Comparing SQL and Prolog

Question

I\'ve started learning Prolog and wondering about the theoretical difference from the SQL language.

For example:

• both are declarative languages
• b

Answer 1

Prolog and SQL are both based on first order logic, but SQL tables are simple binary relations, whereas Prolog predicates are Horn clauses.

This is not some obscure theoretical point. SQL binary relations are statements of fact, of the form:

f(A, B, C ... N)

Where f is the name of the relation and A...N its variables. Prolog binary relations are implications, of the form:

A <- B, C, D ... N

Where A ... N are themselves Horn clauses. SQL relations are an efficient way to describe data. Prolog relations describe complex relationships between data, themselves stored as data.

It is important to understand that in Prolog, there is no separation between data and operation. Prolog facts, rules and queries are all Horn clauses, therefore data, something that gets lost in translation in most university courses. Prolog is not like C but with facts instead of variables and rules instead of functions. On the other hand, SQL is a lot like Prolog without the rules or queries.

SQL queries are also logic predicates, but SQL queries are not stored in the database itself. Rather, they are used to extract data sets from the database of facts. You could store a query as a table row in a SQL database but you couldn't execute it in that form.

Prolog queries are stored in the database like any other Prolog predicate, because they are like any other Prolog predicate. Queries are Horn clauses of the form:

<- B, C, D ... N

So implications with a precedent but no antecedent, therefore always false. Facts are Horn clauses with an antecedent but no precedent, of the form:

A <-

So always true. Prolog proves a query by refutation: if it can't find a fact (or rule) that proves it, it will state the goal is true, since a query is always false. In the process, some variables are bound so result sets can be constructed, much like SQL does with SELECT queries.

Modern SQL DBMSs have features like stored procedures and flow control language so SQL can be used for inference (not that you want to do inference in SQL). Prolog comes ready with an inference engine tuned to its database of Horn clauses, because that's an efficient way to do inference over databases of facts stated as binary relations (and no, not just because it's pretty).

Prolog's homoiconic nature (data is operation is data) means that new new data has to be added to the database, therefore to the program, because the database is the program. So everytime a new fact is added to its database (typically using assert/1) the whole program must be decompiled. This is a huge PITA and makes Prolog inefficient for storing large data sets, though there is no reason why it has to be inefficient at data retrieval and Prolog systems will use the same algorithms as SQL systems for that purpose. SQL on the other hand is well suited for both storage and retrieval.

Finally, Prolog has a few features that SQL just doesn't have, namely its super-pattern-matching called unification, negation as failure and syntactic elements that facilitate list processing and grammar declaration (Definite Clause Grammar notation). Those are just a happy accident and were mostly added to the language because they were en vogue in the time it was first created (thanks to LISP). SQL got recursive queries relatively recently so Prolog can't boast that anymore.

And of course both languages are weak at I/O and maths, although at least you can do some arithmetic in Prolog without having to pull your hair out all the way.

So, really, Prolog and SQL are as much alike as C and Haskell. They're both based on the same root abstraction, first order logic (like C and Haskell are both based on algebra) but things get very different pretty quickly after that. Also, from the point of view of language design, SQL tends to be fractured, with many different language features (pedicates, queries, data manipulation language....). Prolog's design is extremely consistent, so that the whole language is really just predicates and a few punctuation marks.

For me, the most important difference is this: I don't like SQL but I have to work with it. I love Prolog, but I can't use it at work. Life is unfair :)