Global Variables in Prolog

It is not the skilful poet that bemoans the rigid rules of the sonnet.

When you start learning Prolog, the language will initially seem very limited to you. Where are all the features you know from other programming languages? In particular, where are the global variables that you can use in other languages to pass information around?

In Prolog, we pass information around in predicate arguments. We are not using global variables.

Why is this? Think about it: Suppose there were a global state that is modified by a predicate. This would have consequences that clash fundamentally with how we expect relations to behave. For example, when programming in terms of relations, we expect: Therefore, we avoid the use of global state in Prolog. Instead, we describe relations between states of interests. When you are stuck, consider adding an argument to your predicate that lets you express the state of interest. See Thinking in States for examples.

That being said, there are of course ways to modify the global state of your programs. However, using them should be the exception, not the norm. For example, you can modify the global database using assertz/1. The global database is very good for reading information, but quite bad for often modifying the data. This is because the global database performs indexing over its entries, and you pay the price for indexing every time you assert new clauses. In addition, you cannot use this kind of modifications in the other direction. On backtracking, the data will still be there, and running the same code twice may also add the data redundantly.

Using predicate arguments to pass around information therefore more naturally leads to Prolog programs that are pure and efficient. Semicontext notation lets you pass around information without the need to introduce many additional predicate arguments.

More about Prolog

Main page