Relational Algebra & Calculus Zachary G. Ives University of Pennsylvania CIS 550 – Database & Information Systems September 15, 2009 Some slide content courtesy of Susan Davidson & Raghu Ramakrishnan

Codd’s Relational Algebra A set of mathematical operators that compose, modify, and combine tuples within different relations Relational algebra operations operate on relations and produce relations (“closure”) f: Relation  Relation f: Relation x Relation  Relation

Codd’s Logical Operations: The Relational Algebra Six basic operations: Projection  (R) Selection  (R) Union R1 [ R2 Difference R1 – R2 Product R1 £ R2 (Rename) b (R) And some other useful ones: Join R1 ⋈ R2 Semijoin R1 ⋉ R2 Intersection R1 Å R2 Division R1 ¥ R2

Data Instance for Operator Examples STUDENT Takes COURSE sid name 1 Jill 2 Qun 3 Nitin 4 Marty sid exp-grade cid 1 A 550-0109 520-1009 3 C 501-0109 4 cid subj sem 550-0109 DB F09 520-1009 AI S09 501-0109 Arch PROFESSOR Teaches fid name 1 Ives 2 Taskar 8 Martin fid cid 1 550-0109 2 520-1009 8 501-0109

Last Time… We discussed: Projection, (R), specified a set of attributes to include in a new relation

Selection, 

Product X

Join, ⋈: A Combination of Product and Selection

Union 

Difference –

Rename, ab The rename operator can be expressed several ways: The book has a very odd definition that’s not algebraic An alternate definition: ab(x) Takes the relation with schema  Returns a relation with the attribute list  Rename isn’t all that useful, except if you join a relation with itself Why would it be useful here?

Mini-Quiz This completes the basic operations of the relational algebra. We shall soon find out in what sense this is an adequate set of operations. Try writing queries for these: The names of students named “Bob” The names of students expecting an “A” The names of students in Milo Martin’s 501 class The sids and names of students not enrolled

Deriving Intersection Intersection: as with set operations, derivable from difference A Å B ≡ (A [ B) – (A – B) – (B – A) ≡ A – (A – B) A-B B-A A B

The Big Picture: SQL to Algebra to Query Plan to Web Page Web Server / UI / etc STUDENT Takes COURSE Merge Hash by cid Query Plan – an operator tree Execution Engine Optimizer Storage Subsystem SELECT * FROM STUDENT, Takes, COURSE WHERE STUDENT.sid = Takes.sID AND Takes.cID = cid

Hint of Future Things: Optimization Is Based on Algebraic Equivalences Relational algebra has laws of commutativity, associativity, etc. that imply certain expressions are equivalent in semantics They may be different in cost of evaluation! c Ç d(R) ´ c(R) [ d(R) c (R1 £ R2) ´ R1 ⋈c R2 c Ç d (R) ´ c (d (R)) Query optimization finds the most efficient representation to evaluate (or one that’s not bad)

Switching Gears: An Equivalent, But Very Different, Formalism Codd invented a relational calculus that he proved was equivalent in expressiveness Based on a subset of first-order logic – declarative, without an implicit order of evaluation Tuple relational calculus Domain relational calculus More convenient for describing certain things, and for certain kinds of manipulations The database uses the relational algebra internally But query languages (e.g., SQL) are mostly based on the relational calculus

Domain Relational Calculus Queries have form: {<x1,x2, …, xn>| p} Predicate: boolean expression over x1,x2, …, xn Precise operations depend on the domain and query language – may include special functions, etc. Assume the following at minimum: <xi,xj,…>  R X op Y X op const const op X where op is , , , , ,  xi,xj,… are domain variables domain variables predicate

More Complex Predicates Starting with these atomic predicates, build up new predicates by the following rules: Logical connectives: If p and q are predicates, then so are p  q, p  q, p, and p  q (x>2)  (x<4) (x>2)  (x>0) Existential quantification: If p is a predicate, then so is x.p x. (x>2) (x<4) Universal quantification: If p is a predicate, then so is x.p x.x>2 x. y.y>x

Some Examples Faculty ids Subjects for courses with students expecting a “C” All course numbers for which there exists a smaller course number

Logical Equivalences There are two logical equivalences that will be heavily used: p  q  p  q (Whenever p is true, q must also be true.) x. p(x)  x. p(x) (p is true for all x) The second can be a lot easier to check! Example: The highest course number offered

Free and Bound Variables A variable v is bound in a predicate p when p is of the form v… or v… A variable occurs free in p if it occurs in a position where it is not bound by an enclosing  or  Examples: x is free in x > 2 x is bound in x. x > y

Can Rename Bound Variables Only When a variable is bound one can replace it with some other variable without altering the meaning of the expression, providing there are no name clashes Example: x. x > 2 is equivalent to y. y > 2 Otherwise, the variable is defined outside our “scope”…

Safety Pitfall in what we have done so far – how do we interpret: {<sid,name>| <sid,name>  STUDENT} Set of all binary tuples that are not students: an infinite set (and unsafe query) A query is safe if no matter how we instantiate the relations, it always produces a finite answer Domain independent: answer is the same regardless of the domain in which it is evaluated Unfortunately, both this definition of safety and domain independence are semantic conditions, and are undecidable

Safety and Termination Guarantees There are syntactic conditions that are used to guarantee “safe” formulas The definition is complicated, and we won’t discuss it; you can find it in Ullman’s Principles of Database and Knowledge-Base Systems The formulas that are expressible in real query languages based on relational calculus are all “safe” Many DB languages include additional features, like recursion, that must be restricted in certain ways to guarantee termination and consistent answers

Mini-Quiz How do you write: Which students have taken more than one course from the same professor?

Translating from RA to DRC Core of relational algebra: , , , x, - We need to work our way through the structure of an RA expression, translating each possible form. Let TR[e] be the translation of RA expression e into DRC. Relation names: For the RA expression R, the DRC expression is {<x1,x2, …, xn>| <x1,x2, …, xn>  R}

Selection: TR[ R] Suppose we have (e’), where e’ is another RA expression that translates as: TR[e’]= {<x1,x2, …, xn>| p} Then the translation of c(e’) is {<x1,x2, …, xn>| p’} where ’ is obtained from  by replacing each attribute with the corresponding variable Example: TR[#1=#2 #4>2.5R] (if R has arity 4) is {<x1,x2, x3, x4>| < x1,x2, x3, x4>  R  x1=x2  x4>2.5}

Projection: TR[i1,…,im(e)] If TR[e]= {<x1,x2, …, xn>| p} then TR[i1,i2,…,im(e)]= {<x i1,x i2, …, x im >|  xj1,xj2, …, xjk.p}, where xj1,xj2, …, xjk are variables in x1,x2, …, xn that are not in x i1,x i2, …, x im Example: With R as before, #1,#3 (R)={<x1,x3>| x2,x4. <x1,x2, x3,x4> R}

Union: TR[R1  R2] R1 and R2 must have the same arity For e1  e2, where e1, e2 are algebra expressions TR[e1]={<x1,…,xn>|p} and TR[e2]={<y1,…yn>|q} Relabel the variables in the second: TR[e2]={< x1,…,xn>|q’} This may involve relabeling bound variables in q to avoid clashes TR[e1e2]={<x1,…,xn>|pq’}. Example: TR[R1  R2] = {< x1,x2, x3,x4>| <x1,x2, x3,x4>R1  <x1,x2, x3,x4>R2

Other Binary Operators Difference: The same conditions hold as for union If TR[e1]={<x1,…,xn>|p} and TR[e2]={< x1,…,xn>|q} Then TR[e1- e2]= {<x1,…,xn>|pq} Product: If TR[e1]={<x1,…,xn>|p} and TR[e2]={< y1,…,ym>|q} Then TR[e1 e2]= {<x1,…,xn, y1,…,ym >| pq} Example: TR[RS]= {<x1,…,xn, y1,…,ym >| <x1,…,xn> R  <y1,…,ym > S }

What about the Tuple Relational Calculus? We’ve been looking at the Domain Relational Calculus The Tuple Relational Calculus is nearly the same, but variables are at the level of a tuple, not an attribute {Q | 9 S  COURSES, 9 T 2 Takes (S.cid = T.cid Æ Q.cid = S.cid Æ Q.exp-grade = T.exp-grade)}

Limitations of the Relational Algebra / Calculus Can’t do: Aggregate operations Recursive queries Complex (non-tabular) structures Most of these are expressible in SQL, OQL, XQuery – using other special operators Sometimes we even need the power of a Turing-complete programming language

Summary Can translate relational algebra into relational calculus DRC and TRC are slightly different syntaxes but equivalent Given syntactic restrictions that guarantee safety of DRC query, can translate back to relational algebra These are the principles behind initial development of relational databases SQL is close to calculus; query plan is close to algebra Great example of theory leading to practice!