On Wed, 8 May 1996, Howard_Schreier@ITA.DOC.GOV wrote:

>On Wed, 8 May 1996 00:06:21 +0100, John Whittington <johnw@MAG-NET.CO.UK> >wrote (in part): >> Of course, one solution would have been for SAS to reserve this 'odd >> behavior', as Tim regards it, only for the situation of the deafult >> (space) delimiter. > >But now of course it's best not to change anything which will break >anybody's existing code.

Agreed, as a principle - but, as I said in my previous message, I seriously doubt that there is any significant amount of code around which would be broken - since I cannot believe that many people use non-significant multiple delimiters when the delimiter is a 'visible' character - i.e. I doubt whether many people deal with data which looks like:

1#2#3######4##Apple###"Quoted"####6 or even 1,2,3,,,,,,4,,Apple,,,"Quoted",,,,6 (where multiple delimiters are 'non-significant')

>My complaint is that the DSD option changes two different behaviors of >INFILE: (1) makes consecutive delimiters significant and (2) enables quoting >of character values, so that such values can contain instances of the >delimiter character. I regularly get files which are blank delimited with >non-significant multiple blanks and which also contain quoted strings. In >other words, I need to utilize #2 but not #1, and I can't. >If SAS would provide separate options for these behaviors, I for one would >accept any combination of defaults, no matter how strange.

Yes, I agree. I get data like that, too. There is, as far as I know, no very simple solution. If the character variables are of fixed length, then one can obviously read them in with a fixed length format (which copes with embedded 'delimiters'), and then get rid of any quotes with something like:

array x _character_ ; do over x ; x = dequote(x) ; end ;

... but that approach won't work if the variables are (as they usually are!) of varying length. My usual approach is to pre-process to eliminate mutiple delimiters between variables, and then to use DSD, but I agree this is a pain - as is the alternative, to read into SAS character by character and then re-build the variables.

To maintain consistency with the existing state of affairs, I presume that they would have to *add* a new option, which simply dealt with the quoted strings, leaving DSD still doing both things. To complete the set, I suppsoe one ought to have the third option, as well - to treat multiple delimiters as significant, but NOT to handle quoted strings.

John

----------------------------------------------------------- Dr John Whittington, Voice: +44 1296 730225 Mediscience Services Fax: +44 1296 738893 Twyford Manor, Twyford, E-mail: johnw@mag-net.co.uk Buckingham MK18 4EL, UK CompuServe: 100517,3677 -----------------------------------------------------------