Brian,

Reverse the task and view the flat file as a driver and the SAS data set as a look-up table. The latter can take on a variety of forms. Say, you have a flat file and SAS data set concocted from sashelp.zipcode, keyed by ZIP and with about 1/3 of keys mismatched between the two:

filename flat temp ;

data sas (keep = zip state); set sashelp.zipcode ; file flat ; if ranuni (1) < .3 then fzip ++ .5 ; else fzip = zip ; put @1 fzip z7.1 @ 10 city ; run ; ------------------------------- NOTE: 41988 records were written to the file FLAT. NOTE: There were 41988 observations read from the data set SASHELP.ZIPCODE. NOTE: The data set WORK.SAS has 41988 observations and 2 variables.

Now if you would prefer to organize the look-up table as a hash, consider

data merged ; if _n_ = 1 then do ; if 0 then set sas ; dcl hash sas (dataset: 'sas') ; sas.definekey ('zip') ; sas.definedata ('state') ; sas.definedone ('state') ; end ;

infile flat truncover ; input @1 zip 9. @10 city $35. ; call missing (state) ;

_iorc_ = sas.find() ; * left join on flat file ; * if _iorc_ = 0 ; * equijoin ; run ;

If you need an equijoin (output only matching records), erase the asterisk. If you like SAS indices better, you can make the SAS file a self-searching look-up table in itself:

proc sql ; create index zip on sas (zip) ; quit ;

data merged ; infile flat truncover ; input @1 zip 9. @10 city $35. ; set sas key = zip / unique ; if _iorc_ ne 0 then do ; _error_ = 0 ; call missing (state) ; end ; * nothing needed ; * left join on flat ; *if _iorc_ = 0 ; * equijoin ; run ;

Once again, for equijoin, delete the last asterisk. If you need the output in the form of another flat file, just use _NULL_ instead of MERGED and a pair of FILE/PUT statements as dictated by the requirements and common sense.

Kind regards ----------- Paul Dorfman Jax, FL -----------

On Thu, 17 Jul 2008 11:59:00 -0700, Brian <brian8202@GMAIL.COM> wrote:

>Hello, > >I have a SAS dataset driver file that has 100 records. I need to >merge this dataset with another file which is a flat file. >Unfortunately this flat file has about 1 million records. One of the >ways to do it is reading every single record from the file file into a >SAS dataset then merge with the driver file, but this process is very >time consuming and inefficient. I am looking for a way that I am able >to merge the SAS dataset with the records on the flat file on the fly. >Thank you very much for your help. > > >Sincerely, > >Brian