You write that a global variable has a “better” scope. This is incorrect. It has a bigger scope. Bigger is not better.
Bigger may be necessary, if you need an identifier to be visible in more places, but this is often not the case. But a bigger scope means more exposure to errors. Global variables muddle the semantics of routines by making it harder to see what program state they use and change, and it increases the probability of errors caused by failing to declare a local identifier and of other errors.
In particular, an identifier with external linkage will collide with identifiers in other libraries. Consider what happens when you are writing a physics application, have an external identifier named acceleration
, and link with a physics library that also has an external identifier named acceleration
. The program will fail. Because of this, external identifiers are usually bad design.
A significant limit on our ability to develop and maintain complex software is human error. Much of programming language semantics limits the language to prevent errors. With a raw computer, you can add two pointers, trash your stack pointer, accidentally load the bytes of a float
into an integer register, and so on. Good programming languages make these errors difficult to do by mistake.
Global variables were a larger source of errors before scoping rules helped control them. Good programmers limit the scopes of their identifiers.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…