Thank you both for your responses. While datatype="D" does occur in some of the header records, the PNODE_TYPE var doesn't. So the solution with only one infile statement works perfectly. I really appreciate your help - I knew this had to be possible but couldn't get it to work despite much fiddling.

Sarah

-----Original Message----- From: Ian Whitlock [mailto:WHITLOI1@WESTAT.com] Sent: Friday, September 12, 2003 4:09 PM To: 'Crawford' Cc: 'swhittier@ISO-NE.COM' Subject: RE: reading from hierarchical files without datatype identifier

Peter,

It really depends on the data whether two INFILEs are needed. I suspect that DATATYPE="D" does not occur on the first 6 records. If true the step can be simplified and look more "standard".

data new_lmps_0( label="read from &file" ); infile solve1 ; if _n_ = 1 then input /// @37 udscaseid $19. // @38 lmpcaseid $16. ; retain udscaseid lmpcaseid ;

input @1 datatype $1. @72 PNODE_TYPE $1. @; if datatype="D" and PNODE_TYPE in ("H" "I" "Z") ;

input @29 date date11. @41 time time5. @48 pnode_name $15. @80 lmp 9.2 @118 loss 8.2 @127 cong 8.2 ; format date date. time time.; run;

When using the two INFILE buffers it is important to use FILENAME statements and filerefs so the buffers can have different names. You did this, but it is important to know that you have to do it that way.

infile "..." ;

is not good enough when you want two data streams.

IanWhitlock@westat.com

-----Original Message----- From: Crawford [mailto:PeterDOTCrawfordATblueyonder.co.uk@Peter.BITNET] Sent: Friday, September 12, 2003 3:35 PM To: SAS-L@LISTSERV.UGA.EDU Subject: Re: reading from hierarchical files without datatype identifier

Hi Sarah To avoid a step joining to datasets, you could just do the two infiles in one step... like

%let file = c:\lmp\data\finalize\MSS_7020030728205502_0X.0_28-JUL-2003_16_55.lmpsolved ; filename solve1 "&file" lrecl =3000; filename solve2 "&file" lrecl =3000;

data new_lmps_0( label="read from &file" ); infile solve1 firstobs=4 ; input @37 udscaseid $19. // @38 lmpcaseid $16. ;

do until( eof ); infile solve2 firstobs=1 ; input @1 datatype $1. @72 PNODE_TYPE $1. @;

if datatype="D" and PNODE_TYPE in ("H" "I" "Z") then do; input @29 date date11. @41 time time5. @48 pnode_name $15. @80 lmp 9.2 @118 loss 8.2 @127 cong 8.2 ; format date date. time time.; output; end /* handling datatype D*/ ;

end /* handling a record from SOLVED2 */ ; stop; /* end of file on solved2 */ run;

Comment: instead of 2 data steps to read the data, I used 2 infile statements This provides 2 streams of data in the data step. Rather than have to bypass handling the SOLVE1 file on all data step iterations after the first, I put a do until( eof ); loop around the reading of data from SOLVE2 It doesn't add a lot. Just the "end=eof" on the infile, an output statement, and a stop statement to avoid having the data step iterate a second time.

welcome to the wonderful data step

Good Luck

Peter Crawford Crawford Software Consultancy Limited UK >

"Sarah Whittier" <swhittier@ISO-NE.COM> wrote in message news:200309121818.h8CII6t09161@listserv.cc.uga.edu... > I have a problem related to reading data from flat files into SAS. I > am using Windows, and the files are text files. The files are > essentially hierarchical files, however, the rows with the header data > that I want are not easily identified (i.e., they don't have a > consistent datatype identifier). > > I want to read certain key values from the 4th and 6th rows of the > file and > have these values on each row of the data values in the output SAS > dataset. I can achieve this by using two data steps with input statements, > and then a third to set them together. I would prefer to do this in > just one data step, which I hope will make it easier to set up a > program to read > in the 30 or so files that I have. Does anyone have a suggestion for > modifying the program below to use just one data step? I hope this is > sufficiently clear without including data. > > > filename solved > "c:\lmp\data\finalize\MSS_7020030728205502_0X.0_28-JUL- > 2003_16_55.lmpsolved"; > > data new_lmps_id; > infile solved firstobs=4 obs=6; > > input @37 udscaseid $19. > // @38 lmpcaseid $16. ; > run; > > data new_lmps; > infile solved; > > input @1 datatype $1. @72 PNODE_TYPE $1. @; > > if datatype="D" and PNODE_TYPE in ("H" "I" "Z") then do; > input @29 date date11. > @41 time time5. > @48 pnode_name $15. > @80 lmp 9.2 > @118 loss 8.2 > @127 cong 8.2 ; > format date date. time time.; > output; > end; > run; > > data new_lmps_2; > set new_lmps; > if _n_ = 1 then set new_lmps_id; > run; > > > Thank you, > > Sarah Whittier > ISO-NE