nodupkey and SQL

[yudhishtirb]

Hello Team,

I am looking to convert below statatement of SAS in SQL 

proc sort data=strat.master_query nodupkey out=strat.master_query_unq; by wclientcode szaccountnumber raisedmth; run;

Is it like DISTINCT in SQL ?

I tried with 

SELECT DISTINCT * FROM master_query SORT BY wclientcode szaccountnumber raisedmth

But it is not giving correct reocrds with SQL statement.

Can someone please help me ?

Thanks In advance

[Kurt_Bremser]

If you do sort with nodupkey, only one record per by values is kept; if additional variables in the dataset have different values between rows, only one of those values is therefore kept, the others are discarded.

Select distinct, OTOH, looks at all variables in the select list (which happen to be all variables in the dataset when using *), and keeps multiple records per "order by" group if there are non-unique values in the other columns.

IMO, if a SQL distinct creates something different than a sort nodupkey, this is a sign for sloppy programming, as values may be lost with the sort.

Look if all columns from your input dataset are needed; a proper keep in the sort step (that shows what columns are significant and makes for better understanding of the code) and a proper select with only the needed columns might be the right thing to do.

[RW9]

No, they are not the same.

by wclientcode szaccountnumber raisedmth;

This orders the SAS data by the variables given in the by statement, then by _n_ if there are more than one.  Nodupkey will take the first record per that by grouping to output.

Distinct will select a distinct value based on the sort order defined internally, which will consider all variables as * means all variables.  Note that in SQL the syntax is:

order by <variable>{, <variable};

The question is, why do you need to change SAS code which you have working for SQL code which you don't know?

[PGStats]

If you want to see the same number of records as produced by sort nodupkey, try

SELECT DISTINCT wclientcode, szaccountnumber, raisedmth
FROM master_query;

If you want to see the same number of records as your query, try

proc sort data=master_query nodupkey; by _all_; run;

[LinusH]

If you got working code, why change it?
If you are doing a migration from SAS then you are better off asking this question in forum appropriate for your target environment.