The code which Louis posted seems to come pretty close. I don't think it reassembles the title fragments however.

A one-pass solution does seem possible if we have an upper bound on the number of authors.

Aside: yesterday I suggested that Paula probably did not need a RETAIN statement in a DATA step featuring use of trailing "@" signs because she did not seem to be inheriting or accumulating data from record to record. Here is a contrasting example, since (1) the "AU" and "TI" flags implicitly apply to follow-on records and (2) the title fragments must be accumulated.

However, I suspect that the data volume here is pretty small, so I'm going to abandon concern with efficiency in favor of an approach which disentangles the problems at hand.

First I should mention that I took Lynn's illustrative input records and loaded them into a couple of text files with extension "aty" in my C:\SAS directory. The purpose was to allow a simple wildcard-based INFILE. If the real problem does not support this approach it may be necessary to enumerate the files in a list and use FILEVAR= to sweep through them.

Anyway, my first objective was to simply fill in the implicit "AU" and "TI" indicators and also assign sequential ID numbers to the items:

data filldown; infile '*.aty'; input rectype $ 1-2 rest:&$50.; select (rectype); when ('AU') idnum ++1; when (' ') rectype = remember; otherwise; end; * Read following 3 statements aloud 3 times quickly :-) ; remember = rectype; retain remember; run;

Now everything is in a SAS data set designed to support WHERE filtering and BY processing.

PROC TRANSPOSE can directly build the author arrays:

proc transpose data=filldown(where=(rectype='AU')) out=au(drop=_name_) prefix=au; by idnum; var rest; run;

A DATA step with an explicit loop simplifies housekeeping for the process of reassembling the titles:

data ti; keep idnum title; length title $ 200; do until (last.idnum); set filldown(where = (rectype='TI')); by idnum; title = left(trim(title)||' '||rest); end; run;

The years should be the simplest, just a character-to-numeric conversion. But presumably there should be just one year record per item, so it's good to check for exceptions:

data yr; set filldown(where = (rectype='YR')); by idnum; keep idnum yr; if not first.idnum then do; put 'Unexpected follow-on to YR record' / _all_; delete; end; yr = input(rest,4.); run;

Another kind of exception is a record of unknown type. A separate step can do a check:

data _null_; set filldown; where not (rectype in ('AU','TI','YR')); put 'Unknown record type' / _all_; run;

Finally, combine the components:

data autiyr; merge au ti yr; by idnum; run;

Results:

IDNUM AU1 AU2 AU3 TITLE YR

1 Person1 person2 person3 What I know about SAS and procs 2000 2 Person1 What I don't know about SAS and procs 1999 On Thu, 6 Jul 2000 17:54:32 +0200, Laproi, L.G.E (Louis) <Louis.Laproi@REAAL.NL> wrote:

>HI, > >Would this source do? > > >cull > >*=========================================================; > >DATA xxx; > INPUT label $ 1-2 > value $ 4-30; > > RETAIN booknum recnum lastlabl; > > * Recnum = seq number within label; > * booknum = seq number of the book; > * lastlabl = last known label; > > > * First record?; > IF _N_ EQ 1 THEN > DO; > booknum = 1; > END; > > * make sure you have labels; > IF label EQ ' ' THEN > DO; > * Use the last known label; > label = lastlabl; > recnum = recnum + 1; > END; > ELSE > DO; > * Remember this label for future use; > lastlabl = label; > recnum = 1; > END; > > * Make new label; > nlabel = label || PUT(recnum,z2.); > > * After every book add 1 to book number; > IF label EQ 'YR' THEN > DO; > OUTPUT; > booknum = booknum + 1; > END; > ELSE > OUTPUT; > >CARDS; >AU Person1 > person2 > person3 >TI What I know about SAS > and procs >YR 2000 >AU Person1 >TI What I don't know about SAS > and procs >YR 1999 >; >RUN; > > >PROC PRINT; > RUN; > > > > >PROC TRANSPOSE DATA=xxx OUT=yyy ; > BY booknum; > ID nlabel; > VAR value; > RUN; > >PROC PRINT; > RUN; > >-----Oorspronkelijk bericht----- >Van: Foster-Johnson [mailto:Foster-Johnson@DARTMOUTH.EDU] >Verzonden: donderdag 6 juli 2000 16:52 >Aan: SAS-L@LISTSERV.UGA.EDU >Onderwerp: file input statement > > >I'm trying to input a series of txt files (w/ hard returns) >into a SAS dataset; > >the files are structured as follows > >AU Person1 > person2 > person3 >TI What I know about SAS > and procs >YR 2000 >. >. >. >AU Person1 >TI What I don't know about SAS > and procs >YR 1999 > > >The files may have 1 to xx number of authors, so I'm trying >to create code that will read in values where appropriate >and leave blank otherwise. > >I have written some incredibly inelegant code using a >combination of lags and conditional statements on the label >var (AU, TI) to read in au1-au5 (for example see below) and >the title. > >label au1 au2 au3 title1 >AU p1 > p2 > p3 >TI what I know.. >AU p1 >TI What I don't > >My difficulty is this: when I use a retain statement (so I >can output the final record of each set of au-yr data), the >records which only have 1 author retain values from previous >records, thereby keeping values for other authors. I would >like to have the code to reset au1-au5 to a blank value each >time the label value equals a new au. > >Also, if anyone has any suggestions for a more efficient >(and elegant) method to read in these vars.. I'd appreciate >input-- > >Thanks in advance > >Lynn Foster-Johnson