Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 212 Part II Manipulating Data With Select RESULT: SELECT LastName, DATENAME(yy,DateOfBirth) AS [Year], DATENAME(mm,DateOfBirth) AS [Month], DATENAME(dd,DateOfBirth) AS [Day], DATENAME(weekday, DateOfBirth) AS BirthDay FROM dbo.Guide WHERE DateOfBirth IS NOT NULL; LastName Year Month Day BirthDay Frank 1958 September 4 Thursday TABLE 9-2 DateTime Portions Used by Date Functions Portion Abbreviation year yy, yyyy quarter qq, q month mm, m dayofyear dy, d day dd, d week wk, ww weekday dw hour hh minute mi, n second ss, s millisecond ms microsecond mcs nanosecond ns TZoffset tz There are two supported types: DateTime and DateTime2. DateTime2 is new to SQL Server 2008 and represents time to a much finer granularity: within 100 nanoseconds. ■ DatePart(date portion, date): Returns the ordinal number of the selected portion of the datetime value. The following example retrieves the day of the year and the day of the week as integers: 212 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 213 Data Types, Expressions, and Scalar Functions 9 SELECT DATEPART(dayofyear, CURRENT_TIMESTAMP) AS DayCount; Result: DayCount 321 SELECT DATEPART(weekday, CURRENT_TIMESTAMP) AS DayWeek; Result: DayWeek 7 An easy way to obtain just the date, stripping off the time, is to use a couple of string functions: SELECT CONVERT(char(10), CURRENT_TIMESTAMP, 112) AS "DateTime"; ■ DateAdd(DATE PORTION, AMOUNT, BEGINNING DATE) and DateDiff(DATE PORTION, BEGINNING DATE, ENDING DATE): Performs addition and subtraction on datetime data, which databases often need to do. The DATEDIFF() and the DATEADD() functions are designed expressly for this purpose. The DATEDIFF() doesn’t look at the complete date, only the date part being extracted: select DATEDIFF(year,’september 4 2008’,’november 10 2009’) Result 1 select DATEDIFF(month,’september 4 2008’,’november 10 2009’) 2 The following query calculates the number of years and days that my wife, Melissa, and I have been married: SELECT DATEDIFF(yy,’19840520’, CURRENT_TIMESTAMP) AS MarriedYears, DATEDIFF(dd,’19840520’, CURRENT_TIMESTAMP) AS MarriedDays; Result: MarriedYears MarriedDays 17 6390 The next query adds 100 hours to the current millisecond: SELECT DATEADD(hh,100, CURRENT_TIMESTAMP) AS [100HoursFromNow]; Result: 100HoursFromNow 2009-11-21 18:42:03.507 The following query is based on the Family sample database and calculates the mother’s age at the birth of each child, using the DateDiff() function: 213 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 214 Part II Manipulating Data With Select USE Family; SELECT Person.FirstName + ‘ ‘ + Person.LastName AS Mother, DATEDIFF(yy, Person.DateOfBirth, Child.DateOfBirth) AS AgeDiff,Child.FirstName FROM Person INNER JOIN Person AS Child ON Person.PersonID = Child.MotherID ORDER By Age DESC; The DATEDIFF() function in this query returns the year difference between PERSON.DATEOFBIRTH, which is the mother’s birth date, and the child’s date of birth. Because the function is in a column expression, it is calculated for each row in the result set: Mother AgeDiff FirstName Audrey Halloway 33 Corwin Kimberly Kidd 31 Logan Elizabeth Campbell 31 Alexia Melanie Campbell 30 Adam Grace Halloway 30 James This section discusses functions that are new to SQL Server 2008. You will take a look at the functions and then see the results of some queries. ToDateTimeOffset(expression, time_zone): Returns a DateTimeOffset value The following example gets the date and time for a given time zone: SELECT TODATETIMEOFFSET(CURRENT_TIMESTAMP,’-07:00’); Result: 2009-11-05 11:24:15.490 -07:00 String Functions Like most modern programming languages, T-SQL includes many string-manipulation functions: ■ SUBSTRING(string, starting position, length): Returns a portion of a string. The first parameter is the string, the second parameter is the beginning position of the substring to be extracted, and the third parameter is the length of the string extracted: SELECT SUBSTRING(’abcdefg’, 3, 2); Result: cd 214 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 215 Data Types, Expressions, and Scalar Functions 9 ■ STUFF(string, insertion position, delete count, string inserted):The STUFF() function inserts one string into another string. The inserted string may delete a specified number of characters as it is being inserted: SELECT STUFF(’abcdefg’, 3, 2, ‘123’); Result: ab123efg The following code sample uses nested STUFF() functions to format a U.S. social security number: SELECT STUFF(STUFF(’123456789’, 4, 0, ‘-’), 7, 0, ‘-’); Result: 123-45-6789 ■ CHARINDEX(search string, string, starting position): Returns the character position of a string within a string. The third argument is optional and rarely used in practice. It defaults to 1. SELECT CHARINDEX(’c’, ‘abcdefg’, 1); Result: 3 The user-defined function dbo.pTitleCase() later in this section uses CHARINDEX() to locate the spaces separating words. ■ PATINDEX(pattern, string): Searches for a pattern, which may include wildcards, within a string. The following code locates the first position of either a c or a d in the string: SELECT PATINDEX(’%[cd]%’, ‘abdcdefg’); Result: 3 ■ RIGHT(string, count) and Left(string, count): Returns the rightmost or leftmost part of a string: SELECT LEFT(’Nielsen’,2) AS ‘ [Left] ‘, RIGHT(’Nielsen’,2) AS [Right]; Result: Left Right Ni en 215 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 216 Part II Manipulating Data With Select ■ LEN(string): Returns the length of a string: SELECT LEN(’Supercalifragilisticexpialidocious’) AS [Len]; Result: Len 34 ■ RTRIM(string) and LTrim(string): Removes leading or trailing spaces. While it’s difficult to see in print, the three leading and trailing spaces are removed from the fol- lowing string. They are often used together as RTRIM(LTRIM(string). I adjusted the column-header lines with the remaining spaces to illustrate the functions: SELECT RTRIM(’ middle earth ‘) AS [RTrim], LTRIM(’ middle earth ‘) AS [LTrim]; Result: RTrim LTrim middle earth middle earth ■ UPPER(string) and Lower(string): Converts the entire string to uppercase or lowercase. There’s not much to know about these two functions, illustrated here: SELECT UPPER(’one TWO tHrEe’) AS UpperCase, LOWER(’one TWO tHrEe’) AS LowerCase; Result: UpperCase LowerCase ONE TWO THREE one two three ■ REPLACE(string, string):TheReplace() function operates as a global search and replace within a string. Using REPLACE() within an update DML command can quickly fix problems in the data, such as removing extra tabs or correcting string patterns. The follow- ing code sample adds apostrophes to the LastName column in the OBXKITES database’s CONTACT table: USE OBXKites; Create test case by modifying one contact’s last name. UPDATE Contact SET LastName = ‘Adam’’s’ WHERE LastName = ‘Adams’; Check the modified sample data and the replacement. SELECT LastName, REPLACE(LastName, ‘’’’, ‘’) AS Replaced FROM Contact WHERE LastName LIKE ‘%’’%’; 216 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 217 Data Types, Expressions, and Scalar Functions 9 Result: LastName Replaced Adam’s Adams To demonstrate the REPLACE() function using an update command, the next query actually changes the data in place and removes any apostrophes: UPDATE Contact SET LastName = REPLACE(LastName, ‘’’’, ‘’) WHERE LastName LIKE ‘%’’%’; Show that the modification was successful. SELECT LastName FROM Contact WHERE LastName LIKE ‘Adam%’; Result: LastName Adams When working with string literals, it’s generally difficult to insert a quote into the string without ending the string and causing a syntax error. SQL Server handles this situation by accepting two single quotes and converting them into one single quote within the string: ’Life’’s Great! ‘ is interpreted as Life’s Great! ■ dbo.pTitleCase(source, search, replace): T-SQL lacks a function to convert text to title case (first letter of each word in uppercase, and the remainder in lowercase). Therefore, the following user-defined function accomplishes that task: CREATE FUNCTION dbo.pTitleCase ( @StrIn NVARCHAR(MAX)) RETURNS NVARCHAR(MAX) AS BEGIN; DECLARE @StrOut NVARCHAR(MAX), @CurrentPosition INT, @NextSpace INT, @CurrentWord NVARCHAR(MAX), @StrLen INT, @LastWord BIT; SET @NextSpace = 1; SET @CurrentPosition = 1; 217 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 218 Part II Manipulating Data With Select SET @StrOut = ‘’; SET @StrLen = LEN(@StrIn); SET @LastWord = 0; WHILE @LastWord = 0 BEGIN; SET @NextSpace = CHARINDEX(’ ‘, @StrIn, @CurrentPosition + 1); IF @NextSpace = 0 no more spaces found BEGIN; SET @NextSpace = @StrLen; SET @LastWord = 1; END; SET @CurrentWord = UPPER(SUBSTRING(@StrIn, @CurrentPosition, 1)); SET @CurrentWord = @CurrentWord + LOWER(SUBSTRING(@StrIn, @CurrentPosition+1, @NextSpace - @CurrentPosition)); SET @StrOut = @StrOut + @CurrentWord; SET @CurrentPosition = @NextSpace + 1; END; RETURN @StrOut; END; Running a user-defined function requires including the owner name in the function name: SELECT dbo.pTitleCase(’one TWO tHrEe’) AS TitleCase; Result: TitleCase One Two Three The dbo.pTitleCase function does not take into consideration surnames with nonstandard capitalization, such as McDonald, VanCamp, or de Jonge. It would be inadequate to hard- code a list of exceptions. Perhaps the best solution is to store a list of exception phrases (Mc, Van, de, and so on) in an easily updateable list. The code for the pTitleCase user-defined function can be downloaded from www.SQLServerBible.com. Soundex Functions Soundex is a phonetic pattern-matching system created for the American census. Franklin Roosevelt directed the United States Bureau of Archives to develop a method of cataloguing the population that could handle variations in the spelling of similar surnames. Margaret K. Odell and Robert C. Russell developed Soundex and were awarded U.S. patents 1261167 (1918) and 1435663 (1922) for their efforts. The census filing card for each household was then filed under the Soundex method. Soundex has been applied to every census since and has been post-applied to census records back to 1880. 218 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 219 Data Types, Expressions, and Scalar Functions 9 The purpose of Soundex is to sort similar-sounding names together, which is very useful for dealing with contact information in a database application. For example, if I call a phone bank and give them my name (Nielsen), they invariably spell it ‘‘Nelson’’ in the contact lookup form, but if the database uses Soundex properly, then I’ll still be in the search-result list box. For more information concerning Soundex and its history, refer to the following websites: ■ www.nara.gov/genealogy/coding.html ■ www.amberskyline.com/treasuremaps/uscensus.html ■ www.bluepoof.com/soundex/ Here’s how Soundex works. The first letter of a name is stored as the letter, and the following Soundex phonetic sounds are stored according to the following code: 1 = B, F, P, V 2 = C, G, J, K, Q, S, X, Z 3 = D, T 4 = L 5 = M, N 6 = R Double letters with the same Soundex code, A, E, I, O, U, H, W, Y, and some prefixes, are disregarded. Therefore, ‘‘Nielsen’’ becomes ‘‘N425’’ via the following method: 1. The N is stored. 2. The i and e are disregarded. 3. The l sound is stored as the Soundex code 4. 4. The s is stored as the Soundex code 2. 5. The e is ignored. 6. The n is stored as the Soundex code 5. By boiling them down to a few consonant sounds, Soundex assigns ‘‘Nielsen,’’ ‘‘Nelson,’’ and ‘‘Neilson’’ the same code: N425. Following are additional Soundex name examples: ■ Brown = B650 (r = 6, n = 5) ■ Jeffers = J162 (ff = 1, r = 6, s = 2) ■ Letterman = L365 (tt = 3, r = 6, m = 5) ■ Nicholson = N242 (c = 2, l = 4, s = 2) ■ Nickols = N242 (c = 2, l = 4, s = 2) SQL Server includes two Soundex-related functions, SOUNDEX() and DIFFERENCE(). 219 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 220 Part II Manipulating Data With Select Using the SOUNDEX() function The SOUNDEX(string) function calculates the Soundex code for a string as follows: SELECT SOUNDEX(’Nielsen’) AS Nielsen, SOUNDEX(’Nelson’) AS NELSON, SOUNDEX(’Neilson’) AS NEILSON; Result: Nielsen NELSON NEILSON N425 N425 N425 Other, more refined, Soundex methods exist. Ken Henderson, in his book The Guru’s Guide to Transact SQL (Addison-Wesley, 2000), provides an improved Soundex algorithm and stored procedure. If you are going to implement Soundex in a production application, I recommend exploring his version. Alternately, you can research one of the other refined Soundex methods on the websites listed previously and write your own custom stored procedure. There are two possible ways to add Soundex searches to a database. The simplest method is to add the SOUNDEX() function within the WHERE clause, as follows: USE CHA2; SELECT LastName, FirstName FROM dbo.Customer WHERE SOUNDEX(’Nikolsen’) = SOUNDEX(LastName); Result: LastName FirstName Nicholson Charles Nickols Bob While this implementation has the smallest impact on the data schema, it will cause performance issues as the data size grows because the SOUNDEX() function must execute for every row in the database, and an index on the name column (if any) cannot be used with an efficient seek operation, but only with a much more expensive scan. A faster variation of this first implementation method pre-tests for names with the same first letter, thus enabling SQL Server to use any indexes to narrow the search, so fewer rows must be read and the SOUNDEX() function must be performed only for rows selected by the index: SELECT LastName, FirstName FROM dbo.Customer WHERE SOUNDEX(’Nikolsen’) = SOUNDEX(LastName) AND LastName LIKE ‘N%’; The first query executes in 37.7 milliseconds on my test server, while the improved second query exe- cutes in 6.5 milliseconds. I suspect that the performance difference would increase with more data. 220 www.getcoolebook.com Nielsen c09.tex V4 - 07/21/2009 12:40pm Page 221 Data Types, Expressions, and Scalar Functions 9 The second implementation method is to write the Soundex value in a column and index it with a non- clustered index. Because the Soundex value for each row is calculated during the write, the SOUNDEX() function does not need to be called for every row read by the SELECT statement. This is the method I recommend for a database application that heavily depends on Soundex for contact searches. The OBXKITES sample database demonstrates this method. The SoundexCode column is persisted cal- culated column, so it’s automatically calculated for every insert and kept updated with every update. Searching for a row, or all the matching rows, based on the stored Soundex code is extremely fast. First determine the Soundex for ‘‘Smith’’: USE OBXKites; SELECT SOUNDEX(’Smith’); Result: S530 Knowing the Soundex value for ‘‘Smith,’’ the Soundex search is now a fast index seek without ever call- ing the SOUNDEX() function for the row being read during the select statement: SELECT LastName, FirstName, SoundexCode FROM Contact WHERE SoundexCode = ‘S530’; Result: LastName FirstName SoundexCode Smith Ulisius S530 Smith Oscar S530 Using the DIFFERENCE() Soundex function The second SQL Server Soundex function, DIFFERENCE(), returns the Soundex difference between two strings in the form of a ranking from 1 to 4, with 4 representing a perfect Soundex match: USE CHA2 SELECT LastName, DIFFERENCE(’Smith’, LastName) AS NameSearch FROM Customer ORDER BY Difference(’Smith’, LastName) DESC; Result: LastName NameSearch Smythe 4 Spade 3 Zeniod 3 221 www.getcoolebook.com . types: DateTime and DateTime2. DateTime2 is new to SQL Server 2008 and represents time to a much finer granularity: within 100 nanoseconds. ■ DatePart(date portion, date): Returns the ordinal number. the complete date, only the date part being extracted: select DATEDIFF(year,’september 4 2008 ,’november 10 2009’) Result 1 select DATEDIFF(month,’september 4 2008 ,’november 10 2009’) 2 The following. Campbell 30 Adam Grace Halloway 30 James This section discusses functions that are new to SQL Server 2008. You will take a look at the functions and then see the results of some queries. ToDateTimeOffset(expression,