Hi Kevin, Assembler programming showed me the need for clarity to the point where even in Delphi, I comment most lines unless very very obvious. I've also found that the programming language bears little relationship to the readability or the debug-ability. I've had to help C programmers debug nested IF statements even though I've never programmed in C. Invariably, they tried to be clever in using the syntax. OOP does force you to design your code more structurally, but I've found a few dirty tricks (such as typecasting a class to expose private properties) which can led to outright dangerous code.
I like the saying: As soon as you idiot proof something, someone builds a better idiot... Regards Peter kevin wrote: > ----- Original Message ----- > From: "Peter" <[EMAIL PROTECTED]> > To: "Borland's Delphi Discussion List" <[email protected]> > Sent: Monday, April 24, 2006 6:46 PM > Subject: Re: Delphi and Java > > > >> Hi, >> Pauli seems to be heading the same way I am, locking all things into >> classes. With regards to global variables, I've started to create a >> configuration class for my apps. This has the added benefit that I can >> save the program configuration with changes to a config file in one >> clean location. >> > > It's me (Kevin) that's going class locking mad: I even have a class instance > called GlobalVars > containing read only properties (so that the variables' values aren't > modifiable: > so that they behave like global constants: XX := GlobalVars.Const_XX ; > etc.) > > >> Being an assembler programmer, pointers don't bother me too much, but I >> must admit I prefer dynamic arrays in Delphi, it just makes the code >> clearer and it doesn't appear to have much impact on performance. Having >> said that, I have a few GIS apps where I use pointers heavily as class >> overheads are a bit too much and dynamic arrays are a bit messy. Even >> there, I've started exploring ways of using arrays within arrays of >> > records. > > Mostly I don't use very large objectlists (I can see that a million pointers > occupy > a lot less memory than a million class instances: I suppose you could have > classes of say a thousand > pointers in an array so that the class overhead is reduced by a thousand): > in C# even an integer > is a class which implies all sorts of overheads (though there are also > advantages); > what interests me is finding a method for clearly defining what's going on: > as you say the > Delphi (and even more that used in C#) method is very clear; there is also > the advantage > that code completion helps in that the class variable names are suggested to > you instead > of your having to go and look for constants wherever they have been > declared; > You would not believe how incoherent some projects can get if global > variables are not kept in order (some I have seen have been declared in > units declared > in the uses clause of some other unit, but not declared in the project, so > that the original declaration > of the variable does not even turn up in the project file search! This is > possible in delphi but not in C#). > > Delphi is fine in that it allows you to use OOP or not as you wish, but > unfortunately this allows for sloppy practices > within or without an OO structure: C# or Java more or less impose OO: if you > want to be sloppy after that, > it is through sheer bloody mindedness! > > Kevin > > _______________________________________________ > Delphi mailing list -> [email protected] > http://www.elists.org/mailman/listinfo/delphi > > > > _______________________________________________ Delphi mailing list -> [email protected] http://www.elists.org/mailman/listinfo/delphi
