## On Answering Queries in the Presence of Limited Access Patterns (2001)

Venue: | In Proc. of ICDT 2001 |

Citations: | 31 - 2 self |

### BibTeX

@INPROCEEDINGS{Li01onanswering,

author = {Chen Li and Edward Chang},

title = {On Answering Queries in the Presence of Limited Access Patterns},

booktitle = {In Proc. of ICDT 2001},

year = {2001},

pages = {219--233}

}

### Years of Citing Articles

### OpenURL

### Abstract

. In information-integration systems, source relations often have limitations on access patterns to their data; i.e., when one must provide values for certain attributes of a relation in order to retrieve its tuples. In this paper we consider the following fundamental problem: can we compute the complete answer to a query by accessing the relations with legal patterns? The complete answer to a query is the answer that we could compute if we could retrieve all the tuples from the relations. We give algorithms for solving the problem for various classes of queries, including conjunctive queries, unions of conjunctive queries, and conjunctive queries with arithmetic comparisons. We prove the problem is undecidable for datalog queries. If the complete answer to a query cannot be computed, we often need to compute its maximal answer. The second problem we study is, given two conjunctive queries on relations with limited access patterns, how to test whether the maximal answer to...

### Citations

1084 |
Principles of Database and Knowledge-Base Systems, volume I
- Ullman
- 1988
(Show Context)
Citation Context ... the solution to this problem can be used to answer queries efficiently. Given a conjunctive query on relations with limited access patterns, [8, 17] show how to construct a recursive datalog program =-=[24]-=- to compute the maximal answer to the query. That is, we can retrieve tuples from relations by retrieving as many bindings from the relations and the query as possible, then use the obtained tuples to... |

997 | Mec[~a[~rs in the Architecture of Future Information Systems
- Wiederhold
- 1992
(Show Context)
Citation Context ...grams. 1 Introduction The goal of information-integration systems (e.g., [3, 20, 25]) is to support seamless access to heterogeneous data sources. In these systems, a user poses a query on a mediator =-=[26]-=-, which computes the answer by accessing the data at the underlying source relations. One of the challenges for these systems is to deal with the diverse capabilities of sources in answering queries. ... |

466 |
Optimal implementation of conjunctive queries in relational data bases
- Chandra, Merlin
- 1977
(Show Context)
Citation Context ...s, since we cannot bind the variable S in the added subgoal. However, this subgoal is actually redundant, and we can show that Q 2 is equivalent to query Q 1 . That is, there is a containment mapping =-=[4]-=- from Q 2 to Q 1 , and vice versa. Therefore, for any database of the two relations, we can still compute the complete answer to Q 2 by answering Q 1 . Example 2 suggests that testing stability of a c... |

436 | Information integration using logical views
- Ullman
(Show Context)
Citation Context ...e assume that the contained query is a conjunctive query, while in [19] the contained query can be a recursive datalog query. Finally, [19] uses the source-centric approach to information integration =-=[25]-=-, which is different from the query-centric approach [25] that is taken in our framework. However, we can easily extend our technique [16] to the source-centric approach. [6] involves a complex algori... |

420 | Answering queries using views
- Levy, Mendelzon
- 1995
(Show Context)
Citation Context ...nomial-time algorithm for testing the boundedness of the datalog program for a connection query. 7 Related Work Several works consider binding patterns in the context of answering queries using views =-=[8, 14, 1]-=-. Rajaraman, Sagiv, and Ullman [21] propose algorithms for answering queries using views with binding patterns. In that paper all solutions to a query compute the complete answer to the query; thus on... |

249 | On the Power of Magic
- Beeri, Ramakrishnan
- 1987
(Show Context)
Citation Context ... the subgoals \Phi a . There are two cases: 1. If all the distinguished variables are bound by the subgoals \Phi a (arc 4), then the complete answer may be computed even if the supplementary relation =-=[2, 24]-=- (denoted I a ) of subgoals \Phi a is not empty. We compute the supplementary relation I a of these subgoals following a feasible order of \Phi a . (a) If I a is empty (arc 5), then we know that the c... |

222 | Using Schema Matching to Simplify Heterogeneous Data Translation
- Milo, Zohar
- 1998
(Show Context)
Citation Context ... query is contained in the maximal answer to the second one? We show this problem is decidable using the results of monadic programs. 1 Introduction The goal of information-integration systems (e.g., =-=[3, 20, 25]-=-) is to support seamless access to heterogeneous data sources. In these systems, a user poses a query on a mediator [26], which computes the answer by accessing the data at the underlying source relat... |

201 | Answering queries using templates with binding patterns
- Rajaraman, Sagiv, et al.
- 1995
(Show Context)
Citation Context ...ndedness of the datalog program for a connection query. 7 Related Work Several works consider binding patterns in the context of answering queries using views [8, 14, 1]. Rajaraman, Sagiv, and Ullman =-=[21]-=- propose algorithms for answering queries using views with binding patterns. In that paper all solutions to a query compute the complete answer to the query; thus only stable queries are handled. Dusc... |

143 |
Equivalence among relational expressions with the union and dierence operators
- Sagiv, Yannakakis
- 1980
(Show Context)
Citation Context ...Delta \Delta \Delta [Qn be a finite union of CQ's (UCQ for short), and all its CQ's have a common head predicate. It is known that there is a unique minimal subset of Q that is its minimal equivalent =-=[22]-=-. Example 3. Suppose we have three relations r, s, and p, and each relation has only one binding pattern bf . Consider the following three CQ's: Q 1 : ans(X) :- r(a; X) Q 2 : ans(X) :- r(a; X); p(Y; Z... |

140 |
On conjunctive queries containing inequalities
- Klug
- 1988
(Show Context)
Citation Context ... XsA and AsY will then force A = X = Y , and the subgoal r(A; B) becomes answerable! This example suggests that we need to consider all the total orders of the query variables, similar to the idea in =-=[12]-=-. 3 Theorem 7. Let Q be a CQAC, and\Omega (Q) be the set of all the total orders of the variables in Q that satisfy the comparisons of Q. For eachs2\Omega (Q), let Q be the corresponding query that in... |

90 | Query optimization in the presence of limited access patterns
- FLORESCU, LEVY, et al.
- 1999
(Show Context)
Citation Context ...ry Q 1 in Example 1 is a stable query. As illustrated by the example, we might think that we can test the stability of a query by checking the existence of a feasible order of all its subgoals, as in =-=[9, 28]-=-. An order of subgoals is feasible if for each subgoal in the order, the variables bound by the previous subgoals provide enough bound arguments that the relation for the subgoal can be accessed using... |

89 | Recursive plans for information gathering
- M, Levy
- 1997
(Show Context)
Citation Context ...contained in the maximal answer to the second one? Clearly the solution to this problem can be used to answer queries efficiently. Given a conjunctive query on relations with limited access patterns, =-=[8, 17]-=- show how to construct a recursive datalog program [24] to compute the maximal answer to the query. That is, we can retrieve tuples from relations by retrieving as many bindings from the relations and... |

86 |
Decidable Optimization Problems for Database Logic Programs (Preliminary Report
- Cosmadakis, Gaifman, et al.
- 1988
(Show Context)
Citation Context ...of datalog programs is undecidable [23], our problem of query containment seems undecidable. However, in Section 6 we prove this containment problem is decidable using the results of monadic programs =-=[6]-=-. Our results extend the recent results by Millstein, Levy, and Friedman [19], since we loosen the assumption in that paper. We also discuss how to test the containment efficiently when the program fo... |

85 | Obtaining complete answers from incomplete databases
- Levy
- 1996
(Show Context)
Citation Context ..., but the computed answer may not be the complete answer. As we saw in Section 5, we can sometimes use the approach in that paper to compute the complete answer to a nonstable conjunctive query. Levy =-=[13]-=- considers the problem of obtaining complete answers from incomplete databases, and the author does not consider relations with binding restrictions. Acknowledgments: We thank Foto Afrati, Mayank Bawa... |

80 | On the equivalence of recursive and nonrecursive Datalog programs - Chaudhuri, Vardi - 1997 |

73 | Constraint checking with partial information
- Gupta, Sagiv, et al.
- 1994
(Show Context)
Citation Context ...neralize the algorithm CQstable* as follows. Given a CQAC Q, we compute its answerable subquery Q a . We test the stability of Q by testing whether Q a v Q, which can be tested using the algorithm in =-=[11, 29] ("the GZO algo-=-rithm" for short). However, the following example shows that this "algorithm" does not always work. Example 4. Consider query P : ans(Y ) :- p(X); r(X; Y ); r(A; B); A ! B; XsA; AsY whe... |

69 |
Equivalence of datalog queries is undecidable
- Shmueli
- 1993
(Show Context)
Citation Context ...aximal answer to another conjunctive query, we need to test whether the datalog program for the first one is contained in that for the second one. Since containment of datalog programs is undecidable =-=[23]-=-, our problem of query containment seems undecidable. However, in Section 6 we prove this containment problem is decidable using the results of monadic programs [6]. Our results extend the recent resu... |

60 | Query Planning and Optimization in Information Integration
- Duschka
- 1997
(Show Context)
Citation Context ...cates in P 1 and P 2 , named p 1 and p 2 respectively, have arity m. Let Q be the datalog query consisting of all the rules in P 1 and P 2 , and of the rules: 2 The idea of the proof is borrowed from =-=[7]-=-, Chapter 2.3. r 1 : ans(X 1 ; : : : ; Xm ) :- p 1 (X 1 ; : : : ; Xm ); e(Z) r 2 : ans(X 1 ; : : : ; Xm ) :- p 2 (X 1 ; : : : ; Xm ) where e is a new 1-ary relation with the binding pattern b. Variabl... |

55 | Query containment for data integration systems
- Millstein, Levy, et al.
- 2002
(Show Context)
Citation Context ...ems undecidable. However, in Section 6 we prove this containment problem is decidable using the results of monadic programs [6]. Our results extend the recent results by Millstein, Levy, and Friedman =-=[19]-=-, since we loosen the assumption in that paper. We also discuss how to test the containment efficiently when the program for a query in the test is inherently not recursive. 2 Preliminaries Limited ac... |

48 | Computing capabilities of mediators
- Yerneni, Li, et al.
- 1999
(Show Context)
Citation Context ...s complete answer. Li et al. [18] study the problem of generating an executable plan based on source restrictions. [9, 28] study query optimization in the presence of binding patterns. Yerneni et al. =-=[27]-=- consider how to compute mediator restrictions given source restrictions. These four studies do not minimize a conjunctive query before checking its feasibility. Thus, they regard the query Q 2 in Exa... |

45 | Capability based mediation in TSIMMIS - Li, Yerneni, et al. - 1998 |

40 | Optimizing large join queries in mediation systems
- YERNENI, LI, et al.
- 1999
(Show Context)
Citation Context ...ry Q 1 in Example 1 is a stable query. As illustrated by the example, we might think that we can test the stability of a query by checking the existence of a feasible order of all its subgoals, as in =-=[9, 28]-=-. An order of subgoals is feasible if for each subgoal in the order, the variables bound by the previous subgoals provide enough bound arguments that the relation for the subgoal can be accessed using... |

35 | Query planning with limited source capabilities
- Li, Chang
- 2000
(Show Context)
Citation Context ...contained in the maximal answer to the second one? Clearly the solution to this problem can be used to answer queries efficiently. Given a conjunctive query on relations with limited access patterns, =-=[8, 17]-=- show how to construct a recursive datalog program [24] to compute the maximal answer to the query. That is, we can retrieve tuples from relations by retrieving as many bindings from the relations and... |

29 | K.: Answering Queries Using Materialized Views with Disjunctions
- Afrati, Gergatsoulis, et al.
- 1999
(Show Context)
Citation Context ...nomial-time algorithm for testing the boundedness of the datalog program for a connection query. 7 Related Work Several works consider binding patterns in the context of answering queries using views =-=[8, 14, 1]-=-. Rajaraman, Sagiv, and Ullman [21] propose algorithms for answering queries using views with binding patterns. In that paper all solutions to a query compute the complete answer to the query; thus on... |

23 | Computing complete answers to queries in the presence of limited access patterns
- Li
(Show Context)
Citation Context ...to the query. 1 Corollary 1. A CQ is stable if it has an equivalent query that is feasible. 1 We do not provide all the proofs of the lemmas and theorems in this paper due to space limitations. Refer =-=[15, 16]-=- for details. The query Q 2 in Example 2 is the minimal equivalent query of Q 1 . A CQ is minimal if it has no redundant subgoals, i.e., removing any of its subgoals will yield a nonequivalent query. ... |

22 | On efficient reasoning with implication constraints
- Zhang, OzsoyoË‡glu
- 1993
(Show Context)
Citation Context ...neralize the algorithm CQstable* as follows. Given a CQAC Q, we compute its answerable subquery Q a . We test the stability of Q by testing whether Q a v Q, which can be tested using the algorithm in =-=[11, 29] ("the GZO algo-=-rithm" for short). However, the following example shows that this "algorithm" does not always work. Example 4. Consider query P : ans(Y ) :- p(X); r(X; Y ); r(A; B); A ! B; XsA; AsY whe... |

5 | Query answering using views for data integration over the web
- Calvanese, Giacomo, et al.
- 1999
(Show Context)
Citation Context ... query is contained in the maximal answer to the second one? We show this problem is decidable using the results of monadic programs. 1 Introduction The goal of information-integration systems (e.g., =-=[3, 20, 25]-=-) is to support seamless access to heterogeneous data sources. In these systems, a user poses a query on a mediator [26], which computes the answer by accessing the data at the underlying source relat... |

5 | Testing query containment in the presence of limited access patterns
- Li, Chang
- 1999
(Show Context)
Citation Context ...to the query. 1 Corollary 1. A CQ is stable if it has an equivalent query that is feasible. 1 We do not provide all the proofs of the lemmas and theorems in this paper due to space limitations. Refer =-=[15, 16]-=- for details. The query Q 2 in Example 2 is the minimal equivalent query of Q 1 . A CQ is minimal if it has no redundant subgoals, i.e., removing any of its subgoals will yield a nonequivalent query. ... |