8. An example row consists of constants and example elements which are really domain variables.

Transcription

1 CMPT Lecture Notes June 9, 1998 Chapter 4 Other Relational Languages 4.1 Quer-b-Eample èqbeè 1. QBE is both a quer language and the name of a DB sstem including it. The sstem is no longer in use, but the language is part of IBM's Quer Management Facilit èqmfè Basic Structure 1. QBE has ëtwo-dimensional" snta. 2. Queries are epressed b eample. 3. Close correspondence with domain relational calculus 4. Non-procedural. 5. Queries are epressed using skeleton tables. 6. User selects the skeletons needed. 7. User ælls in skeletons with eample rows. 8. An eample row consists of constants and eample elements which are reall domain variables. 9. Domain variables are preceded b an underscore character. 10. Constants appear without an qualiæcation. We'll look at eamples from the tet Simple Queries 1. For eample, to ænd all customers having an account at the branch: æ A P. before the variable causes printing. æ A P.ALL. preæ suppresses duplicate elimination. æ AP. in front of the row prints all attributes. 37

2 38 CHAPTER 4. OTHER RELATIONAL LANGUAGES æ The domain variable ma be omitted if it is not used elsewhere. æ Arithmetic epressions are allowed. æ Comparison operators are allowed, space on left hand side is left blank. 2. To ænd the names of all branches not located in Burnab: P. : Burnab 3. To ænd all customers having an account at both the and the MetroTown branch: MetroTown 4. To ænd all customers having an account at either branch or both: MetroTown P. 5. Find all customers having an account at the same branch as Jones: Jones P Queries on Several Relations 1. Queries on several relations require several skeleton tables. 2. To ænd the name and cit of all customers having a loan at the branch: customer cname street ccit P. 3. Find the name of all customers having an account at the branch, but no loan from that branch. Queries involving negation can be epressed b putting a : sign under the relation name beside an eample row: 4. To ænd all customers who have accounts at two diæerent branches: : The Condition Bo : 1. When it is diæcult or impossible to epress all constraints on the domain variables within the skeleton tables, the condition bo ma be used. 2. To add the constraint that we are onl interested in customers other than Jones to the above quer, we include the condition bo: := Jones

3 4.1. QUERY-BY-EXAMPLE èqbeè To ænd all account numbers with balances between $1,300 and $1,500: ç 1300 ç Logical epressions and and or ma appear in the condition bo. 5. To ænd all account numbers where the balance is between $1,300 and $2,000, but is not $1,500: =èç1300 and ç 2000 and : 1500è 6. An unconventional use of the or construct allows comparison with several constant values: = èburnab or Richmondè The Result Relation 1. If the result of a quer includes attributes from several relation schemes, we need a wa of displaing the result in a single table. 2. We can declare a temporar result relation including the attributes to be displaed. We put the print command onl in that table. 3. To ænd the customer names and cities and account numbers for all customers having an account at the branch: z customer cname street ccit result cname ccit accountè z Ordering the Displa of Tuples 1. The order in which tuples are displaed can be controlled b adding the command AO. èascending orderè or DO. èdescending orderè to the print command: P.AO. 2. To sort ærst b name, and then b balance for those with multiple accounts: P.AOè1è. P.DOè2è.

4 40 CHAPTER 4. OTHER RELATIONAL LANGUAGES Aggregate Operations 1. QBE includes the aggregate operators AVG, MAX, MIN, SUM and CNT. As QBE eliminates duplicates b default, the must have ALL. appended to them. 2. To ænd the total balance of all accounts belonging to Jones: Jones P.SUM.ALL. 3. All aggregate operators must have ALL. appended, so to override the ALL. we must add UNQ. èuniqueè. ènote: a number of eamples in the tet incorrectl show UNQ. replacing ALL.è Main P.CNT.UNQ.ALL. 4. To compute functions on groups, we use the G. operator. To ænd the average balance at each branch: P.G. P.AVG.ALL. 5. To ænd the average balances at onl branches where the average is more than $1,200, we add the condition bo: AVG.ALL. é To ænd all customers who have an account at all branches located in Burnab, we can do: z CNT.UNQ.ALL. = CNT.UNQ.ALL. z P.G. Burnab Burnab Modifing the Database 1. QBE has facilities for modifing the database. Deletion 1. We simpl use D. instead of the P. operator. Whole tuples ma be deleted, or onl some columns. 2. Delete all of Smith's account records: D. Smith 3. Delete the branch-cit value for the branch: D.

5 4.2. QUEL Delete all loans with loan numbers between 1300 and 1500: D. =èç1300 and ç 1500è 5. Delete all accounts at branches located in Burnab: D. Burnab Insertion 1. Insertion uses the I. operator. 2. To insert an account tuple for Smith: I Smith If values are missing, nulls are inserted. 4. To provide all loan customers in the branch with a $200 savings account: I. 200 Updates 1. We can update individual attributes with the U. operator. Fields left blank are not changed. 2. To update the assets of the branch to $10,000,000: U To make interest paments of 5è on all balances: U. * Quel 1. We will not cover this section, aside from making a few remarks about this language. æ Quel was the original quer language for the Ingres dbms. Ingres is now available with SQL. æ Quel closel resembles the tuple relational calculus. æ Queries use the range of, retrieve and where clauses.

6 42 CHAPTER 4. OTHER RELATIONAL LANGUAGES æ A tpical quer: range of t is borrow range of s is deposit retrieve unique ès.cnameè where t.bname=ë" and s.bname=ë" and t.cname=s.cname This ænds the names of all customers who have both a loan and an account at the branch. æ There is no representation for 8 or 9 in Quel. æ Quel has the power of the relational algebra b means of the an aggregate function and the use of insertion and deletion into temporar relations. Converting Queries Easil Into An Language 1. I've found that students èand mselfè have trouble getting from a database quer epressed in English to a quer epressed in one of the languages we have covered. 2. I've found a method that seems to help, involving an intermediate step. èaè Decide on the relations required to answer the quer. æ You'll need relations containing attributes eplicitl mentioned, plus relations needed to ëtraverse" between needed relations. æ In some cases ou will need more than one cop of a relation. æ Don't include unneeded relations. èbè Draw them on a piece of paper. æ It helps to draw them in a sensible order. æ Draw them in the order ou would ëtraverse" them. This will simplif the drawing of links. ècè Draw in links and constant values. æ Put links between attributes in diæerent relations wherever the attributes are required to satisf some comparison operator èequals, less than, etc.è. æ I use ordinar lines for equals, and write an other comparison operator on the line at some convenient spot. æ Write in constant values, where some attribute must have a speciæc value. èdè Now take the diagram, and convert it into the language required. We'll discuss this stage in more detail. 3. Converting the Diagram to a Speciæc Language Quer For simpler queries, the following advice works. Where ou need set operations or division, a little more thought is needed. èaè Relational Algebra: we'll do a correct but not necessaril optimal quer. æ Do an appropriate combination of Cartesian products and natural joins of the relations required. æ Do a select where the predicate demands that all the links and constants in our diagram be true. æ Don't forget that natural joins will take care of some of our diagram's links. æ Finall, do a project of the attributes to be printed out. èbè Tuple Relational Calculus: æ Create a tuple variable for each of the relations in our diagram. æ Make sure the parentheses give ou the required scope. æ Ensure each link and constant in our diagram corresponds to some part of our predicate. æ Make sure t gets the attributes that should be printed out.

7 4.2. QUEL 43 ècè Domain Relational Calculus: æ Create domain variables. Name them sensibl. æ Remember that equalit is forced b using the same domain variable in several places. æ Other comparison operators ma be eplicitl stated, e.g. s ç u. æ Remember to use the eistential qualiæer for domain variables, and to make sure our scoping is correct. èdè SQL: similar to relational algebra. æ Put all the relations needed in the from clause. æ Remember to use tuple variables when ou have more than one cop of a relation, or for general convenience. æ Epress each of the links and constants in our diagram as part of the predicate in the where clause. æ State the attributes to be printed out in the select clause. èeè QBE: our diagram is almost QBE to start with. æ Select the skeleton tables needed. æ Remember that ou onl need one skeleton table per relation. You can put more than one line in a skeleton table. æ Force equalit on links b using the same domain variables in diæerent places èsee the connection to domain relational calculus?è. æ Use the condition bo where necessar. æ Use P. to print out the attributes. Remember to use a result relation if attributes are printed out from more than one skeleton table. More complicated queries will take more thought, but I believe this intermediate step of making a diagram is alwas helpful.