將行添加到使用前一行的當前密鑰

Question

我知道標題有點難以理解，我很抱歉。找不到更好的方式來解釋我的問題。

我有這樣的事情：

DATA HAVE; 
INPUT POLID PRODID TOSNUM PREMID X_INSURER_SERIAL_NO; 
DATALINES; 
13102100206 310 0 20 . 
13102100207 310 0 20 . 
13102100207 310 0 21 . 
13102100207 310 1 1 . 
13102100207 310 1 2 . 
13102100207 310 1 3 . 
13102100207 310 1 20 . 
13102100207 310 1 21 . 
13102100207 310 2 1 . 
13102100207 310 2 2 . 
13102100207 310 2 20 . 
13102100207 310 2 21 . 
13102100207 310 3 1 . 
13102100207 310 3 2 . 
13102100207 310 3 9 . 
13102100207 310 3 20 . 
13102100207 310 3 21 . 
13102100209 310 500 20 5 
13102100210 310 100 20 1 
13102100210 310 101 21 1 
13102100210 310 100 1 1 
13102100210 310 101 2 1 
13102100210 310 101 3 1 
13102100211 310 100 20 1 
13102100211 310 101 21 1 
13102100211 310 200 1 2 
13102100211 310 201 2 2 
13102100211 310 201 3 2 
;RUN; 

我需要做的是，

附加premid如果丟失當前行中，但上一行存在了每polid +的產品編號。

例如， polid = 13102100207，PRODID = 310

當tosnum = 1，I有5種不同的premid（1,2,3,20,21）。 當tosnum = 2時，我有4個不同的前體（1,2,20,21）。 premid = 3缺失。

我需要添加此前貼圖，並在fictive_ind列中添加1，每當添加前貼圖時。

polid = 13102100206，例如， 只有一行，並且因爲它沒有以前的tosnum，所以排除在計算之外。

當X_INSURER_SERIAL_NO沒有丟失時，它應該替換tosnum。

所以， 例如，polid = 13102100210 sholud從計算， 已排除以及它不具有任何其他先前的值（x_insurer_serial_no總是等於1）。

這是我做了什麼：

首先，更換時x_insurer_serial_no不缺少tosnum：

PROC SQL; 
CREATE TABLE AAA 
AS 
SELECT DISTINCT POLID, PRODID , 
CASE WHEN X_INSURER_SERIAL_NO = . THEN TOSNUM ELSE X_INSURER_SERIAL_NO END as X_INSURER_SERIAL_NO 
FROM HAVE 
;QUIT; 

然後，

我用下面的代碼，如果polid找出包括或不包括sholud：

PROC SQL; 
CREATE TABLE BBB AS 
SELECT POLID, PRODID, COUNT(*) AS COUNT 
FROM AAA 
GROUP BY 
POLID, PRODID 
;QUIT; 

我只考慮count> 1的計算行。

PROC SQL; 
CREATE TABLE ABC AS 
SELECT * FROM HAVE 
WHERE 
CATS(POLID,PRODID) IN 
SELECT CATS(POLID,PRODID) FROM BBB 
WHERE COUNT>1; 
QUIT; 

現在我發現最小值和最大值爲循環：

proc sql; 
select min(TOSNUM), max(TOSNUM) , max(TOSNUM) - min(TOSNUM) into: min_TOSNUM, :max_TOSNUM, :diff 
from abc; 
quit; 

然後循環：

％宏fictive_premid（min_TOSNUM =，DIFF =）;

proc sql; 創建表fictive_premid 作爲選擇POLID，PRODID，TOSNUM，PREMID，0作爲FICTIVE_IND 從ABC 其中TOSNUM = & min_TOSNUM; quit;

%do i = &min_TOSNUM %to %eval(&diff-1) ; 
    proc sql; 
    create table fictive_premid_t as 
    select * from 
    (
    select polid, prodid, %eval(&i+1) as TOSNUM, premid, 1 as FICTIVE_IND 
    from fictive_premid 
    where TOSNUM = &i 
    except 
    select polid, prodid, %eval(&i+1) as TOSNUM , premid, 1 as FICTIVE_IND 
    from abc 
    where TOSNUM = %eval(&i+1) 
    ) 
    union 
    select polid, prodid,%eval(&i+1) as TOSNUM, premid, 0 as FICTIVE_IND 
    from abc 
    where TOSNUM = %eval(&i+1); 
    quit; 

    proc append base=fictive_premid data=fictive_premid_t; run; 

%end; 


by polid prodid TOSNUM premid 
%mend; 


%fictive_premid (min_TOSNUM = &min_TOSNUM, diff = &diff); 

現在，

我的問題begings，例如，當 的&值i = 4 和polid = 13102100207 ....

去年tosnum等於3， 我得到6個新的行，其中tosnum = 4，fictive_ind = 1。 這發生了，當然，直到&我達到& diff的值。

什麼，我問你，請

是一種方式，當到達最後tosnum或x_insurer_serial_no停止itterating。

非常感謝您的幫助。

Hopefuly我解釋說我的需求夠好。

Answer 1

如果我理解你的問題，我認爲這是你想要的。這依賴於TOSNUM值是增量的（即1,2,3或4,5,6）而沒有跳過。

/*Overwrite TOSNUM with X_INSURER_SERIAL_NO 
    if it exists*/ 
data have; 
set have; 
tosnum = coalesce(X_INSURER_SERIAL_NO,TOSNUM); 
run; 

proc sort data=have; 
by POLID PRODID TOSNUM PREMID; 
run; 

proc sql noprint; 
/*Find polid values to delete*/ 
create view dist as 
select distinct polid, 
     prodid, 
     tosnum 
    from have; 

create table toDelete as 
    select polid, 
      count(*) as count 
     from dist 
     group by polid 
     having count = 1; 

/*Delete single values*/ 
delete from have 
    where polid in 
     (select polid from toDelete); 

create table lastTosnum as 
select POLID, PRODID, max(TOSNUM)+1 as Tosnum 
    from have 
    group by POLID, PRODID; 
quit; 

/*Increment TOSNUM by 1*/ 
data nextVals; 
set have(drop=X_INSURER_SERIAL_NO); 
TOSNUM = tosnum + 1; 
run; 

/*Delete the last TOSNUM from the set above*/ 
proc sql noprint; 
delete from nextVals as nv 
    where nv.tosnum in 
     (select tosnum 
      from lastTosnum as lts 
      where lts.polid = nv.polid 
       and lts.prodid = nv.prodid 
     ); 
quit; 

/*Merge to get the final set.*/ 
data want; 
merge have (in=h) nextVals (in=next); 
by POLID PRODID TOSNUM PREMID; 

if next and ^h then 
    fictive_ind = 1; 
else 
    fictive_ind = 0; 
/*decrement tosnum*/ 
tosnum = tosnum-1; 

run;