Conditionally delete observations and drop variable(s) from a given dataset.

%ds_delete(dsn, var=, cond=, obs=, firstobs=0, lib=WORK);

dsn : a dataset reference;
var : (option) List of variable(s) to delete (drop) from the dataset; if more variables are to be deleted, var should be defined as an unformatted list; default: not used;
cond : (option) an expression that resolves to a boolean (0/1) so that all observations for which cond is true (1) will be deleted; default: 0, i.e. no observations is deleted;
firstobs, obs : (option) indexes of the first and the last observations to consider for the delete operation resp.; all obsevation whose index is <firstobs or >obs will be automatically deleted; see DATA step options; by default, options are not used;
lib : (option) name of the input library; by default: empty, i.e. WORK is used.

The table dsn is updated with the conditional deletion.

Let us consider the following _dstest31 table:

we will delete all VALUEs and keep only observations for which UNIT is EUR:

%ds_delete(_dstest31, var=value, cond=%quote(unit="EUR"));

so that we have:

geo	unit
BG	NAC
FR	NAC

Note that the command can be used to delete more than one variable at a time, e.g.:

%ds_delete(_dstest31, var=value unit, cond=%quote(unit="EUR"));

will return instead:

geo
BG
FR

Run macro %_example_ds_delete for more examples.

In short, the macro sequentially runs two operations that summarizes to the following DATA step:

DATA &lib..&dsn (DROP=&var);
    SET &lib..&dsn(FIRSTOBS=&firstobs OBS=&obs);
    IF &cond THEN DELETE;
run;

It shall be noticed that in practice: first the options firstobs and obs are applied, then the condition cond is evaluated (though it occurs inside a unique DATA step), and then the variable var is dropped from the dataset. This matters in the cases where cond is an expression based on var values.