Complex SQL rename query

[Swapnil_21]

I have dataset with 200 columns. Out of that I need to replace values in original columns with modified columns which always start with  _M_.

E.g. out of 200 columns in datasets, Account_ID,  ACCOUNT_CLASSIFICATION_CODE ,BS_CENTRE_ID has corresponding columns as _M_ACCOUNT_ID,_M_ACCOUNT_CLASSIFICATION_CODE,_M_BS_CENTRE_ID. However, there could be any number of columns which has such corresponding _M_ pair

So we need to identify how many columns have such pairs and then replace values in original columns with modified columns. 

Below is the sample Have dataset 

MONTH_KEY	PERIOD_ID	ACCOUNT_ID	ACCOUNT_CLASSIFICATION_CODE	BILLING_CODE	PRODUCT_CODE	BS_CENTRE_ID	_M_ACCOUNT_ID	_M_ACCOUNT_CLASSIFICATION_CODE	_M_BS_CENTRE_ID
202109	20210930	CC 5163104001923574	65031	871	13334	161730	MAJ12345	11433	113
202109	20210930	CC 5163190010421759	66031	V05	13255	161730	MAJ87564	67845	345
202109	20210930	CC 5163230005429159	28051	G02	13253	161730	JAK89167	98365	765
202109	20210930	CC 5163202036863805	43151	V05	13249	161730	PPT11765	41897	897
202109	20210930	CC 4715726286102320	12011	105	13124	161730	RAM90145	34950	950
202109	20210930	CC 4564729003710930	95021	S02	13126	161730	TPT66387	12854	684

And below is what is expected in Want dataset

MONTH_KEY	PERIOD_ID	ACCOUNT_ID	ACCOUNT_CLASSIFICATION_CODE	BILLING_CODE	PRODUCT_CODE	BS_CENTRE_ID
202109	20210930	MAJ12345	11433	871	13334	113
202109	20210930	MAJ87564	67845	V05	13255	345
202109	20210930	JAK89167	98365	G02	13253	765
202109	20210930	PPT11765	41897	V05	13249	897
202109	20210930	RAM90145	34950	105	13124	950
202109	20210930	TPT66387	12854	S02	13126	684

I have tried using coalesce and also using rename , but somehow I am not able to find correct solution.

Any help is really appreciated

[Sajid01]

Hello
In the first para of your question para of your question you are requesting assistance in renaming columns.
Then you want assistance if finding columns to be renamed.
Can you please be more  clear?
Renaming columns is a simple process in SAS if that is what you are looking for. I have given two examples below.

proc sql;
create table class1 as
select Name as _M_Name,  Sex as _M_Sex ,  Age as _M_Age
 from sashelp.class;
quit;
data class2;
set sashelp.class(Rename=(Name=_M_Name Sex=_M_  Age=_M_Age) );
run;

[Swapnil_21]

So precisely my question is
1. Out of 200 column how do I know which columns start with _M_ and then remaining portion matches with actual column
2. When matches, rename them with original column.

Sample datasets provided earlier

[Sajid01]

Hello @Swapnil_21 
Once can find the columns of any data set using SAS dictionaty table or proc contents.
I am giveng and example below using sashelp.class. 
Here libname is the library  (SAS HELP) where the data set memname (CLASS in the example )resides.

proc sql;
select  name from dictionary.columns
where libname="SASHELP" and memname="CLASS" and memtype="DATA";
quit;

Thus once you have the column names, you can decide what columns to rename ans shown earlier.

[andreas_lds]

---

@Swapnil_21 wrote:
So precisely my question is
1. Out of 200 column how do I know which columns start with _M_ and then remaining portion matches with actual column
2. When matches, rename them with original column.

Sample datasets provided earlier

---

Your second step won't work without dropping the original variable first.

Before starting to solve such task, i recommend writing the code for one m-Variable, skipping any existence checks. The most obvious, but also not working solution is:

proc datasets nolist;
  modify your_dataset;
    drop old_var;
    rename _m_old_var = old_var;
quit;

Unfortunately proc datasets doesn't support the drop statement. Sad.

Dropping variable is possible using proc sql alter table, but the alter statement can't rename variables. Sad²

So you need both: proc sql to drop the unwanted variables and proc datasets to rename the m-variables. If the dataset is really small, you could use a data null step with call execute generating a proc  sql to creating a new dataset.

Will return later, as soon as i had the time to create something working.

[andreas_lds · Accepted solution]

Unfortunately you didn't provide data in usable form, so i used a modified version sashelp.class. The subsequent steps extract the information of interest.

data work.class;
   set sashelp.class;
   
   length _m_Age _m_Weight 8;
   
   _m_Age = Age + 10;
   _m_Weight = round(Weight * 1.5, 0.01);
run;

data work.Variables;
   set sashelp.vcolumn(where=(libname = 'WORK' and MemName = 'CLASS' and  not ('_m_' =: Name)));
      
   bob = _n_;
   
   keep Name bob;
run;

data work.M_Variables;
   set sashelp.vcolumn(where=(libname = 'WORK' and MemName = 'CLASS' and '_m_' =: Name));
   
   length mVar $ 32;
   
   mVar = Name;
   Name = substr(Name, 4);
   
   keep Name mVar;
run;

The first solution will not change your dataset, but will have a higher runtime, if the dataset is large.

data _null_;
   set work.Variables end= jobDone;
   
   if _n_ = 1 then do;
      if 0 then set work.M_Variables;
      
      declare hash h(dataset: 'work.M_Variables');
      h.defineKey('Name');
      h.defineData('mVar');
      h.defineDone();
      
      call execute('proc sql noprint; create table work.fixed as select');
   end;
   
   mVar = ' ';
   
   if _n_ > 1 then do;
      call execute(',');
   end;
   
   if h.find() = 0 then do;
      call execute(catx(' ', mVar, 'as', Name));
   end;
   else do;
      call execute(Name);
   end;
   
   if jobDone then do;
      call execute(' from work.class; quit;');
   end;
run;

And here's another solution modifying your dataset, because there is no undo, you should be 100% sure, that you understand what happens.

data _null_;
   set work.Variables end= jobDone;
   
   if _n_ = 1 then do;
      if 0 then set work.M_Variables;
      
      declare hash h(dataset: 'work.M_Variables');
      h.defineKey('Name');
      h.defineData('mVar');
      h.defineDone();
   end;
   
   length 
      dropList $ 6600 /* 200 Variables, max length 32 chars, +1 to hold the separator */
      renameList $ 14000  /* 66 chars for each pair oldname/newname incl. separators */
   ;
   
   retain dropList renameList;
   
   if h.find() = 0 then do;
      dropList = catx(',', dropList, Name);
      renameList = catx(' ', renameList, cats(mVar, '=', Name));
   end;

   if jobDone and not missing(dropList) then do;
      call execute(catx(' ', 'proc sql noprint; alter table work.Class drop', dropList, '; quit;'));
      call execute(catx(' ', 'proc datasets nolist library=work; modify Class; rename', renameList, ';quit;'));
   end;
run;