PROC DATASETS set var mixing up variables

[sassy_seb]

Hello, really new to SAS. I am updating a table full of variables using PROC DATASETS and XATTR SET VAR.

The first statements worked fine, and I added the column with the proper values. However, as I added more statements, I noticed that some of the values got swapped around between the variables. 

For example:

AGE_40 (Values="39:Below 40yo; 4099(40-99yo)")
AGE_45 (Values="44:Below 45yo; 4599(45-99yo)")
AGE_55 (Values="54:Below 55yo; 5599(55-99yo)")

get swapped with:

AGE_0184262 (Values="1841(18-41yo); 4261(42-61yo); 6299(62-99yo)")

AGE_0184565 (Values="17(10-17yo);1844(18-44yo); 4564(45-64yo); 6599(65-99yo)")

AGE_0204565 (Values="19(10-19yo);2044(20-44yo); 4564(45-64yo); 6599(65-99yo)")

So now AGE_40 has the values of AGE_0184262.

I am not getting any errors or warnings in the log. Any help would be greatly appreciated.

[Reeza]

Can you show more info on how you're applying these statements to your datasets? Any possibility of missing a semicolon somewhere?

[sassy_seb]

DATA CHT_ALL;
	SET SAS_CHT.CHT_MERGEALLNOQC;
RUN;

PROC DATASETS LIBRARY=WORK;
	MODIFY CHT_ALL;
	XATTR SET VAR
	A1C65PLUSAGE60 (Values="Missing")
		AGE184560_DM (Values="Missing")
		AGE1856_DM (Values="Missing")
		AGE_40 (Values="39:Below 40yo; 4099(40-99yo)")
		AGE_45 (Values="44:Below 45yo; 4599(45-99yo)")
		AGE_55 (Values="54:Below 55yo; 5599(55-99yo)")
		AGE_1851 (Values="1850(18-50yo);5099(51-99yo)")
		AGE_1856 (Values="1855(18-55yo); 5699(56-99yo)")
		AGE_1860 (Values="1859(18-59yo); 6099(60-99yo)")
		AGE_1865 (Values="1864(18-64yo); 6099(65-99yo)")
		AGE_183660 (Values="0(0-17yo); 1(18-35yo); 2(36-59yo); 3(60+yo)")
		AGE_0184262 (Values="1841(18-41yo); 4261(42-61yo); 6299(62-99yo)")
		AGE_184560 (Values="1844(18-44yo); 4559(45-59yo); 6099(60-99yo)")
		AGE_184565 (Values="1844(18-44yo); 4564(45-64yo); 6599(65-99yo)")
		AGE_0184565 (Values="17(10-17yo);1844(18-44yo); 4564(45-64yo); 6599(65-99yo)")
		AGE_204565 (Values="2044(20-44yo); 4564(45-64yo); 6599(65-99yo)")
		AGE_0204565 (Values="19(10-19yo);2044(20-44yo); 4564(45-64yo); 6599(65-99yo)")
		AGE_18355065 (Values="2035(20-35yo); 3650(36-50yo); 5160(51-60yo); 6169(61-69yo)")
		AGE_18516065 (Values="1850(18-50yo);5159(51-59yo); 6064(60-64yo); 6599(65-99yo)")
		AGE_0_183560 (Values="0(0-17yo); 1(18-35yo); 2(36-59yo); 3(60+yo)")
		AGE_0_204060 (Values="19(0-19yo); 2039(20-39yo); 4059(40-59yo); 6099(60-99yo)")
		AGE_10Y_080p (Values="1019(10-19yo);2029(20-29yo);3039(30-39yo);4049(40-49yo);5059(50-59yo);6069(60-69yo);7079(70-79yo);8089(80-99yo);")
		AGE_10Y_35to65 (Values="0(0-19yo); 20(20-29yo); 30(30-34yo); 35(35-39yo); 40(40-49yo); 50(50-59yo); 60(60-65yo); 66(66+yo)")
		AGE_10Y_70P (Values="0(0-17yo); 1(18-29yo); 2(30-39yo); 3(40-49yo); 4(50-59yo); 5(60-69yo); 70(70+yo)")
		AGE_10Y_80m (Values="0(0-17yo); 1(18-29yo); 2(30-39yo); 3(40-49yo); 4(50-59yo); 5(60-69yo); 6(70-79yo); 80(80+yo)")
		AGE_10Y_80p (Values="0(0-17yo); 1(18-29yo); 2(30-39yo); 3(40-49yo); 4(50-59yo); 5(60-69yo); 6(70-79yo); 80(80+yo)")
		AGE_18e5160 (Values="1850(18-50yo); 6099(60-99yo)")

		ALC_12M (Values="1(Daily);2(2 to 6 times/wk);3(Weekly);4(2 to 3 times/mth);5(Monthly)");
RUN;

QUIT;

It's a lot but here's what I have. Like I said, the first couple of lines worked fine. 

Also provided some of the output as you can see how it swaps the position of the variables. 

[Reeza]

---

@sassy_seb wrote:

Also provided some of the output as you can see how it swaps the position of the variables. 

---

Can you clarify how this is occurring? Can you show a screenshot of before and after? I find it hard to believe that proc datasets would do that somehow. 

[Tom]

If you want PROC CONTENTS to display the variables in position order (instead of alphabetical order) then add the VARNUM option to the PROC statement.

proc contents data=sashelp.class varnum;
run;

[Tom]

Why are you using XATTR SET VAR?

What is it that you are trying to do?

The listing you posted the photograph of is just of the normal variable attributes, like the NAME, TYPE, LENGTH, LABEL and attached FORMAT.  If you want to see extended attributes you will have to look somewhere else.

[sassy_seb]

I am using it to add an attribute to each variable. In this case, adding an attribute named Values and adding the valid values for each variable. Overall, I am trying to create a data dictionary, however all of the variables are missing their values.

[Tom]

Your code style works fine for me.  Perhaps you are just looking at the wrong part of the output of PROC CONTENTS?

data class;
  set sashelp.class;
run;

PROC DATASETS LIBRARY=WORK nolist;
	MODIFY class;
	XATTR SET VAR
	NAME (Values="Name - Missing")
  SEX (Values='SEX - Not Missing')
  ;
run;
quit;

proc contents data=class ;
run;

[sassy_seb]

I think I am  merging the data incorrectly. The values I added are displayed properly, it is actually the labels that are getting mixed around.

[Reeza · Accepted solution]

How are you merging? By Variable name or variable number?

[sassy_seb]

Oh! This was totally the issue. I had commented out the "by variableName" by mistake

[Tom]

The variable labels in your first photograph look fine to me.

I do not understand where you are merging anything. 

Are you trying to get the output of proc contents into a dataset?  What code did you use?  How did you attempt to merge the extended attribute values back to the normal variable attributes?

[ChrisHemedinger]

You can also save the extended attributes to another data set as you go, maybe making it easier to review. Adapting the example from Tom:

data class;
  set sashelp.class;
run;

PROC DATASETS LIBRARY=WORK nolist;
	MODIFY class;
	XATTR SET VAR
	NAME (Values="Name - Missing")
  SEX (Values='SEX - Not Missing')
  ;

 /* save these extended attributes for review */
 ods output ExtendedAttributesVar = work.savedAttrs;
 contents data=class;
quit;

After this, work.SAVEDATTRS will be a data set that contains your new attributes.

[Reeza]

data class;
set sashelp.class;
run;

proc datasets lib=work;
modify class;
xattr set var sex(values='M:Male;F:Female');
run;quit;

proc sql;
create table lookup as select * from sashelp.vxattr where memname='CLASS' and libname='WORK';
quit;

FYI - if you have formats created you could do this programmatically, by listing out the format values instead of manually listing the values. 

Just a suggestion. If you use data dictionary to find all the custom formats and then use proc format to list the formatted values is possible to create  a custom data dictionary. 

[Reeza]

Would this at all be useful for you?

https://communities.sas.com/t5/SAS-Communities-Library/Library-Datasets-Summary-Macro-DATA-SPECS/ta-...