr) do
// if (MyObject.FArr[i].Val1 < MyObject.FArr[i].Val2) or
// (MyObject.FArr[i].Val3 < MyObject.FArr[i].Val4) then
// Tag := MyObject.FArr[i].Val1 + MyObject.FArr[i].Val2 - MyObject.FArr[i].Val3
// + MyObject.FArr[i].Val4;
// END OF APPROACH 1
// APPROACH 2: WITH INTERFACE (SLOW!)
for i := 0 to high(MyInterface.Arr) do
if (MyInterface.Arr[i].Val1 < MyInterface.Arr[i].Val2) or
(MyInterface.Arr[i].Val3 < MyInterface.Arr[i].Val4) then
Tag := MyInterface.Arr[i].Val1 + MyInterface.Arr[i].Val2 - MyInterface.Arr[i].Val3
+ MyInterface.Arr[i].Val4;
// END OF APPROACH 2
QueryPerformanceCounter(v2);
QueryPerformanceFrequency(f);
ShowMessage(FloatToStr((v2-v1) / f));
finally
MyInterface := nil;
MyObject.Free;
end;
end;
{ TMyObject }
function TMyObject.GetArray: TArrayOfRecord;
begin
result := FArr;
end;
procedure TMyObject.InitArr;
var
i: Integer;
begin
SetLength(FArr, N);
for i := 0 to N - 1 do
with FArr[i] do
begin
Val1 := Random(high(integer));
Val2 := Random(high(integer));
Val3 := Random(high(integer));
Val4 := Random(high(integer));
end;
end;
end.
When I read the data directly, I get times like 0.14 seconds. But when I go through the interface, it takes 1.06 seconds.
Is there no way to achieve the same performance as before with this new design?
I should mention that I tried to set PArrayOfRecord = ^TArrayOfRecord and redefined IMyInterface.arr: PArrayOfRecord and wrote Arr^ etc in the for loop. This helped a lot; I then got 0.22 seconds. But it is still not good enough. And what makes it so slow to begin with?
Simply assign the array to a local variable before iterating through the elements.
What you're seeing is that the interface methods calls are virtual and have to be called through an indirection. Also, the code has to pass-through a "thunk" that fixes up the "Self" reference to now point to the object instance and not the interface instance.
By making only one virtual method call to get the dynamic array, you can eliminate that overhead from the loop. Now your loop can go through the array items without the extra overhead of the virtual interface method calls.
You're comparing oranges with apples, as the first test reads a field (FArr), while the second test reads a property (Arr) that has a getter assigned with it. Alas, interfaces offer no direct access to their fields, so you really can't do it any other way than like you did.
But as Allen said, this causes a call to the getter method (GetArray), which is classified as 'virtual' without you even writing that because it's part of an interface.
Thus, every access results in a VMT-lookup (indirected via the interface) and a method call.
Also, the fact that you're using a dynamic array means that both the caller and the callee will do a lot of reference-counting (you can see this if you take a look at the generated assembly code).
All this is already enough reasons to explain the measured speed difference, but can indeed easily be overcome using a local variable and read the array only once. When you do that, the call to the getter (and all the ensueing reference counting) is taking place only once. Compared to the rest of the test, this 'overhead' becomes unmeasurable.
But note, that once you go this route, you'll loose encapsulation and any change to the contents of the array will NOT reflect back into the interface, as arrays have copy-on-write behaviour. Just a warning.
Patrick and Allen's answers are both perfectly correct.
However, since your question talks about improved OO design, I feel a particular change in your design that would also improve performance is appropriate to discuss.
Your code to set the Tag is "very controlling". What I mean by this is that you're spending a lot of time "poking around inside another object" (via an interface) in order to calculate your Tag value. This is actually what exposed the "performance problem with interfaces".
Yes, you can simply deference the interface once to a local variable, and get a massive improvement in performance, but you'll still be poking around inside another object. One of the important goals in OO design is to not poke around where you don't belong. This actually violates the Law of Demeter.
Consider the following change which empowers the interface to do more work.
IMyInterface = interface
['{C0070757-2376-4A5B-AA4D-CA7EB058501A}']
function GetArray: TArrayOfRecord;
function GetTagValue: Integer; //<-- Add and implement this
property Arr: TArrayOfRecord read GetArray;
end;
function TMyObject.GetTagValue: Integer;
var
I: Integer;
begin
for i := 0 to High(FArr) do
if (FArr[i].Val1 < FArr[i].Val2) or
(FArr[i].Val3 < FArr[i].Val4) then
begin
Result := FArr[i].Val1 + FArr[i].Val2 -
FArr[i].Val3 + FArr[i].Val4;
end;
end;
Then inside TForm3.FormCreate, //APPROACH 3 becomes:
Tag := MyInterface.GetTagValue;
This will be as fast as Allen's suggestion, and will be a better design.
Yes, I'm fully aware you simply whipped up a quick example to illustrate the performance overhead of repeatedly looking something up via interface. But the point is that if you have code performing sub-optimally because of excessive accesses via interfaces - then you have a code smell that suggests you should consider moving the responsibility for certain work into a different class. In your example TForm3 was highly inappropriate considering everything required for the calculation belonged to TMyObject.
your design use huge memory. Optimize your interface.
IMyInterface = interface
['{C0070757-2376-4A5B-AA4D-CA7EB058501A}']
function GetCount:Integer:
function GetRecord(const Index:Integer):TRecord;
property Record[Index:Integer]:TRecord read GetRecord;
end;
Related
I have types declared as:
TPlayer = class(TObject)
TBullet = class(TObject)
TEnemy = class(TObject)
and objects:
Player: TPlayer;
PlayerBullets: Array[1..20] of TBullet;
Enemies: Array[1..20] of TEnemy;
EnemyBullets: Array[1..20] of TBullet;
Now I want to create TBullet constructor, which can process informations from both Player and Enemies. In short, I want this constructor to handle both TPlayer and TEnemy objects.
My idea is:
constructor TBullet.Create(const Source: TObject);
Sadly it does not work. How to do this?
EDIT:
My exact problem is: when I pass TPlayer or TEnemy object to this constructor it doesn't see atributes of those objects. For example: TPlayer has attr xPos. If I use Bullet.Create(Player) and in TBullet.Create I use Source.xPos I get an error.
I can think of 3 ways to achieve that.
Have TPlayer and TEnemy both derive from the same base class that have all the information TBullet's constructor need, and have the constructor parameter of that type.
Define an interface contains all the information needed by TBullet and have TPlayer and TEnemy implement that interface
Leave everything "as is" and manage the different class in a "hard coded" manner in TBullet's constructor.
By that I mean:
constructor TBullet.Create(const Source: TObject);
var
vPlayer : TPlayer;
vEnemy : TEnemy;
begin
if Source is TPlayer then
begin
vPlayer := TPlayer(Source);
[Do whatever with vPlayer]
end else if Source is TEnemy then
begin
vEnemy := TEnemy(Source);
[Do Whatever with vEnemy]
end;
end;
Which solution is the best? That could be a debate in itself and largely dependant on your specific situation. Based solely on the name of your class, I'd guess option 1 could be valid. A "TCharacter" class could be created and use as a base calss for both TCharacter and TEnemy. But this is mere speculation at this point.
In windows there is little difference in inheriting from a parent with the common methods (could be abstract in the parent ) or implementing an interface (when again the behavior could be customized ).
If you have an enumeration commonParticipant( cpPlayer, cpEnemy) then Windows allows access to the ultimate parent IUnknown and then down again to a child interface that identifies the methods peculiar to that child, i.e. you can pass an ICommonParticipant including the commonParticipant and then work with either a iPlayer or IEnemy interface
I'm trying to define class methods all inside of the class declaration in Free Pascal, which I haven't been able to find any examples of online. Currently I have to do it like so:
unit Characters;
{$mode objfpc}{$H+}
// Public
interface
type
TCharacter = class(TOBject)
private
FHealth, FAttack, FDefence: Integer;
procedure SetHealth(newValue: Integer);
public
constructor Create(); virtual;
procedure SayShrek();
function GetHealth(): Integer;
published
property Health: Integer read GetHealth write SetHealth;
end;
// Private
implementation
constructor TCharacter.Create;
begin
WriteLn('Ogres have LAYERS!');
end;
procedure TCharacter.SayShrek;
begin
WriteLn('Shrek!');
end;
procedure TCharacter.SetHealth(newValue: Integer);
begin
FHealth:= FHealth + newValue;
end;
function TCharacter.GetHealth() : Integer;
begin
GetHealth:= FHealth;
end;
end.
Is there any possible way to make this a little cleaner? Defining everything elsewhere looks messy and is unorganized.
EDIT:
To clarify, I'd like to do something along the lines of this:
TMyClass = class(TObject)
public
procedure SayHi();
begin
WriteLn('Hello!');
end;
end;
Instead of having to define it further down. Is that possible?
That is not possible in Pascal. It is just not allowed by its grammar.
It is a fundamental design in Pascal that units are divided in interface (What can be done?) and implementation (How is something done?).
The compiler reads all interface sections before parsing the implementation parts. You might know this from C language. implementation could be described as *.c-files, whereas interface is equivalent to *.h-files in C.
Furthermore such code would heavily decrease readability of interface sections (f.i. class declaratons).
What benefits do you hope to get with that?
No, you can not do this. Pascal has a single-pass compiler from the outset was designed for the single-pass compilation so you can not use something before it will be declared.
As a simple example in pseudocode:
MyClass = class
procedure MethodA;
begin
MethodB; <== At this point the compiler knows nothing about MethodB
end;
procedure MethodB;
begin
end;
end;
It is why each unit have at least two sections: interface (declarations, you can think about it as about C++ header files) and implementation.
However there are some tricks in the language syntax for implementing cyclic declarations where you can use forward declarations.
For the pointers:
PMyRec = ^TMyRec; // Here is TMyRec is not declared yet but compiler can to handle this
TMyRec = record
NextItem: PMyRec;
end;
For the classes:
MyClassA = class; // Forward declaration, class will be fully declared later
MyClassB = class
SomeField: MyClassA;
end;
MyClassA = class
AnotherField: MyClassB;
end;
In the IDE you can use Shift+Ctrl+Up/Down keys to navigate between declaration and implementation of the item.
So i've been going trough types at my new work and they pretty much all look like this
create or replace TYPE BODY T_Some_Type AS
CONSTRUCTOR FUNCTION T_Some_Type
RETURN SELF AS RESULT AS
BEGIN
INTIALIZE();
RETURN;
END T_Some_Type;
MEMBER PROCEDURE INTIALIZE AS
BEGIN
var1 := 0;
var2 := 0;
var3 := 0;
END INTIALIZE;
END;
Being skilled in OOP but new to pl/sql, i keep wondering why use extra procedure to initialize variables, when it can be done directly in constructor making objects interface simpler and lighter on memory.
This is how i would normally do it :
create or replace TYPE BODY T_Some_Type AS
CONSTRUCTOR FUNCTION T_Some_Type
RETURN SELF AS RESULT AS
BEGIN
var1 := 0;
var2 := 0;
var3 := 0;
RETURN;
END T_Some_Type;
END;
Is there any advantage or this is recommended for some reason?
Please advise.
Please bear in mind that the object features in Oracle have evolved slowly since version 8.0, and yet are still pretty limited in some respects. This is because Oracle is a relational database with a structured and procedural programming paradigm: object-orientation is not a good fit.
So. In a language like Java we can override a method in a sub-type but execute the code in the parent's implementation with a super() call.
Before 11g Oracle had no similar mechanism for extending member functions. If we wanted to override a super-type's method we had to duplicate the code in the sub-type, which was pretty naff. There was a workaround which is not much less naff: extract the common code into a separate method in the super-type, and call that method in the super- type and its dependents.
In 11g Oracle introduced "Generalized Invocation". This allows us to call super-type code on a sub-type. Find out more. There is one major limitation with Generalized Invocation, and that is that we cannot use it with Constructor methods. Therefore, in its sub-types' constructors too our only choice is to put that code in a method like the initialize() in your question.
In my simulation, I want to have RW access to signals whereever there are in the project. To get the write access, I use the "signal_force" procedure from the modelsim_lib library. But to get the read access I havn't find the corresponding function.
The reason why signal_force fit my needs is that I'm working with input text files, so I have the name and the value of the signal from a "string" or a "line" variable and I can directly give these variable to the fonction.
I cannot use the "init_signal_spy" procedure because this procedure doesn't give back a value into a string but just duplicates the behavior of a signal onto an other. As my project has to be as generic as possible, I work with variables declared into procedures and I cannot link a signal onto a variable.
Thanks for your help
edited
Sorry, I win the "did not read very carefully" award for the day...
Just for completeness, I'm leaving the part of my answer that deals with signal spy (which is a proprietary ModelSim method), even though you said it wouldn't work for you:
library modelsim_lib;
use modelsim_lib.util.all;
architecture ...
signal local_sig ...
begin
process
begin
init_signal_spy("/sim/path/to/signal/internal_sig", "local_sig");
With VHDL-2008 (if you have support for it), the standard way to access signals not in scope is hierarchical/external names, and as a bonus, it does both "write" and "read". I may be a bit rusty on the nuances, but you access them like:
<<signal .sim.path.to.signal.internal_sig : std_logic>>
And you should be able to use that in place of any normal in-scope identifier, I believe. Aliases, assignments, etc.
If you're comfortable writing C code it should be straightforward to achieve what you want using the VHPI, although sadly despite being part of the VHDL standard Mentor are not planning to implement it. However it will also be possible using FLI although you're locked into a proprietary interface.
Something like this:
procedure get_signal_value_as_string(
vhdl_path : IN string;
vhdl_value: OUT string);
attribute FOREIGN of get_signal_value_as_string : procedure is “my_func mylib.so”;
procedure get_signal_value_as_string(
vhdl_path : IN string;
vhdl_value: OUT string) is
begin
report “ERROR: foreign subprogram get_signal_value_as_string not called”;
end;
Then in C:
#include <stdio.h>
#include "mti.h"
/* Convert a VHDL String array into a NULL terminated string */
static char *get_string(mtiVariableIdT id)
{
static char buf[1000];
mtiTypeIdT type;
int len;
mti_GetArrayVarValue(id, buf);
type = mti_GetVarType(id);
len = mti_TickLength(type);
buf[len] = 0;
return buf;
}
void my_func (
mtiVariableIdT vhdl_path /* IN string */
mtiVariableIdT vhdl_value /* OUT string */
)
{
mtiSignalIdT sigID = mti_FindSignal(get_string(vhdl_path));
mtiInt32T value = mti_GetSignalValue(sigID);
...
}
Plenty of example code in the FLI manual.
Here is code:
procedure DisableContrlOL(const cArray : array of string; ReEnable : boolean = False);
// can be called from VKP / RAW / Generation clicks
var
AComponent: TComponent;
CompListDis, CompListEna : TStringList;
begin
CompListDis := TStringList.Create;
CompListEna := TStringList.Create;
for i := Low(cArray) to High(cArray) do begin
AComponent := FindComponent(cArray[i]);
if Assigned(AComponent) then
if (AComponent is TControl) then begin
if TControl(AComponent).Enabled then
CompListEna.Add(TControl(AComponent).Name)
else
CompListDis.Add(TControl(AComponent).Name);
ShowMessage(TControl(AComponent).Name);
if ReEnable then begin // if reenabling needed, then all whi
if not TControl(AComponent).Enabled then
TControl(AComponent).Enabled := True;
end else if (TControl(AComponent).Enabled) then
TControl(AComponent).Enabled := False;
end;
end;
end;
I think no more explanations are needed.
The ShowMessage correctly shows name of each component, but nothing is added in StringLists. Why?
UPDATE: As question has gone pretty wild, I did confirm answer, which a bit helped me.
I understand that I did write things pretty unclear, but I am very limited, because these code lines is part of commercial project, and my hobby and heart thing. The main problem was found already 6h ago, but Rob just wanted to extend this whole question :D No, no offense, mate, it's OK. I am happy to receive so willing and helpful posts. Thanks again.
How do you know that nothing is added to the lists? You create them in this code and the only references to them are in local variables. The objects are leaked when this function returns, so you never actually use the lists anywhere.
You've said you have code for "modular testing." Since that code isn't here, I must assume the code is not part of this function. But if you have external code that's supposed to check the contents of the lists, then the lists can't be just local variables. No other code can access them. You need to either return those lists or accept lists from outside that you then fill. Here's an example of the latter:
procedure DisableContrlOL(const cArray: array of string;
Reenable: Boolean
CompListDis, CompListEna: TStrings);
// can be called from VKP / RAW / Generation clicks
var
AComponent: TComponent;
AControl: TControl;
i: Integer;
begin
for i := Low(cArray) to High(cArray) do begin
AComponent := FindComponent(cArray[i]);
if not Assigned(AComponent) or not (AComponent is TControl) then
continue;
AControl := TControl(AComponent);
if AControl.Enabled then
CompListEna.Add(AControl.Name)
else
CompListDis.Add(AControl.Name);
ShowMessage(AControl.Name);
AControl.Enabled := Reenable;
end;
end;
The caller of this function will need to provide a TStrings descendant for each list. They could be TStringList, or they could be other descendants, such as TMemo.Lines, so you can directly observe their contents in your program. (They can't be just TStrings, though, since that's an abstract class.)
As you can see, I made some other changes to your code. All your code using the Reenable parameter can be simplified to a single statement. That's because enabling a control that's already enabled, and disabling a control that's already disabled, are no-ops.
Also, Name is a public property of TComponent. You don't need to type-cast to TControl before reading that property, but since you're type-casting so often elsewhere, it made sense to introduce a new variable to hold the type-casted TControl value, and that can make your code easier to read. Easier-to-read code is easier-to-understand code, and that makes it easier to debug.
Emphasizing that this is largely based on Rob's excellent suggestions, it looks as though you could simplify the code to:
procedure DisableContrlOL(const cArray : array of string;
ReEnable : boolean = False);
var
AComponent: TComponent;
begin
for i := Low(cArray) to High(cArray) do
begin
AComponent := FindComponent(cArray[i]);
if Assigned(AComponent) then
if (AComponent is TControl) then
begin
ShowMessage(TControl(AComponent).Name);
TControl(AComponent).Enabled := ReEnable;
end;
end;
end;
Not clear what the stringlists were for, since their contents were lost when execution left the scope of this procedure. If you want to return them, you should create and free them in the calling code.
That sure looks like it ought to work. This is the sort of thing that the debugger can probably help with more than we can here.
Try breaking the problematic line down into multiple lines, like so:
if TControl(AComponent).Enabled then
CompListEna.Add(TControl(AComponent).Name)
else CompListDis.Add(TControl(AComponent).Name);
Rebuild with the "Use Debug DCUs" option on, and place a breakpoint on the if statement. Then use F7 to trace your way through the logic and see what's going on.