CSCI 403 Database Management. IN in WHERE. The IN Crowd. IN with Tuples 9/11/2018 SUBQUERIES WITH IN. 10 Subqueries

Transcription

1 CSCI 403 Database Management 10 Subqueries IN 2 The IN Crowd Another operator: IN Tests for presence in a list of values or tuples Can be used man places, but usuall in WHERE Trivial eamples: SELECT 17 in (42, 99, 103);?column? f SELECT 42 in (42, 99, 103);?column? t IN in WHERE SELECT course_id, section, instructor, title FROM mines_courses WHERE course_id IN ('CSCI261', 'CSCI262', 'CSCI999'); course_id section instructor title CSCI262 R01 Anderson, Ale DATA STRUCTURES CSCI262 R02 Painter-Wakefield, Christopher DATA STRUCTURES CSCI262 B Painter-Wakefield, Christopher DATA STRUCTURES CSCI261 B Gruchow, Colten PROGRAMMING CONCEPTS CSCI261 C Paone, Jeffre PROGRAMMING CONCEPTS CSCI261 D Paone, Jeffre PROGRAMMING CONCEPTS CSCI261 E Schilling, Samuel PROGRAMMING CONCEPTS CSCI262 A Anderson, Ale DATA STRUCTURES CSCI261 A Gallegos, Lorenzo PROGRAMMING CONCEPTS CSCI262 R03 Painter-Wakefield, Christopher DATA STRUCTURES 3 4 IN with Tuples IN can also match tuples ordered lists of values: SELECT course_id, section, instructor, title FROM mines_courses WHERE (instructor, section) IN (('Fisher, Wend','A'), ('Han, Qi','B')); course_id section instructor title CSCI341 B Han, Qi COMPUTER ORGANIZATION CSCI250 A Fisher, Wend BUILDING A SENSOR SYSTEM CSCI303 A Fisher, Wend DATA SCIENCE SUBQUERIES WITH IN 5 6 1

2 IN and Subqueries Queries can return a list of values (or tuples). We can substitute a quer for a list: IN and Subqueries (SELECT name FROM mines_cs_facult); (SELECT name FROM mines_cs_facult); To interpret this, replace the subquer with the list resulting from the subquer: ('Han, Qi', 'Painter-Wakefield, Christopher', 'Paone, Jeffre', ) A subquer, also known as a nested quer. 7 8 Here we match tuples: Another Eample WHERE (instructor, section) IN (SELECT name, 'A' FROM mines_cs_facult); Of course, for this eample we could instead do: (SELECT name FROM mines_cs_facult) AND section = 'A'; Subqueries vs Joins Subquer queries are often equivalent to join queries: (SELECT name FROM mines_cs_facult); vs SELECT mc.* FROM mines_courses AS mc, mines_cs_facult AS mcf WHERE mc.instructor = mcf.name; 9 10 Subqueries vs Joins: A Difference foo cherr 99 bar wh zee WHERE IN (SELECT wh FROM bar); SELECT foo.* FROM foo, bar WHERE foo. = bar.wh; OTHER SUBQUERY SETTINGS

3 What Can a Subquer Return? A table (a list of tuples) A single tuple A (scalar) value Nothing Subqueries Returning a Table This is partl interpretation: all four cases fall under this categor. However, onl some operators work with this general case: [NOT] IN [NOT] EXISTS [NOT] UNIQUE (not implemented in PostgreSQL) EXISTS & UNIQUE Unlike IN, EXISTS and UNIQUE do not compare with epressions. EXISTS returns true iff a subquer returns anthing: WHERE EXISTS (SELECT * FROM bar WHERE wh = 77); EXISTS & UNIQUE UNIQUE returns true iff a subquer returns unique values onl. Both of these operators are more useful in correlated subqueries (net section) Subqueries Returning a Single Tuple A quer returning a scalar value is a special case of this. A quer returning nothing acts like a quer returning a tuple of NULL values. These subqueries can be used in equalit comparison epressions (eample net page): Single Tuple Eample foo cherr 99 WHERE = (SELECT wh FROM bar WHERE zee = ' '); banana 17 wh bar zee

4 Single Tuple Countereample Non-Scalar Tuple foo cherr 99 wh WHERE = (SELECT wh FROM bar WHERE zee > ' '); ERROR: more than one row returned b a subquer used as an epression bar zee If ou have multiple value tuples (not just a scalar as in previous eample), use parentheses: SELECT FROM tablename WHERE (epr1, epr2, ) = (SELECT sq_epr1, sq_epr2, ); Scalar Value Scalar values can be used in an epression: WHERE > (SELECT wh FROM bar WHERE zee = ' '); Doesn t have to be in WHERE, either: SELECT (SELECT wh FROM bar WHERE zee = ' '); CORRELATED SUBQUERIES Again, a nothing result is interpreted as NULL Correlated Subqueries 1 In a correlated subquer, the subquer accesses attributes from rows in the outer quer. Here s an eample from Wikipedia: SELECT emploee_number, name FROM emploees AS e1 WHERE salar > (SELECT AVG(salar) FROM emploees AS e2 WHERE e2.department = e1.department); Correlated Subqueries 2 SELECT AVG(salar) FROM emploees AS e2 WHERE e2.department = e1.department This subquer uses an operator we haven t covered et, which provides an aggregate value over a table. AVG(salar) gives the average of the salar values in all rows matching the WHERE condition

5 Correlated Subqueries 3 SELECT emploee_number, name FROM emploees AS e1 WHERE salar > (SELECT AVG(salar) FROM emploees AS e2 WHERE e2.department = e1.department); The highlighted comparison shows the correlation. Conceptuall, a correlated subquer is run once for ever row in the outer quer. The epression e1.department, then comes from some row in the outer quer. Correlated Subqueries 4 SELECT emploee_number, name FROM emploees AS e1 WHERE salar > (SELECT AVG(salar) FROM emploees AS e2 WHERE e2.department = e1.department); So what is this doing? For each emploee: get the average of salaries in the emploee s department; if the emploee s salar is greater than average, include it in the result Another Eample Here s an eample from the csci403 DB: SELECT instructor, course id, section FROM mines_courses AS mc1 WHERE course id IN (SELECT course id FROM mines_courses AS mc2 WHERE mc2.course id = mc1.course id AND mc2.instructor <> mc1.instructor); Another Eample 2 The above quer is equivalent to: SELECT instructor, course id, section FROM mines_courses AS mc1 WHERE EXISTS (SELECT course id FROM mines_courses AS mc2 WHERE mc2.course id = mc1.course id AND mc2.instructor <> mc1.instructor); Subqueries in FROM Not sure wh ou d want to, but this is legal: SUBQUERIES IN OTHER CLAUSES SELECT course_id FROM (SELECT course_id, instructor FROM mines_courses) AS mc WHERE mc.instructor LIKE 'Painter%';

6 Subqueries in SELECT and SET Subquer in SET Eample We saw an eample using a subquer returning a scalar in a SELECT clause epression. More usefull, we can use single-tuple subquer results in a SET clause in an UPDATE quer (especiall, with correlation). This gives us something like a join that works with UPDATE! From the PostgreSQL docs: --Update contact names in an accounts table to match the currentl assigned salespeople: UPDATE accounts SET (contact_first_name, contact_last_name) = (SELECT first_name, last_name FROM salesperson WHERE salesperson.id = accounts.sales_id); Up Net Net lecture: Grouping and aggregation. 33 6