Board index » cppbuilder » Re: Deriving from TCustomListView

2003-09-19 03:05:45 AM
cppbuilder70
"Remy Lebeau (TeamB)" < XXXX@XXXXX.COM >wrote:
the constuctos/destructors are void. But the rest of the lib could do
astonishing extras, or not?
wrong, as it is not so important: including from hpp is faster and does not
cause that kind of problems.
the objector inspector 'reads back' the values?
those properties which are set by the constructor?
originally it was part of the component like project.
It is not in all-capital, but also the original was. It worked with BCB6,
but specified creating the component.
A final question: how can you reload the component after modifications? I
worked out a way ( Components | Install Package to remove the entire bpl,
then Components | Add Component to reinstall it), but doubt it is the
best/correct way.
Rgds, Martin
 

"Remy Lebeau (TeamB)" < XXXX@XXXXX.COM >wrote in message
Ralph
 

"Martin Moeller" < XXXX@XXXXX.COM >wrote in message
something that is not declared first. The HPP header files are just
IDE-generated C++ header files for the VCL's Pascal code, and as such as
(almost) an exact mirror of that code. If you do not see something declared
in the HPP files, then they most likely are not declared in the Pascal code
either. In order for a class to do something extra on top of its base
class, whether the class is written in Pascal or C++, the class's
declaration in the header must include declarations for the new
functionality, whether that be new members, overridden menbers, etc. As you
can clearly see, TListView does not declare anything new which
TCustomListView does not already have. And its generated
constructors/destructor are empty (which, btw, they don't even exist in the
Pascal code at all, the C++ IDE auto-generated them only for the sake of the
HPP code being compliant with the C++ language).
Besides, Boland distributes the VCL source code anyway, so I can tell you
with 100% certainty that TListView does not do anything extra at all.
you? I notice that you were trying to publish the Canvas and Column[]
(singular) properties before. You cannot publish those, as 1) they are both
read-only and you can only published read/write properties, and 2) you
cannot publish properties that use array notation. They both have to remain
public instead of published. And in fact they are both already declared as
public in TCustomListView, so there is no need to promote them anyway.
They have been constructed normally and they reside in memory normally and
they are executing as if you were running outside of the IDE normally. So
yes, the Object Inspector is merely displaying the current values from an
actual running instance of the component. The only difference is that the
VCL has support for detecting the presence of the IDE so that is can
internally control how certain component actions behave, such as mouse
clicks and such. In all other regards, the component instance is real and
its coding is being executed as it would be outside of the IDE.
reassign a value that is already assigned.
components for years and it works fine from BCB3 all the way of to BCB6
without problem.
Some will say that it is possible to use DCR files. Some will say that they
have done it successfully. I believe them. However, I have never had a
single DCR file work correctly for me, which is why I always recommend RES
instead. Besides, DCR is a Delphi file anyway. For a C++ component, RES is
just easier to work with anyway, in my opinion.
start using the new library files immediately.
Gambit
 

{smallsort}

"Remy Lebeau (TeamB)" < XXXX@XXXXX.COM >wrote:
Now I have the res file, well named (in caps, same name as component) in the
same directory as the cpp.
It could be addded once but showed not the icon, so I deleted it from the
package window and checked.
Now I can not add it again, it opens an error window with following text:
' the project already contains a form or module named MLSVVREP1. '
Pls help, thanks, Martin
 

"Martin Moeller" < XXXX@XXXXX.COM >wrote:
- The bmp within the resource file MUST be all caps and have the same name
as the component
- The res file does not have to be all caps and MUST NOT have the same name
as cpp file.
That allows loading and now displays fine. As a matter of fact, I have the
same name for component and its cpp file.
Thanks, Martin
 

"Martin Moeller" < XXXX@XXXXX.COM >wrote in message
sounds like you named the file. You are supposed to be naming the actual
bitmap instead, not the RES file itself. That is another of the beauties I
like about using RES files - the file can be named anything you want.
Whereas with DCR, the file has to be specifically named after the component
it belongs to.
Gambit
 

"Remy Lebeau (TeamB)" < XXXX@XXXXX.COM >wrote in message
than the component. As long as the bitmap is named after the component (all
caps) it'll work.
Ralph
 

"Ralph Kazemier" < XXXX@XXXXX.COM >wrote:
must be named differently res and cpp, and res does not have to be all-caps.
The naming in my project, as asked by Remy, is very simple: same name all
over (component class & component source file, bitmap all-caps and res
file - which does not work). Also tried removing both files from package;
adding res file first is ok, but get the same error on adding cpp file
later.
Anyhow, what should the strategy be (question) :
- very limited number of packages, for component types
- single res file for bitmap or several bitmaps in one res file (a- is it
possible, b- does it make sense)
?
Tks & Regards, Martin
 

"Martin Moeller" < XXXX@XXXXX.COM >wrote in message
1) Open Image Editor
2) Select 'File|New...|Component Resource File (.dcr)'
3) Right click 'Contents' in the treeview
4) Select 'New|Bitmap'
5) Set Width and Height to 24. Set Colors to 256.
6) Rename the new bitmap entry to the name of the component in all caps
7) Draw the bitmap
Repeat steps 3-7 for each component you're installing
8) Select 'File|Save'
9) Typically the .dcr file is saved in the same directory
where the package resides
10) Open the package in BCB
11) Open the main .cpp file
You'll probably see something like;
//----...
#include <vcl.h>
#pragma hdrstop
USERES("myPackage.res");
12) Add the your resource file, so that it looks something like;
//----...
#include <vcl.h>
#pragma hdrstop
USERES("myPackage.res");
USERES("myResource.dcr");
13) Build & Install the package
Ralph
 

"Remy Lebeau (TeamB)" < XXXX@XXXXX.COM >wrote
Suppose that what I should have noticed is that TListView declares no new
members nor methods,
from TListView and TCustomListView. If at all, deriving from a Custom class
to promote all first is better done pasting from the header file rather than
from the help anyhow.
is anyhow assigned by the compiler creating a class instance (with or
without beeing a component), thus no assignement is needed while it should
still be specified as { default } in order to avoid unnecessary saving of
that value in the resource file. If I want a different value, I need to
write the assignement statement into the constructor (typically), in which
case I can/should as well specify the {default}. Values without {default} or
set to a different value than what specified, will all be saved to projects
res file.
With the aim to adress specifically a virtual TListView in Report mode, I
was considering the opportunity to enhance the TListColumn to include more
references to the data displayed and their coordinates.
But then there is a general uncertity I have: it is quite easy to derive a
TListColumn adding my wanted features, but then the pointer in the
TCustomListView is still of type original TListColumn. Therefore, without
rewriting the whole column part, he will try to allocate standard
TListColums and crash as my class has a different size. Thus no way to
override TListColumn maintaining the rest of the original methods. Or is
there another trick ?
And considering my very limited BCB skills, would it be possible at all (for
me) ?
Finally, I learn that components are compiled to obj when inserted and
linked from there. How about the debugg information, will he somehow
automatically strip them as per the general project setting ( and recompile
the component if necessary), or will I have to do this manually loading the
non-debugg version once it is stable enough.
And as they are pre-compiled, there can be no template components as well as
any template component members. What I'm thinking of is to at least
partially automate the data flow between my virtual report viewer and the
std::container holding the data. It's quite vague still, but somehow I'm
looking for the possibility to set the pointer/refrence of the container
into the ListView and call some standard interface functions.
Can I hardcode an event handler? Not publishing it, but beeing called to
execute my fixed build in code?
Tks & Regds, Martin
 

"Martin Moeller" < XXXX@XXXXX.COM >wrote in message
the same and they work fine :-p.
What you're referring to is only the IDE trying to side-step you. Although
the IDE doesn't like files being named the same, the compiler and linker are
perfectly fine with it. I just modify the project's main files manually to
get around the IDE's complaining. Works fine.
Perticularly useful when you have multiple components in the same package.
I don't thing DCR supports that, every component would need its own
individual DCR file instead.
Gambit
 

"Ralph Kazemier" < XXXX@XXXXX.COM >wrote in message
What you describe is what the "Add to Project" option of the IDE already
does.
Gambit
 

"Martin Moeller" < XXXX@XXXXX.COM >wrote in message
zeros out its own memory during its construction. That is not the case for
non-TObject classes. The memory for those classes is left uninitialized, it
is your own responsibility to initialize your class members yourself
appropriately.
No, not typically - always. That is the only place default values can
safely by assigned. That is what constructors are meant for.
You have no control over it.
TListColumns collection before your own code has any chance to run.
Your only option is to reimplement the entire Columns mechanism, either as
an override to the existing Columns property, or as a secondary property.
TCustomComboBoxEx has to do something like that in BCB6 for its new ItemsEx
property, for example.
packages in both Release and Debug modes. The compiling mode of a package
is independant of the compiling of the application that is using the
package. You can have a Debug package in a Release application, and vice
versa. Of course, to actually step through and debug the package, it has to
be compiled in Debug mode, its TDS file has to be present on the IDE's
search path (since Borland project do not compile the debug info directly
into the code, they use a separate TDS file instead), and the application
usually has to be compiled in Debug mode as well.
the user's code from assign its own handlers. What you should be doing
instead is overriding the inherited methods that are responsible for
triggering the events in the first place. For example, to handle the OnData
event, override the OwnerDataFetch() method instead.
publishing it manually with another descendant class if they really want
their own access.
Gambit
 

"Remy Lebeau (TeamB)" < XXXX@XXXXX.COM >wrote in message
accept .c, .cpp, .obj, .asm, .rc, .res, .pas, .lib, .dcu or .def files.
Ralph
 

"Remy Lebeau (TeamB)" < XXXX@XXXXX.COM >wrote:
and whenever a restart is required.
part) to override the columns property by a MyColumns, but then that must be
handling all what the Columns did.
Otherwise, given that MyColumns derives from Columns and adds both a few
variables and a few functions, I just can not immagine how that could step
into the shorter allocate space. Also, provided I allocated entirely by
myself the MyColumns array (or list), would it then be sufficient for the
rest of the component to declare a MyColumns function calling the
Columns::function, ( plus what is needed for the added performance of
course) ?
And, sorry for a standard C++ question, will I have only to bother about the
protected/public members and methods such as listed in the help as per
object oriented programming logic?
No problem with secondary property (suppose you mean a second array/list
complementing those data of the Columns); it's just that such a solution
wouldn't be too smart.
OnDataFetch() generates OnData event, from the documentation it is unknown
when it is called and what it does with the item passed. Therefore, having
to replace it and without calling TCustomListView::OnDataFetch() to avoid
the OnData event, I can override it but how can I know what to put into it?
Rgds, Martin
P.S. will be away for the rest of the week - don't cancel the thread pls.