Re: [Lazarus] Extending TRect breaks Lazarus

[Marco van de Voort]

On Wed, Nov 11, 2015 at 07:53:33PM +0100, Marco van de Voort wrote:
> > I replaced many endangered with-blocks in the LCL.
> > ToDo: check the rest of the Lazarus sources.
> 
> Ok, I plan finalizing and committing the trect stuff in the coming week.

The first batch is committed. Tpoint* is reasonably complete but TRect 
and -F need some more methods.

The problem is also a bit that TRect is more complicated and my sources span
various Delphi versions and are sometimes conflicting.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Marco van de Voort]

On Wed, Nov 11, 2015 at 10:37:48PM +0100, Bart wrote:
> On 11/11/15, Marco van de Voort  wrote:
> 
> > Ok, I plan finalizing and committing the trect stuff in the coming week.
> 
> Should these be merged to 1.4 fixes branch?

If you want to run 1.4 with trunk yes.

It won't be in 3.0.0, and even 3.0.0 fixes branch is to be decided when
the impact is clearer.

So for now FPC trunk only, and see how much dust it throws up.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Mattias Gaertner]

On Sat, 7 Nov 2015 00:21:09 +0100
Marco van de Voort  wrote:

>[...]
> Anyway, I hunted down one case by raise tobject.create, and running in the
> debugger, and I committed a fix. The Win32 lazarus now seems to function
> normally at first glance.

I replaced many endangered with-blocks in the LCL.
ToDo: check the rest of the Lazarus sources.

Mattias

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Anthony Walter]

My solution is to define my own TRectI type which can implicitly convert
between TRectI <-> TRect. Also I use a TRectF which does the same thing,
but with floats as the dimenions ... MyRectInt := TRectI(MyRectFloat); ...
There is also TPointI and TPointF ...

{ TRectI }

  TRectI = record
  private
function GetEmpty: Boolean;
procedure SetTop(Value: Integer);
procedure SetLeft(Value: Integer);
function GetRight: Integer;
procedure SetRight(Value: Integer);
function GetBottom: Integer;
procedure SetBottom(Value: Integer);
function GetSize: TPointI;
function GetTopLeft: TPointI;
function GetBottomLeft: TPointI;
function GetBottomRight: TPointI;
function GetTopRight: TPointI;
function GetMidPoint: TPointI;
  public
X, Y, Width, Height: Integer;
class operator Implicit(const Value: TRectI): TRect;
class operator Implicit(const Value: TRect): TRectI;
class operator Equal(const A, B: TRectI): Boolean;
class operator NotEqual(const A, B: TRectI): Boolean;
class function Create: TRectI; overload; static;
class function Create(Size: TPointI): TRectI; overload; static;
class function Create(W, H: Integer): TRectI; overload; static;
class function Create(X, Y, W, H: Integer): TRectI; overload; static;
function Equals(const Value: TRectI): Boolean;
function Contains(X, Y: Integer): Boolean; overload;
function Contains(const P: TPointI): Boolean; overload;
procedure Center(X, Y: Integer); overload;
procedure Center(const P: TPointI); overload;
procedure Inflate(X, Y: Integer); overload;
procedure Inflate(const P: TPointI); overload;
procedure Offset(X, Y: Integer); overload;
procedure Offset(const P: TPointI); overload;
property Empty: Boolean read GetEmpty;
property Left: Integer read X write SetLeft;
property Top: Integer read Y write SetTop;
property Right: Integer read GetRight write SetRight;
property Bottom: Integer read GetBottom write SetBottom;
property Size: TPointI read GetSize;
property TopLeft: TPointI read GetTopLeft;
property BottomLeft: TPointI read GetBottomLeft;
property BottomRight: TPointI read GetBottomRight;
property TopRight: TPointI read GetTopRight;
property MidPoint: TPointI read GetMidPoint;
  end;
  PRectI = ^TRectI;
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Marco van de Voort]

On Wed, Nov 11, 2015 at 04:39:55PM +0100, Mattias Gaertner wrote:
> >[...]
> > Anyway, I hunted down one case by raise tobject.create, and running in the
> > debugger, and I committed a fix. The Win32 lazarus now seems to function
> > normally at first glance.
> 
> I replaced many endangered with-blocks in the LCL.
> ToDo: check the rest of the Lazarus sources.

Ok, I plan finalizing and committing the trect stuff in the coming week.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Bart]

On 11/11/15, Marco van de Voort  wrote:

> Ok, I plan finalizing and committing the trect stuff in the coming week.

Should these be merged to 1.4 fixes branch?

Bart

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Juha Manninen]

On Wednesday, November 11, 2015, Bart  wrote:
>
> Should these be merged to 1.4 fixes branch?
>

No, Lazarus 1.4 dot versions are released with FPC 2.6.4. The TRect changes
will happen in FPC trunk only.

Juha
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Sven Barth]

Am 10.11.2015 09:40 schrieb "Jürgen Hestermann" :
>
> Am 2015-11-09 um 14:26 schrieb Sven Barth:
>>
>>
>> > The docs already note that the compiler uses a temporary register
>> > for 'With'.
>> > http://www.freepascal.org/docs-html/ref/refsu58.html#x155-16500013.2.8
>> > Of course if there is no register left the compiler uses the stack.
>>
>> Please note that this is an implementation detail. The point a user
should care about is that the expression of the with is evaluated only once
(before the block of the with is entered). Everything else is up to the
compiler.
>> (I would prefer if it wouldn't be mentioned as is in the documentation)
>>
>>
> This information is very important IMO.
> The programmer should know which coding has which performance.

Which is solved by saying that the expression is only evaluated once. The
user doesn't need to know the detail that the value is kept in a temporary
register, because that statement by itself is utterly useless for the user,
as on load/store architectures like ARM you /always/ have values in
registers if they are worked with and the compiler will happily spill
around the contents if necessary.

Regards,
Sven
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Jürgen Hestermann]

Am 2015-11-09 um 14:26 schrieb Sven Barth:


> The docs already note that the compiler uses a temporary register
> for 'With'.
> http://www.freepascal.org/docs-html/ref/refsu58.html#x155-16500013.2.8
> Of course if there is no register left the compiler uses the stack.

Please note that this is an implementation detail. The point a user should care 
about is that the expression of the with is evaluated only once (before the 
block of the with is entered). Everything else is up to the compiler.
(I would prefer if it wouldn't be mentioned as is in the documentation)



This information is very important IMO.
The programmer should know which coding has which performance.

The reason for many slow programs is that
the programmer(s) do not know about such details.
If i.e. a loop over a huge array can be speeded up with
a slight coding change  it should be done.
The attitude: "don't worry about the details" is wrong IMO.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Mattias Gaertner]

On Tue, 10 Nov 2015 12:26:56 +0100
Jürgen Hestermann  wrote:

>[...]
> Does that mean, that local variables and with statements are treated 
> completely identically?

A local var can be changed. Without optimization the compiler puts the
local var on the stack. With optimization enabled the compiler can
find out that the local var is only read, which is also true for a 
with-expression and applies the same optimizations.

Optimizing read only variables is basic code optimization.

> Are both held in registers with the same mechanism and will there never be an 
> (speed) adavantage of one of them?
> If there is only the chance that it may speed up the code (even when only on 
> certain platforms) then I would like to know it.

If speed matters you should enable optimization. When optimization is
enabled there is no difference.

Mattias

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Jürgen Hestermann]

Am 2015-11-10 um 11:32 schrieb Sven Barth:

The user doesn't need to know the detail that the value is kept in a temporary 
register, because that statement by itself is utterly useless for the user, as 
on load/store architectures like ARM you /always/ have values in registers if 
they are worked with and the compiler will happily spill around the contents if 
necessary.


Does that mean, that local variables and with statements are treated completely 
identically?
Are both held in registers with the same mechanism and will there never be an 
(speed) adavantage of one of them?
If there is only the chance that it may speed up the code (even when only on 
certain platforms) then I would like to know it.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Sven Barth]

Am 10.11.2015 12:27 schrieb "Jürgen Hestermann" :
>
> Am 2015-11-10 um 11:32 schrieb Sven Barth:
>>
>> The user doesn't need to know the detail that the value is kept in a
temporary register, because that statement by itself is utterly useless for
the user, as on load/store architectures like ARM you /always/ have values
in registers if they are worked with and the compiler will happily spill
around the contents if necessary.
>
>
> Does that mean, that local variables and with statements are treated
completely identically?
> Are both held in registers with the same mechanism and will there never
be an (speed) adavantage of one of them?
> If there is only the chance that it may speed up the code (even when only
on certain platforms) then I would like to know it.

It always depends on the specific code (its complexity) and the specific
compiler settings (optimizations) and maybe also the target architecture
(because of register pressure and such). There is nothing general I can say
/except/ that the with expression will be evaluated only once. Thus
depending on the circumstances a local variable and the usage of with might
behave the same. In simple cases the with /might/ be faster because it
won't result in the expression result being stored on the stack, but that's
by no means a given.

Regards,
Sven
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Sven Barth]

Am 10.11.2015 14:37 schrieb "Sven Barth" :
>
> Am 10.11.2015 12:27 schrieb "Jürgen Hestermann" :
> >
> > Am 2015-11-10 um 11:32 schrieb Sven Barth:
> >>
> >> The user doesn't need to know the detail that the value is kept in a
temporary register, because that statement by itself is utterly useless for
the user, as on load/store architectures like ARM you /always/ have values
in registers if they are worked with and the compiler will happily spill
around the contents if necessary.
> >
> >
> > Does that mean, that local variables and with statements are treated
completely identically?
> > Are both held in registers with the same mechanism and will there never
be an (speed) adavantage of one of them?
> > If there is only the chance that it may speed up the code (even when
only on certain platforms) then I would like to know it.
>
> It always depends on the specific code (its complexity) and the specific
compiler settings (optimizations) and maybe also the target architecture
(because of register pressure and such). There is nothing general I can say
/except/ that the with expression will be evaluated only once. Thus
depending on the circumstances a local variable and the usage of with might
behave the same. In simple cases the with /might/ be faster because it
won't result in the expression result being stored on the stack, but that's
by no means a given.

So... I quickly rechecked the code and can now make a more fundated
statement. In case of a more complex expression (e.g. not a field that's a
class or something like that) the expression will be stored in a temporary
/variable/ not a register. Thus the generated code should be the same as
for a local variable. For non complex expressions the value is used
directly (e.g. as if the with didn't happen), though further optimizations
(e.g. the regvars optimization) might improve that.

Regards,
Sven
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Sven Barth]

Am 09.11.2015 13:11 schrieb "Mattias Gaertner" :
>
> On Mon, 9 Nov 2015 10:48:02 +
> Lukasz Sokol  wrote:
>
> >[...]
> > Thanks, I believe, this should be added into the docs somewhere ;) if
it's not already there.
>
> The docs already note that the compiler uses a temporary register
> for 'With'.
>
> http://www.freepascal.org/docs-html/ref/refsu58.html#x155-16500013.2.8
>
> Of course if there is no register left the compiler uses the stack.

Please note that this is an implementation detail. The point a user should
care about is that the expression of the with is evaluated only once
(before the block of the with is entered). Everything else is up to the
compiler.
(I would prefer if it wouldn't be mentioned as is in the documentation)

Regards,
Sven
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Lukasz Sokol]

On 07/11/15 11:01, Mattias Gaertner wrote:
> On Sat, 7 Nov 2015 11:39:44 +0100
> Jürgen Hestermann  wrote:
> 
[...]
>> -
>> with PathArray[High(PathArray)]^ do
>> fillchar(StatisticOfFiles,sizeof(StatisticOfFiles),0);
>> -
>>
>> instead of this:
>>
>> -
>> fillchar(PathArray[High(PathArray)]^.StatisticOfFiles,sizeof(PathArray[High(PathArray)]^.StatisticOfFiles),0);
>> -
[...]
> 
> LastPath:=PathArray[High(PathArray)];
> fillchar(LastPath^.StatisticOfFiles,sizeof(LastPath^.StatisticOfFiles),0);
> 

Will this form produce same code (as in memory footprint and machine 
code/so-called performance) ?

(but then, the last form, will have the variable declared explicitly, which 
probably means more complicated code around this region etc.?)

> Use Ctrl+Shift+C on LastPath to create the pointer variable.
> 
> Mattias
> 
el es



--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Lukasz Sokol]

On 09/11/15 10:19, Mattias Gaertner wrote:

> The above three are not the same.
> 
> The second version executes the With-Expression multiple times, so
> unless the compiler optimizes a lot it will create more code and will be
> slower.
> 
> The first and third versions have the same amount of reads/writes and
> with optimizations (-O2 or higher) create the same assembler code.
> Without optimizations the first is a bit faster, because the compiler
> stores the With-Pointer in a register, while in third version it stores
> it on the stack.
> 
>  
>> (but then, the last form, will have the variable declared explicitly, which 
>> probably means more complicated code around this region etc.?)
> 
> A local variable is bread and butter for the compiler. There is
> nothing complicated about it.
> I would argue that giving an expression a describing name can make code
> less complicated.
> 
> Mattias
> 

Thanks, I believe, this should be added into the docs somewhere ;) if it's not 
already there.

el es


--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Mattias Gaertner]

On Mon, 9 Nov 2015 09:38:27 +
Lukasz Sokol  wrote:

>[...] On 07/11/15 11:01, Mattias Gaertner wrote:
> > On Sat, 7 Nov 2015 11:39:44 +0100
> > Jürgen Hestermann  wrote:
> > 
> [...]
> >> -
> >> with PathArray[High(PathArray)]^ do
> >> fillchar(StatisticOfFiles,sizeof(StatisticOfFiles),0);
> >> -
> >>
> >> instead of this:
> >>
> >> -
> >> fillchar(PathArray[High(PathArray)]^.StatisticOfFiles,sizeof(PathArray[High(PathArray)]^.StatisticOfFiles),0);
> >> -
> [...]
> > 
> > LastPath:=PathArray[High(PathArray)];
> > fillchar(LastPath^.StatisticOfFiles,sizeof(LastPath^.StatisticOfFiles),0);
> > 
> 
> Will this form produce same code (as in memory footprint and machine 
> code/so-called performance) ?

The above three are not the same.

The second version executes the With-Expression multiple times, so
unless the compiler optimizes a lot it will create more code and will be
slower.

The first and third versions have the same amount of reads/writes and
with optimizations (-O2 or higher) create the same assembler code.
Without optimizations the first is a bit faster, because the compiler
stores the With-Pointer in a register, while in third version it stores
it on the stack.

 
> (but then, the last form, will have the variable declared explicitly, which 
> probably means more complicated code around this region etc.?)

A local variable is bread and butter for the compiler. There is
nothing complicated about it.
I would argue that giving an expression a describing name can make code
less complicated.

Mattias

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Mattias Gaertner]

On Mon, 9 Nov 2015 10:48:02 +
Lukasz Sokol  wrote:

>[...]
> Thanks, I believe, this should be added into the docs somewhere ;) if it's 
> not already there.

The docs already note that the compiler uses a temporary register
for 'With'.

http://www.freepascal.org/docs-html/ref/refsu58.html#x155-16500013.2.8

Of course if there is no register left the compiler uses the stack.

Mattias

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Florian Klaempfl]

Am 07.11.2015 um 06:40 schrieb Alexander Klenin:
> On Sat, Nov 7, 2015 at 5:58 AM, Florian Klaempfl  
> wrote:
>> Am 06.11.2015 um 20:22 schrieb Luca Olivetti:
>>> El 06/11/15 a les 18:10, Martin Schreiber ha escrit:
>>>
 You are joking, no? It has been discussed on fpc-pascal/fpc-devel several
 times. An IIRC Italian community member then usually shows a syntax
 description of a safe "with" statement from a maybe more than 20 year old
 pascal dialect.
> Here is one "recent" discussion:
> http://free-pascal-lazarus.989080.n3.nabble.com/Lazarus-with-is-evil-isn-t-it-td3219144.html
> 
>> It is similiar to the proc. var code which is also one of the places
>> only very brave people look at ;)
> One nice thing about FPC/Lazarus is that I may disappear for a few years,
> then look back and see same topics discussed by the same people with
> the same conclusions ;-)
>

Indeed, unfortunatly nothing changed: bug count increases continously
and the fixing is done mainly by the long term comitters but people are
still proposing to extend the language with more sugar increasing
maintainance work.


--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Florian Klaempfl]

Am 07.11.2015 um 06:42 schrieb Ondrej Pokorny:
> 
> Do not forget that you can overload operators! In particular you can
> overload the comparison operator to return a record, so your
> "with a=somerecord do" can already be a valid pascal code now:
> 
> *type**
> **  TMyRec = record**
> **I: Integer;**
> **  end;**
> **
> **  TMyResult = record**
> **Res: Integer;**
> **  end;**
> **
> **operator = (z1: TMyRec; z2 : TMyRec) b : TMyResult;**
> **begin**
> **  Result.Res := z1.I+z2.I;**
> **end;**
> **
> **var**
> **  m1, m2: TMyRec;**
> **begin**
> **  m1.I := 10;**
> **  m2.I := 1;**
> **  with m1=m2 do**
> **writeln(Res);**
> **  readln;**
> **end.*
> 
> The assignment operator would be better, since it doesn't return a value
> (in pascal, of course) and thus cannot be used in a with statement now:

Good that we removed another "great" language extennsion already years
ago, the support for a C-like assignment operator which returns a result
:)))

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Jürgen Hestermann]

Am 2015-11-06 um 18:26 schrieb Ondrej Pokorny:
> So yes, let's not use with :)
> ... There is no problem with long variable names at all. There are CodeTools 
:)

CodeTools do not help you *read* the code.

I prefer to have this:

-
with PathArray[High(PathArray)]^ do
   fillchar(StatisticOfFiles,sizeof(StatisticOfFiles),0);
-

instead of this:

-
fillchar(PathArray[High(PathArray)]^.StatisticOfFiles,sizeof(PathArray[High(PathArray)]^.StatisticOfFiles),0);
-

Also, the problem of the WITH statement is the same as for identifiers declared 
in other units.
Do you also preceed variables from a used unit with the unit name?



--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Mattias Gaertner]

On Sat, 7 Nov 2015 11:39:44 +0100
Jürgen Hestermann  wrote:

>[...]
> CodeTools do not help you *read* the code.

It can help you writing the code more readable and helps to read (i.e.
understand) the code by hints and find declaration.

 
>[...]
> -
> with PathArray[High(PathArray)]^ do
> fillchar(StatisticOfFiles,sizeof(StatisticOfFiles),0);
> -
> 
> instead of this:
> 
> -
> fillchar(PathArray[High(PathArray)]^.StatisticOfFiles,sizeof(PathArray[High(PathArray)]^.StatisticOfFiles),0);
> -

An one liner with only one used variable is a pretty safe 'with' and
is mostly a matter of taste.

But a complicated expression like PathArray[High(PathArray)]^ can be
made more readable (easier to understand) like this:

LastPath:=PathArray[High(PathArray)];
fillchar(LastPath^.StatisticOfFiles,sizeof(LastPath^.StatisticOfFiles),0);

Use Ctrl+Shift+C on LastPath to create the pointer variable.

Mattias

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Michael Van Canneyt]

On Sat, 7 Nov 2015, Alexander Klenin wrote:


On Sat, Nov 7, 2015 at 5:58 AM, Florian Klaempfl  wrote:

Am 06.11.2015 um 20:22 schrieb Luca Olivetti:

El 06/11/15 a les 18:10, Martin Schreiber ha escrit:


You are joking, no? It has been discussed on fpc-pascal/fpc-devel several
times. An IIRC Italian community member then usually shows a syntax
description of a safe "with" statement from a maybe more than 20 year old
pascal dialect.

Here is one "recent" discussion:
http://free-pascal-lazarus.989080.n3.nabble.com/Lazarus-with-is-evil-isn-t-it-td3219144.html


It is similiar to the proc. var code which is also one of the places
only very brave people look at ;)

One nice thing about FPC/Lazarus is that I may disappear for a few years,
then look back and see same topics discussed by the same people with
the same conclusions ;-)


Well, assuming God exists and visits from time to time to look at her creation,
she probably thinks the same about humanity.

So FPC really is just a representative sample of humanity. With the same 
factions.

Michael.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Marco van de Voort]

On Fri, Nov 06, 2015 at 07:25:01PM +0100, Mattias Gaertner wrote:
> > width and height of the trect are assigned instead of the component's ?
> 
> Thanks for the hint.
> 
> Does Delphi alter some other records too, e.g. TPoint and TSize?
> 
> Hint:
> Cody has a tool 'Explode a "With" block':
> http://wiki.lazarus.freepascal.org/Cody#Explode_a_.22With.22_block

Would have been nice to run that on the code that I replied to Michael.
(with the local procedure)

Anyway, I hunted down one case by raise tobject.create, and running in the
debugger, and I committed a fix. The Win32 lazarus now seems to function
normally at first glance.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Sven Barth]

Am 06.11.2015 17:59 schrieb "Marco van de Voort" :
>
> On Fri, Nov 06, 2015 at 05:42:39PM +0100, Michael Van Canneyt wrote:
> > > by Delphi compatibility...
> >
> > Ehm. How can "with" ever be safe ?
>
> Btw the case that I fixed was like this
>
> procedure TSomeForm.Getwind...
>
> procedure setwidthheight(const r : TRect); inline;
> begin
>  with R do
>begin
> width:=right-left;
> height:=bottom-top;
>   end;
> end;
>
> begin
>  ...
>   setwidtheight(r);
> end
>
> In such case fixing with SELF doesn't work, because self isn't allowed in
> the nested function. I had to strip the WITH.

Self really isn't allowed in nested functions? O.o

Regards,
Sven
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Ondrej Pokorny]

On 06.11.2015 22:45, Mark Morgan Lloyd wrote:

Luca Olivetti wrote:

 with a=somerecord, b=someotherrecord do
 begin
   a.x:=b.x;
 end;


Just because something is 20+ years old doesn't necessarily make it 
bad. That syntax looks OK to me (i.e. make a and b look like consts) 
but I also find myself wondering what the type rules are, i.e.


with pod: TTabSheet= Pages.Objects[deleteAt] do

vs

with pod= TTabSheet(Pages.Objects[deleteAt]) do


Do not forget that you can overload operators! In particular you can 
overload the comparison operator to return a record, so your

"with a=somerecord do" can already be a valid pascal code now:

*type**
**  TMyRec = record**
**I: Integer;**
**  end;**
**
**  TMyResult = record**
**Res: Integer;**
**  end;**
**
**operator = (z1: TMyRec; z2 : TMyRec) b : TMyResult;**
**begin**
**  Result.Res := z1.I+z2.I;**
**end;**
**
**var**
**  m1, m2: TMyRec;**
**begin**
**  m1.I := 10;**
**  m2.I := 1;**
**  with m1=m2 do**
**writeln(Res);**
**  readln;**
**end.*

The assignment operator would be better, since it doesn't return a value 
(in pascal, of course) and thus cannot be used in a with statement now:


*with a:=somerecord, b:=someotherrecord do**
**begin**
**  a.x:=b.x;**
**end; **
*
IMO there is no need to set a type of a/b on the left side of the with 
assignment - the left side should always gain the type from the the 
right side. In this case you can decide whether the "AS" or the direct 
re-type should be used:


*with pod := Pages.Objects[deleteAt] as TTabSheet do **
**with pod := TTabSheet(Pages.Objects[deleteAt]) do **
*
-> both are valid and the syntax is already in pascal - why not to use it.

Ondrej
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Alexander Klenin]

On Sat, Nov 7, 2015 at 5:58 AM, Florian Klaempfl  wrote:
> Am 06.11.2015 um 20:22 schrieb Luca Olivetti:
>> El 06/11/15 a les 18:10, Martin Schreiber ha escrit:
>>
>>> You are joking, no? It has been discussed on fpc-pascal/fpc-devel several
>>> times. An IIRC Italian community member then usually shows a syntax
>>> description of a safe "with" statement from a maybe more than 20 year old
>>> pascal dialect.
Here is one "recent" discussion:
http://free-pascal-lazarus.989080.n3.nabble.com/Lazarus-with-is-evil-isn-t-it-td3219144.html

> It is similiar to the proc. var code which is also one of the places
> only very brave people look at ;)
One nice thing about FPC/Lazarus is that I may disappear for a few years,
then look back and see same topics discussed by the same people with
the same conclusions ;-)

-- 
Alexander S. Klenin

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Luca Olivetti]

El 06/11/15 a les 18:10, Martin Schreiber ha escrit:


You are joking, no? It has been discussed on fpc-pascal/fpc-devel several
times. An IIRC Italian community member then usually shows a syntax
description of a safe "with" statement from a maybe more than 20 year old
pascal dialect.


I don't know if that's me, but, yes, I used a variant of pascal (by 
texas instruments) that allowed to declare aliases with "with", e.g


 with a=somerecord, b=someotherrecord do
 begin
   a.x:=b.x;
 end;


I even found the (scanned) reference manual to confirm the syntax but 
now I cannot find it.

It is definitely more than 20 years old (and more than 30 ;-):

http://www.computinghistory.org.uk/det/11554/TI-990-System/


Bye
--
Luca Olivetti
Wetron Automation Technology http://www.wetron.es/
Tel. +34 93 5883004 (Ext.3010)  Fax +34 93 5883007

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Dmitry Boyarintsev]

On Fri, Nov 6, 2015 at 1:25 PM, Mattias Gaertner 
wrote:

>
> Does Delphi alter some other records too, e.g. TPoint and TSize?
>
> It does
http://docwiki.embarcadero.com/Libraries/Seattle/en/System.Types.TPoint
http://docwiki.embarcadero.com/Libraries/Seattle/en/System.Types.TSize

so it's just a matter of time when FPC RTL replicates them.

thanks,
Dmitry
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Martin Schreiber]

On Friday 06 November 2015 18:19:14 Michael Van Canneyt wrote:
>
> But presumably you mean some construct as
>
> With QMyVeryLongQueyObjectName as Q do
>try
>  // Do things with Q
>finally
>  Q.Close;
>end;
>
> ?
>
Correct. There have been some more notations suggested.

Martin

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Graeme Geldenhuys]

On 2015-11-06 17:19, Michael Van Canneyt wrote:
> With QMyVeryLongQueyObjectName as Q do
>try
>  // Do things with Q
>finally
>  Q.Close;
>end;

var
  Q: TQuery absolute QMyVeryLongQueyObjectName;
begin
  with Q do
try
  // Do things with Q
finally
  Q.Close;
end;


problem solved! ;-)

And because Q is such a short name, I wouldn't even both with the WITH
statement above.  It would make for easier debugging too. eg: debug
tooltips evaluation etc.


Regards,
  - Graeme -

-- 
fpGUI Toolkit - a cross-platform GUI toolkit using Free Pascal
http://fpgui.sourceforge.net/

My public PGP key:  http://tinyurl.com/graeme-pgp

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Michael Van Canneyt]

On Fri, 6 Nov 2015, Martin Schreiber wrote:


On Friday 06 November 2015 17:42:39 Michael Van Canneyt wrote:

On Fri, 6 Nov 2015, Martin Schreiber wrote:

On Friday 06 November 2015 17:20:24 Ondrej Pokorny wrote:

Nevertheless I am of the opinion that WITH is evil and should be removed
from the IDE/LCL code (unless only identifiers from within the with
scope are used; but even then they could be removed...).


Or implement a safe "with" in Free Pascal without waiting until it is
enforced by Delphi compatibility...


Ehm. How can "with" ever be safe ?


You are joking, no? It has been discussed on fpc-pascal/fpc-devel several
times.


Erm... I am not joking.

Do you really think I recall every discussion or braindead proposal from 
these mailing lists over the last 20 years ?


I appreciate your high esteem of my memory, but alas, age takes it's toll... ;-)

But presumably you mean some construct as

With QMyVeryLongQueyObjectName as Q do
  try
// Do things with Q
  finally
Q.Close;
  end;

?

That is a modification of the current "with". 
I assumed you meant to make an unchanged "with" safe.


That said, a construct such as "With QMyVeryLongQueyObjectName as Q do" has my 
support...

Michael.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Ondrej Pokorny]

On 06.11.2015 17:42, Michael Van Canneyt wrote:

Ehm. How can "with" ever be safe ?


You can make sure that if some property is ADDED into the with scope, it 
doesn't break anything by explicitely using Self (or the unit name or 
whatever):


procedure TForm1.DoMySize;
begin
  //...
  with MyRect do
Self.Width := Right - Left;
end;

But you cannot do anything if a property is REMOVED (e.g. Right in this 
case)...


So yes, let's not use with :)

+ I don't use with in my code any more. I confess that in the beginning 
I did use it, but decided to ignore the with statement at all. There is 
no problem with long variable names at all. There are CodeTools :)


Ondrej

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Michael Van Canneyt]

On Fri, 6 Nov 2015, Martin Schreiber wrote:


On Friday 06 November 2015 18:33:54 Graeme Geldenhuys wrote:

On 2015-11-06 17:19, Michael Van Canneyt wrote:

With QMyVeryLongQueyObjectName as Q do
   try
 // Do things with Q
   finally
 Q.Close;
   end;


var
  Q: TQuery absolute QMyVeryLongQueyObjectName;
begin
  with Q do
try
  // Do things with Q
finally
  Q.Close;
end;


problem solved! ;-)


Not a good idea IMHO, that hinders readability because one doesn't see the
real container at the "with" block start. Another important use of "with" is
to make complex address calculations only once at the "with" block start
without the need to define a pointer variable in "var" and to dereference in
every statement.


Indeed, and besides this:

a) it does not compile for properties. See below.
b) I did mention the requirement about non-trivial readable variable names...

Michael.

program Project1;

{$mode objfpc}

Type
  TMyType = class(TObject)
  private
FM: TMyType;
function GetM: TMyType;
procedure SetM(AValue: TMyType);
  Public
Procedure DoIt;
Property MyVeryLongObjectPropertyName : TMyType Read GetM Write SetM;
  end;

function TMyType.GetM: TMyType;
begin
  Result:=Nil;
end;

procedure TMyType.SetM(AValue: TMyType);
begin
  //
end;

procedure TMyType.DoIt;

Var
  Q : TMyType absolute MyVeryLongObjectPropertyName; // Error !!!

begin

end;

begin
end.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Mattias Gaertner]

On Fri, 6 Nov 2015 16:45:07 +0100
Marco van de Voort  wrote:

>[...]
> I yesterday tried to implement some of the TRect helper methods in 3.1.1
>[...]
> I haven't investigated deeper yet, but I suspect the heavy use of WITH in
> interfaces etc is the cause.  (if you assign width and height of a
> component in a component method in a WITH with a TRECT as argument, then now
> width and height of the trect are assigned instead of the component's ?

Thanks for the hint.

Does Delphi alter some other records too, e.g. TPoint and TSize?

Hint:
Cody has a tool 'Explode a "With" block':
http://wiki.lazarus.freepascal.org/Cody#Explode_a_.22With.22_block

Mattias

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Bart]

@marcov: your patch for this issue in r50234 seems to have mangled
some utf8-codepoints in a comment about Bézier curves (line 2739 and
2749)?

Bart

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Martin Schreiber]

On Friday 06 November 2015 18:33:54 Graeme Geldenhuys wrote:
> On 2015-11-06 17:19, Michael Van Canneyt wrote:
> > With QMyVeryLongQueyObjectName as Q do
> >try
> >  // Do things with Q
> >finally
> >  Q.Close;
> >end;
>
> var
>   Q: TQuery absolute QMyVeryLongQueyObjectName;
> begin
>   with Q do
> try
>   // Do things with Q
> finally
>   Q.Close;
> end;
>
>
> problem solved! ;-)
>
Not a good idea IMHO, that hinders readability because one doesn't see the 
real container at the "with" block start. Another important use of "with" is 
to make complex address calculations only once at the "with" block start 
without the need to define a pointer variable in "var" and to dereference in 
every statement.

Martin

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Florian Klaempfl]

Am 06.11.2015 um 20:22 schrieb Luca Olivetti:
> El 06/11/15 a les 18:10, Martin Schreiber ha escrit:
> 
>> You are joking, no? It has been discussed on fpc-pascal/fpc-devel several
>> times. An IIRC Italian community member then usually shows a syntax
>> description of a safe "with" statement from a maybe more than 20 year old
>> pascal dialect.
> 
> I don't know if that's me, but, yes, I used a variant of pascal (by
> texas instruments) that allowed to declare aliases with "with", e.g
> 
>  with a=somerecord, b=someotherrecord do
>  begin
>a.x:=b.x;
>  end;
> 

While I think this is something were a patch would be accepted, my main
issue is that the with handling code in the compiler is somehow messy.
Not necessarily due to bad coding but simply the way how "with" works.
It is similiar to the proc. var code which is also one of the places
only very brave people look at ;)


--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

[Lazarus] Extending TRect breaks Lazarus

[Marco van de Voort]

FYI:

I yesterday tried to implement some of the TRect helper methods in 3.1.1 and
property and Lazarus(/win32) refused to start due to WMSIZE loops.

I haven't investigated deeper yet, but I suspect the heavy use of WITH in
interfaces etc is the cause.  (if you assign width and height of a
component in a component method in a WITH with a TRECT as argument, then now
width and height of the trect are assigned instead of the component's ?

Probably method calling (setwidth/setheight?) is dangerous too.

Request for TRect and patch that I used : 
http://bugs.freepascal.org/view.php?id=21041


--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Dmitry Boyarintsev]

On Fri, Nov 6, 2015 at 10:45 AM, Marco van de Voort  wrote:

> I haven't investigated deeper yet, but I suspect the heavy use of WITH in
> interfaces etc is the cause.  (if you assign width and height of a
> component in a component method in a WITH with a TRECT as argument, then
> now
> width and height of the trect are assigned instead of the component's ?
>

This is odd. How adding  the method to TRect didn't break much for Delphi?
(or maybe it did, we just don't know it).
I'd think that Delphi /3d party components code also uses a lot of WITHs
with TRects

thanks,
Dmitry
--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Ondrej Pokorny]

On 06.11.2015 17:08, Dmitry Boyarintsev wrote:

This is odd. How adding  the method to TRect didn't break much for Delphi?


It did break a lot of code for Delphi.

I'd think that Delphi /3d party components code also uses a lot of 
WITHs with TRects.


Yes, it did break a lot of code in 3rd party components as well.

---

Nevertheless I am of the opinion that WITH is evil and should be removed 
from the IDE/LCL code (unless only identifiers from within the with 
scope are used; but even then they could be removed...). The 
Width/Height (+ other useful functions) are a good TRect enhancement.


Ondrej

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Marco van de Voort]

On Fri, Nov 06, 2015 at 11:08:29AM -0500, Dmitry Boyarintsev wrote:
> > I haven't investigated deeper yet, but I suspect the heavy use of WITH in
> > interfaces etc is the cause.  (if you assign width and height of a
> > component in a component method in a WITH with a TRECT as argument, then
> > now
> > width and height of the trect are assigned instead of the component's ?
> >
> 
> This is odd. How adding  the method to TRect didn't break much for Delphi?
> (or maybe it did, we just don't know it).

Maybe they were more careful with WITH or had problems and simply fixed
them, always used self. to qualify such cases. I don't routine look at
Delphi source, so I don't know which one.

> I'd think that Delphi /3d party components code also uses a lot of WITHs
> with TRects

Well, not anymore I guess :-) Or there is a flaw in my thinking, but if so,
then I don't see it. I'm currently building fpc with raise tobject.create in
the relevant methods to see if I can quickly identify some locations. 

Yup:

a#0  0x77055b68 in RaiseException () from C:\WINDOWS\SysWOW64\KernelBase.dll
#1  0x00416039 in fpc_raiseexception (OBJ=0xe11f3c8, ANADDR=0x0,
# AFRAME=0xe11f730) at seh32.inc:93
#2  0x00417cb6 in RECT__SETWIDTH (AVALUE=236058416, this=...) at
# c:/repo/fpc/rtl/win/wininc/func.inc:2414
#3  0x00531d0a in TWIN32WIDGETSET__GETWINDOWSIZE (HANDLE=986038,
# WIDTH=986038, HEIGHT=5,
this=) at
C:/repo/lazarus/lcl/interfaces/win32/win32winapi.inc:2395
#4  0x004f0c0e in GETWINDOWSIZE (HANDLE=986038, WIDTH=986038, HEIGHT=5) at
# C:/repo/lazarus/lcl/include/winapi.inc:564
#5  0x0052ad58 in TWINDOWPROCHELPER__DOMSGSIZE (this=)
at C:/repo/lazarus/lcl/interfaces/win32/win32callback.inc:1903
#6  0x0052be5b in TWINDOWPROCHELPER__DOWINDOWPROC (this=)
at C:/repo/lazarus/lcl/interfaces/win32/win32callback.inc:2358
#7  0x0052c767 in WINDOWPROC (WINDOW=986038, MSG=5, WPARAM=0, LPARAM=0)

Maybe for the 3rd party support it is smart to tag them with experimental or
so till this code is really release (as 3.2), so that people at least get
an hint that something is/could be wrong.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Martin Schreiber]

On Friday 06 November 2015 17:20:24 Ondrej Pokorny wrote:
>
> Nevertheless I am of the opinion that WITH is evil and should be removed
> from the IDE/LCL code (unless only identifiers from within the with
> scope are used; but even then they could be removed...).

Or implement a safe "with" in Free Pascal without waiting until it is enforced 
by Delphi compatibility...

Martin

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Graeme Geldenhuys]

On 2015-11-06 16:20, Ondrej Pokorny wrote:
> Nevertheless I am of the opinion that WITH is evil 

+1

Especially if you mix more than one scope/variable in the with block.

Regards,
  - Graeme -


--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Michael Van Canneyt]

On Fri, 6 Nov 2015, Martin Schreiber wrote:


On Friday 06 November 2015 17:20:24 Ondrej Pokorny wrote:


Nevertheless I am of the opinion that WITH is evil and should be removed
from the IDE/LCL code (unless only identifiers from within the with
scope are used; but even then they could be removed...).


Or implement a safe "with" in Free Pascal without waiting until it is enforced
by Delphi compatibility...


Ehm. How can "with" ever be safe ?

Don't get me wrong:

I use it abundantly. If I have QMYQueryObjectWithAVeryLongName then being able 
to write

With QMYQueryObjectWithAVeryLongName do
  try
// Really nice code
  finally
Close;
  end;

is really good (tm).

NOT being able to use with would seriously impede on the readability of my code.

Add this to the observation that the lazarus team demands 'clear' variable 
names,
and does not like/accept i,m,f,p,q or whatever as variable names, and using "with" 
becomes very attractive indeed...


Michael.

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Marco van de Voort]

On Fri, Nov 06, 2015 at 05:42:39PM +0100, Michael Van Canneyt wrote:
> > by Delphi compatibility...
> 
> Ehm. How can "with" ever be safe ?

Btw the case that I fixed was like this

procedure TSomeForm.Getwind...

procedure setwidthheight(const r : TRect); inline;
begin
 with R do
   begin
width:=right-left;
height:=bottom-top;
  end;
end;

begin
 ... 
  setwidtheight(r);
end

In such case fixing with SELF doesn't work, because self isn't allowed in
the nested function. I had to strip the WITH.

Luckily it wasn't as bad as expected, and only this one in the win32 widget
set.

> 
> Don't get me wrong:
> 
> I use it abundantly. If I have QMYQueryObjectWithAVeryLongName then being 
> able to write

Me too, but I like to qualify fieldnames with self when the they are short,
common words.
 


> With QMYQueryObjectWithAVeryLongName do
>try
>  // Really nice code
>finally
>  Close;
>end;
> 
> is really good (tm).
> 
> NOT being able to use with would seriously impede on the readability of my 
> code.
> 
> Add this to the observation that the lazarus team demands 'clear' variable 
> names,
> and does not like/accept i,m,f,p,q or whatever as variable names, and using 
> "with" 
> becomes very attractive indeed...
> 
> Michael.
> 
> --
> ___
> Lazarus mailing list
> Lazarus@lists.lazarus.freepascal.org
> http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Flávio Etrusco]

On Fri, Nov 6, 2015 at 1:42 PM, Michael Van Canneyt
 wrote:
>
>
> On Fri, 6 Nov 2015, Martin Schreiber wrote:
>
>> On Friday 06 November 2015 17:20:24 Ondrej Pokorny wrote:
>>>
>>>
>>> Nevertheless I am of the opinion that WITH is evil and should be removed
>>> from the IDE/LCL code (unless only identifiers from within the with
>>> scope are used; but even then they could be removed...).
>>
>>
>> Or implement a safe "with" in Free Pascal without waiting until it is
>> enforced
>> by Delphi compatibility...
>
>
> Ehm. How can "with" ever be safe ?
>
> Don't get me wrong:
>
> I use it abundantly. If I have QMYQueryObjectWithAVeryLongName then being
> able to write
>
> With QMYQueryObjectWithAVeryLongName do
>   try
> // Really nice code
>   finally
> Close;
>   end;
>
> is really good (tm).
>
> NOT being able to use with would seriously impede on the readability of my
> code.
>
> Add this to the observation that the lazarus team demands 'clear' variable
> names,
> and does not like/accept i,m,f,p,q or whatever as variable names, and using
> "with" becomes very attractive indeed...
>
> Michael.

I hate "with" with passion because maintaining code written using it
(by other people ;-) ) is very often a living hell.
I, for one, always use local variables in that case, but I'd really
really love to be able to declare an alias for "with" expression, no
matter how many people say it's unpascalish...

WRT a "safe with", well, the compiler could continue the identifier
search as if it didn't find in the With expression and then generate
an error or warning...

-Flávio

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Graeme Geldenhuys]

On 2015-11-06 16:42, Michael Van Canneyt wrote:
> I use it abundantly. If I have QMYQueryObjectWithAVeryLongName then being 
> able to write

Can't you simply use an "absolute" variable, or alias variable with a
shorter name - instead of WITH all together.

I hate WITH, but do use it on occasion - primarily in fpGUI's Form
Designer (auto generated code). But if I do have to use WITH, I always
only use one variable with it.

For me at least, WITH falls into the same category as GOTO. It has its
pitfalls, can easily be abused, but useful on occasions.  ;-)

Regards,
  - Graeme -

-- 
fpGUI Toolkit - a cross-platform GUI toolkit using Free Pascal
http://fpgui.sourceforge.net/

My public PGP key:  http://tinyurl.com/graeme-pgp

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Graeme Geldenhuys]

On 2015-11-06 16:42, Michael Van Canneyt wrote:
> Ehm. How can "with" ever be safe ?

I forgot to add in my previous message... I also can't see how WITH can
ever be made safe.

Regards,
  - Graeme -

-- 
fpGUI Toolkit - a cross-platform GUI toolkit using Free Pascal
http://fpgui.sourceforge.net/

My public PGP key:  http://tinyurl.com/graeme-pgp

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Martin Schreiber]

On Friday 06 November 2015 17:42:39 Michael Van Canneyt wrote:
> On Fri, 6 Nov 2015, Martin Schreiber wrote:
> > On Friday 06 November 2015 17:20:24 Ondrej Pokorny wrote:
> >> Nevertheless I am of the opinion that WITH is evil and should be removed
> >> from the IDE/LCL code (unless only identifiers from within the with
> >> scope are used; but even then they could be removed...).
> >
> > Or implement a safe "with" in Free Pascal without waiting until it is
> > enforced by Delphi compatibility...
>
> Ehm. How can "with" ever be safe ?
>
You are joking, no? It has been discussed on fpc-pascal/fpc-devel several 
times. An IIRC Italian community member then usually shows a syntax 
description of a safe "with" statement from a maybe more than 20 year old 
pascal dialect.
But probably such language additions are not sexy enough.

Martin

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus

Re: [Lazarus] Extending TRect breaks Lazarus

[Mark Morgan Lloyd]

Luca Olivetti wrote:

El 06/11/15 a les 18:10, Martin Schreiber ha escrit:


You are joking, no? It has been discussed on fpc-pascal/fpc-devel several
times. An IIRC Italian community member then usually shows a syntax
description of a safe "with" statement from a maybe more than 20 year old
pascal dialect.


I don't know if that's me, but, yes, I used a variant of pascal (by 
texas instruments) that allowed to declare aliases with "with", e.g


 with a=somerecord, b=someotherrecord do
 begin
   a.x:=b.x;
 end;


Just because something is 20+ years old doesn't necessarily make it bad. 
That syntax looks OK to me (i.e. make a and b look like consts) but I 
also find myself wondering what the type rules are, i.e.


with pod: TTabSheet= Pages.Objects[deleteAt] do

vs

with pod= TTabSheet(Pages.Objects[deleteAt]) do

--
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk

[Opinions above are the author's, not those of his employers or colleagues]

--
___
Lazarus mailing list
Lazarus@lists.lazarus.freepascal.org
http://lists.lazarus.freepascal.org/mailman/listinfo/lazarus