Date:         Fri, 3 Jun 2005 12:22:17 -0400
Reply-To:     Venky Chakravarthy <swovcc@HOTMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Venky Chakravarthy <swovcc@HOTMAIL.COM>
Subject:      Re: using Perl regex to validate values in SAS 9.1
Comments: To: rune@FASTLANE.NO
Content-Type: text/plain; charset=ISO-8859-1

What are the rules for Van Husen, Jeff who has both Phone1 and Phone2
missing?

On Fri, 3 Jun 2005 15:40:21 +0200, Rune Runnestø <rune@FASTLANE.NO> wrote:

>Yet another small justification of the conditions.
>The criteria for the records in the dataset WRONGRECORDS is that there
shall
>be one or more non-numbers in one or both of the columns phone1 and phone2.
>The code seems to work OK with one exception: The record with Mary O'Reily
>should not be in WRONGRECORDS. Because blank values are allowed (phone1). I
>*suppose* it has come here because of the blank value in phone1.
>
>
>data staff;
>    attrib  name    length=$20
>            phone1  length=$12
>            phone2  length=$12;
>    infile cards;
>    input   @1 name $20.
>            @21 phone1 $12.
>            @34 phone2 $12.;
>datalines;
>O'Reily, Mary                    123456789888
>Pyles, Jane         2B3667567715 987654328887
>Hoffman, Valerie    2C268660808  234234234234
>Dawn, Jennifer      718638361549 5B7898765476
>Van Husen, Jeff
>Sim-Smith, Angela   201681265665 5C9876567865
>Timmons, David      5866806      34543234
>Benjamin, Catherine 203678161777 234567432334
>Windsor, Stephen    718638462849 876422345564
>Richardson, Larry   718638468816 556768432234
>Bellum, Sarah       2066756434   3454799665
>Garcia Tracy        21268761247  348564332
>Montgomery, Adam    212658763622 789023456677
>George, Clara       203678161212 345656667777
>Sabatini, Anthony   203678160019 3E5678932RT5
>;
>run;
>
>data wrongrecords;
>    set staff;
>    where not missing(phone1) or not missing(phone2); /* we want to push
>away
>            the records where *both* phone1 and phone2 are blank.
>            If we had used 'and' instead of 'or', we would wrongly
>            pushed away Mary O'Reily too. */
>    *alternatively, we could have used this sentence:;
>    *where phone1 ne "" or phone2 ne "";
>    if _n_ = 1 then
>        do;
>            retain re;
>            re = prxparse('/[^\d]/'); /* match only non-numbers*/
>            if missing(re) then
>                do;
>                    putlog 'ERROR: malformed regex';
>                    stop;
>                end;
>        end;
>    if prxmatch(re,trim(phone1)) or prxmatch(re,trim(phone2));
>run;
>
>
>Regards
>Rune