topic Re: unwanted pos changed in SAS Programming

Dataset corrupted after simple copy

Adambo — Wed, 17 Aug 2016 18:47:12 GMT

When changing the name of a dataset using a datastep to further manipulate the dataset, I see that some of the variables pos have changed. This results in some of the values shifting to the wrong variables.

How can I prevent changes other than the dataset's name when performing the following steps?

DATA want;

SET have;

RUN;

Thanks,

Adam

Re: unwanted pos changed

PGStats — Wed, 17 Aug 2016 17:24:58 GMT

In my experience with SAS, what you describe does not happen. Please post a reproducible example.

Re: unwanted pos changed

Adambo — Wed, 17 Aug 2016 18:00:55 GMT

Here is a screen shot of the dataset before and after the datastep. A value seems to be appear in the first collumn. Below are the output of the Proc Contents for the two datasets. Note how the position changes for some variables.

LAB1 Proc Contents

BIOMEDIC Proc Contents:

Re: unwanted pos changed

PGStats — Wed, 17 Aug 2016 18:23:31 GMT

Very confusing! Nothing matches after a simple copy operation? What is your theory as to what might have happened?

Re: unwanted pos changed

Adambo — Wed, 17 Aug 2016 18:32:48 GMT

My original file might be carrying some value from previous manipulations (ex, Retain), hidden values, or be corrupted. I have no idea. Thanks for confirming this is abnormal.

Re: unwanted pos changed

Reeza — Wed, 17 Aug 2016 19:18:33 GMT

@Adambo wrote:

My original file might be carrying some value from previous manipulations (ex, Retain), hidden values, or be corrupted. I have no idea. Thanks for confirming this is abnormal.

That doesn't make sense.

If the code is as shown, with no retain or any other process in between the position in the buffer shouldn't change. Note that your variable order is staying the same.

Can you post the code that causes including the proc contents before and after?

If you can replicate it I would contact SAS Tech Support.

Not sure what version your on, but In SAS 9.4 POS isn't even included in the output from proc contents, there's a NPOS though.

Re: unwanted pos changed

Shmuel — Wed, 17 Aug 2016 19:20:57 GMT

Please run the compare procedure:

Proc compare data=LAB1 compare=BIOMEDIC;

ID I_LABIM_KITID;

run;

Look at the output of this procedure. Are there unequal values ?

If you don't understand the output report, post it and I shall try to clarify it.

You should not be aware of changes in posions of variables, as long as all observations are equal.

Shmuel

Re: unwanted pos changed

Adambo — Wed, 17 Aug 2016 19:31:12 GMT

Someone else created the file i'm working on. I don't know what specific manipulations were done before it got to me.

Re: unwanted pos changed

Reeza — Wed, 17 Aug 2016 19:38:16 GMT

@Adambo wrote:

Someone else created the file i'm working on. I don't know what specific manipulations were done before it got to me.

Those don't matter, they're not maintained with the dataset.

What matters is what you're doing. Are you seeing a note in log? About CEDA access perhaps?

Please post the code and log.

Re: unwanted pos changed

Adambo — Wed, 17 Aug 2016 19:39:54 GMT

Re: unwanted pos changed

Adambo — Wed, 17 Aug 2016 19:43:23 GMT

6635      DATA SAS.BIOMEDIC;
6636      SET SAS.LAB1;
6637      RUN;

NOTE: 51588 observations were read from "SAS.LAB1"
NOTE: Data set "SAS.BIOMEDIC" has 51588 observation(s) and 32 variable(s)
NOTE: The data step took :
real time : 0.118
cpu time : 0.109

6638 quit; run;
6639 ODS _ALL_ CLOSE;

Re: unwanted pos changed

Reeza — Wed, 17 Aug 2016 19:46:28 GMT

POS -> Position in Buffer

NUMBER -> Position in Dataset

This results in some of the values shifting to the wrong variables.

According to Proc Compare output all variables and values are identical.

Personally, I don't see an issue. If you want to understand why the location in the buffer has changed it could be as simple as you're writing the data to a different system. I would consider asking tech support for an answer if interested, but given everything provided the change in POS should not impact your work.

Re: unwanted pos changed

SASKiwi — Wed, 17 Aug 2016 19:53:09 GMT

The PROC COMPARE proves that the 2 datasets are identical as far as column names and values are concerned which is what you would expect. It doesn't explain the change in column positions.

Re: unwanted pos changed

Shmuel — Wed, 17 Aug 2016 19:53:59 GMT

Clarification of results:

You got the note thal all observations are duplicate, that is because I_LABIM_KITID is not a unique key.

Change the ID line and add the other key varables in a format of: ID key1 key2 key3 ... ;

You got also the message "Nomber of Observations with Some Compared Variables Unequal = 0" and

"Nomber of Observations with All Compared Variables Equal = 51588" which is the whole dataset.

Those 2 messages are the proof that the 2 datasets are equal by values.

Regards, Shmuel

Re: unwanted pos changed

Adambo — Wed, 17 Aug 2016 20:28:33 GMT

I tried Proc Compare with different KeyIDs. I thought this would not work as the KeyID values are different in the two datasets (as displayed to me) but it appears SAS still sees them as the same regardless.

So I reckon the two datasets are "technically" the same. So I guess this is something related to the display and my system.

I reckon the two datasets are "technically" the same. So I guess this is something related to the display provided by my system.

Re: unwanted pos changed

Shmuel — Wed, 17 Aug 2016 20:50:31 GMT

In oredr to eliminate duplicates one need know the meaning of each variable and

is there a unique keys combination and what variables create it.

I suggest, try to find/guess the minimal number of variables which give the combination of a unique key

so you don'n get duplicates on proc compare report.

Shmuel

Re: unwanted pos changed

Reeza — Thu, 18 Aug 2016 00:04:04 GMT

@Adambo What do you mean by related to your display? You shouldn't be seeing any difference in your data. If you are that would be an issue.

Your pictures res earlier were not clear enough to view.

Re: unwanted pos changed

Adambo — Thu, 18 Aug 2016 14:25:16 GMT

I hope this one is clearer.

Note how values appear in the leftmost collumn and all other variable values seem to shift left. This does not happen to all records, only the first 19.

Re: unwanted pos changed

Reeza — Thu, 18 Aug 2016 14:31:52 GMT

My guess is your row order isn't actually the same between the datasets.

It does look like a bad import from a text file to me though.

I would contact tech support unless your able to actually post datasets.

Beyond that I don't think there's anything we ca help with anymore.

Re: unwanted pos changed

Adambo — Thu, 18 Aug 2016 15:29:16 GMT

OK,

Thank you all for your insights.