Thread: About using global vars

Originally Posted by esbo

Are you serious? "The name, documentation, and general project design should make it clear what each variable does. ".
The variable does what it does, the documention etc is what is 'supposed' to do.

What if there is no documentation?

Most variable are not documented.
The only thing which will tell you what a function does is the code.
There may be errors in the design at all levels.

Most variables are not documented because you can tell what they do given their name and the function they are in. If that isn't enough, then a comment is required. If the variable is part of the external API of the component, then more documentation is needed.

If you want to find out what something is supposed to do, you look at the documentation, the name, and the context.

If you are trying to find a bug, then you are correct that the documentation usually won't show it (unless the bug is in the design).

Originally Posted by esbo

There may not be a debugger aavailable.
What a 'trace' is is unclearly defined, there may well not be one which does what you
just described, if indeed at all of if I even understand what you mean. (unclear).

Tracing is when a debugger steps through your code line by line as your program runs. It is an extremely common feature.

Certainly there are environments where you cannot use a debugger, and you are forced to examine code manually. Even then it is often most useful to follow code flow, rather than what variable is changing.

This benefit of semi-isolating when a variable changes is small compared to the other benifits of local variables (mentioned in the other thread). Indeed in most cases having function local variables is itself a way of isolating when a variable changes, as then you know that the variable can only change within the function it is declared in, and the functions called by it.

Originally Posted by esbo

It is really not a problem for me at all.
It might be for everyone else but that's their problem not mine.
I can easilly write code to fit into a big project.
If there are any scoping problems they will be easy enough to sort out.
It's not something I would want to waste time on beforehand.

If it's everyone else problem and not yours then perhalps you have not worked on enough of such projects. In which case you should follow the design pattern, before you run into the problem that it solves on your own.

huh?

If you write code without formal design, then you wind up with crappy design. At least for big projects.

Originally Posted by esbo

Ever noticed how most big formally designed projects end up having to be scrapped

There are plenty of successful large software projects out there. I'd wager all of them had a formal design phase.

But what does that have to do with global variables?

Originally Posted by esbo

I can easilly write code to fit into a big project.

Define "big project".

Originally Posted by esbo

I prefer to write code rather than talking about it.

Me too.

Originally Posted by esbo

That way once it is written at least everyone knows exactly what they are talking about, rather
than some ill defined definitions.

Unfortunately, real world projects do not work that way. Even if you do not go with a full formal design up front, you still need to talk first and get the requirements right, otherwise you could all end up knowing exactly what you are talking about, except that what you are talking about is nothing like what the customer wanted.

Originally Posted by whiteflags

My attempt to explain how abstraction often makes globals unnecessary, and how they can actually break a program by introducing side effects has obviously failed entirely as opposed to just esbo's special case.

You realy can't talk in general terms about programs as they are all different.
Programs tend to have a fair bit of global data but then ever program is different.
The only thing which can break a program is the person(s) who write it.
Anyway it's all rather too vague, nobody knows what anyone else is talkin about because
we are talking about the undefined.
We can waffle on endlessly about the underfined but it is all rather pointless, unless we are
talking about something specific nothing has any real meaning.
It's all just too vague.

Originally Posted by esbo

You realy can't take in general terms about programs as they are all different.

That's rubbish. Programs are different, but they have common characteristics in their requirements, design, development and maintenance (though these may differ depending on the exact area of computing).

Originally Posted by esbo

You realy can't talk in general terms about programs as they are all different.
Programs tend to have a fair bit of global data but then ever program is different.
The only thing which can break a program is the person(s) who write it.
Anyway it's all rather too vague, nobody knows what anyone else is talkin about because
we are talking about the undefined.
We can waffle on endlessly about the underfined but it is all rather pointless, unless we are
talking about something specific nothing has any real meaning.
It's all just too vague.

This seems a shrewd way to opt out of a discussion. First you make your point, then somebody else makes theirs, then you say that the subject is to broad an unconcrete to meaningfully talk about it in the first place.

Originally Posted by robwhit

> I prefer to write code rather than talking about it.
> That way once it is written at least everyone knows exactly what they are talking about, rather
than some ill defined definitions.

> We can waffle on endlessly about the underfined but it is all rather pointless, unless we are
talking about something specific nothing has any real meaning.
> It's all just too vague.

You claim to have worked on projects before, and not using local variables is your style. Could you post or otherwise show a non-trivial example of a project that uses the techniques you are describing?

Not really.
I don't want my code stolen
What would it prove if I did?
I have a prog with over 100 global and about 30 local.
If I posted what would it prove?
I could post the code and the data it works on to prove that it works, then what are
you going to do?
Tell me it does not work????
Tell me I won't be able to modify or add to it?
Well I could post all the previous versions of it with less functionality which also work.

So what exactly are you going to say is bad about it????

Or maybe say it is only 1200 lines long and that I would have probllems if it was a million lines long?
The reason it is not a milion line long is because it is well written