个人中心
退出
SAS宏分析Log找出最费时间的步骤
我经常遇到的问题是想找出成百上千个DATA步和过程步中花费时间最长的步骤（对程序进行优化），但SAS输出的Log文件可读性不强，也不能从里面方便的找到我要的信息。
我写的这个宏是把Log文件里每一个步骤的名称、所在Log文件的行号、所花的实际时间和CPU时间提取出来，方便分析。也可以对日志进行分析找出错误信息和警告信息，这个我们留到以后再说。
Log文件大致是下面这个样子：
NOTE: Deleting WORK.SS_TEMP (memtype=DATA).
NOTE: PROCEDURE DELETE used (Total process time):
real time 0.00 seconds
cpu time 0.01 seconds
NOTE: The data set WORK.SS_TEMP has 1 observations and 1 variables.
NOTE: DATA statement used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
NOTE: There were 1 observations read from the data set WORK.SS_TEMP.
NOTE: The data set WORK.SS_TEMP has 1 observations and 2 variables.
NOTE: DATA statement used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
NOTE: PROCEDURE SQL used (Total process time):
real time 0.00 seconds
cpu time 0.00 seconds
此宏的作用是读取SAS系统生成的Log文件，并将得到的步骤名称和对应的运行时间保存至结果表格中。其中，InFilePath指定的Log文件，要写全从根目录至最后的文件名称；TargetTable是结果表格。最终得到的是原始Log文件中包含的步骤名称和对应的运行时间(实际时间和CPU时间)，其中N是步骤在原始Log文件中的行号。
%macro GetTimeTableForSASLog(InFilePath,TargetTable);
/**********************************************************************/
/* 检查InFilePath的合法性，后缀必须为TXT或LOG */
%if %SYSFUNC(FIND(&InFilePath,%STR(.))) EQ 0 OR (%SYSFUNC(FIND(&InFilePath,%STR(.))) NE 0 AND %UPCASE(%SCAN(&InFilePath,-1,%STR(.))) NE TXT AND %UPCASE(%SCAN(&InFilePath,-1,%STR(.))) NE LOG) %then
%do;
%put ERROR: The suffix of InFilePath should be txt or log, please check it again.;
%goto exit;
%end;
/* 检查TargetTable的存在性 */
%if &TargetTable EQ %STR() %then
%do;
%put ERROR: The TargetTable should not be blank, please check it again.;
%goto exit;
%end;
/* 开始进行计算 */
/* 第一步：导入LOG文件 */
proc import datafile="&InfilePath" out=&TargetTable replace;
getnames=no;
run;
/* 第二步：处理结果表格 */
data &TargetTable;
set &TargetTable;
N=_N_;
if SUBSTR(UPCASE(VAR1),1,10) EQ 'NOTE: DATA' OR SUBSTR(UPCASE(VAR1),1,15) EQ 'NOTE: PROCEDURE' OR
SUBSTR(UPCASE(VAR1),1,9) EQ 'REAL TIME' OR SUBSTR(UPCASE(VAR1),1,8) EQ 'CPU TIME';
run;
data &TargetTable;
set &TargetTable;
if SUBSTR(UPCASE(VAR1),1,5) EQ 'NOTE:' then
do;
Name=SUBSTR(VAR1,7,FIND(UPCASE(VAR1),'USED')-8);
end;
else if SUBSTR(UPCASE(VAR1),1,9) EQ 'REAL TIME' then
do;
Real_Time=SUBSTR(VAR1,PRXMATCH('/\d+.\d+/',VAR1));
end;
else if SUBSTR(UPCASE(VAR1),1,9) EQ 'CPU TIME' then
do;
CPU_Time=SUBSTR(VAR1,PRXMATCH('/\d+.\d+/',VAR1));
end;
run;
data &TargetTable(keep=Col_N Col_Name Col_Content);
set &TargetTable;
retain Col_N;
length Col_Name $50;
length Col_Content $100;
if Name NE '' then
do;
Col_N=N;
Col_Name='Name';
Col_Content=Name;
output;
end;
else if Real_Time NE '' then
do;
Col_Name='Real_Time_Temp';
Col_Content=Real_Time;
output;
end;
else if CPU_Time NE '' then
do;
Col_Name='CPU_Time_Temp';
Col_Content=CPU_Time;
output;
end;
run;
proc transpose data=&TargetTable out=&TargetTable;
by Col_N;
id Col_Name;
var Col_Content;
run;
/* 第三步：处理结果表格中的时间变量 */
data &TargetTable(keep=N Name Real_Time CPU_Time);
retain N Name Real_Time CPU_Time;
set &TargetTable(rename=(Col_N=N));
format Real_Time TIME10.4;
format CPU_Time TIME10.4;
if FIND(UPCASE(Real_Time_Temp),'SECONDS') NE 0 then
Real_Time=HMS(0,0,SCAN(Real_Time_Temp,1,' '));
else if LENGTH(COMPRESS(Real_Time_Temp,'.','d')) EQ 2 then
Real_Time=HMS(SCAN(Real_Time_Temp,1,':'),SCAN(Real_Time_Temp,2,':'),SCAN(Real_Time_Temp,3,':'));
else if LENGTH(COMPRESS(Real_Time_Temp,'.','d')) EQ 1 then
Real_Time=HMS(0,SCAN(Real_Time_Temp,1,':'),SCAN(Real_Time_Temp,2,':'));
if FIND(UPCASE(CPU_Time_Temp),'SECONDS') NE 0 then
CPU_Time=HMS(0,0,SCAN(CPU_Time_Temp,1,' '));
else if LENGTH(COMPRESS(CPU_Time_Temp,'.','d')) EQ 2 then
CPU_Time=HMS(SCAN(CPU_Time_Temp,1,':'),SCAN(CPU_Time_Temp,2,':'),SCAN(CPU_Time_Temp,3,':'));
else if LENGTH(COMPRESS(CPU_Time_Temp,'.','d')) EQ 1 then
CPU_Time=HMS(0,SCAN(CPU_Time_Temp,1,':'),SCAN(CPU_Time_Temp,2,':'));
run;
%exit:
%mend;
分类导航