如何更优雅命名SAS临时数据集？替代temp系列命名方案咨询

Great question! Numbered temp tables like temp1, temp2 get the job done, but they’re hard to follow as your code grows—especially if you come back to it weeks later. Let’s walk through cleaner, more maintainable alternatives for your SAS workflow:

1. Descriptive, Action-Oriented Names (Best for Readability) #

Instead of generic tempX, name each table based on what it does. This makes your code self-documenting—you’ll instantly know what each step accomplishes without reading the code itself.

Here’s how your code would look with this approach:

* Calculate sum of A, B, C and create base dataset;
data summed_abc;
    set test;
    A = sum(A, B, C);
run;

* Sort dataset by variable E;
proc sort data=summed_abc out=sorted_by_e;
    by E;
run;

* Retain co_date and update E to match M;
data retained_co_date;
    set sorted_by_e;
    retain co_date;
    E = M;
run;

* Aggregate metrics by group V;
proc sql;
    create table aggregated_by_v as
    select *, sum(h) as p, sum(a) as m, sum(r) as c
    from retained_co_date
    group by v;
quit;

* Filter to only records where B=1;
data filtered_b1;
    set aggregated_by_v;
    if b=1;
run;

2. Use SAS's _LAST_ System Variable (Quick Iterations) #

SAS automatically tracks the last created dataset with the _LAST_ keyword. This lets you reference the previous output without typing its name—perfect for quick scripting or when you don’t want to manage intermediate names.

Example with _LAST_:

data temp1; set test; A=sum(A,B,C); run;
proc sort data=_LAST_ out=temp_sorted; by E; run;
data temp_retained; set _LAST_; retain co_date; E=M; run;
proc sql; create table temp_aggregated as select * ,sum(h) as p ,sum(a) as m ,sum(r) as c from _LAST_ group by v; quit;
data final_dataset; set _LAST_; if b=1; run;

3. Minimize Intermediate Tables (Streamline Workflow) #

Whenever possible, combine steps to avoid creating unnecessary temp tables. For example, you can use proc sort’s out= parameter to feed directly into the next data step, or merge logic in proc sql with subqueries (note: retain works best in data steps, so you may still need one step for that):

* Combine sum and sort in a single pipeline;
data summed_abc;
    set test;
    A = sum(A,B,C);
run;

proc sort data=summed_abc out=sorted_by_e; by E; run;

* Retain co_date and filter in one step (if logic allows);
data final_data;
    set sorted_by_e;
    retain co_date;
    E = M;
    if b=1;
run;

* Aggregate in SQL using the filtered dataset;
proc sql;
    create table final_aggregated as
    select *, sum(h) as p, sum(a) as m, sum(r) as c
    from final_data
    group by v;
quit;

Final Recommendation #

For long-term code maintainability, descriptive names are the way to go. They make your code easier to debug, share, and revisit later. The _LAST_ variable is great for quick tests or throwaway scripts, but avoid it in production code where clarity matters most.

内容的提问来源于stack exchange，提问作者user1481397