The following code does what you asked for. data _null_; have='yyy y www qqq '; have=tranwrd(have,'y y','yy'); put have=; want=scan(have,1); put want=; run; More general solutions would be possible - especially with RegEx - but you would have to tell us a bit more in detail what the problem is about you want to solve. ... View more

You define delimiters the way you did it. You defined '3' as delimiter. BUT: Your data doesn't look as if the digit '3' is the separator between you raw data fields. You are reading the data using fixed column style input. I.e. you read the data for class always from column 1-6. Using columns style you don't need to define a delimiter at all (look it up in the Online Doc under INPUT Statement). To read the data as given in your example, try something like this: %let myfile=datalines; data WORK.want; infile &myfile firstobs=2; input Class:$6. Age:$6. Sex:$7. Survived:$4.; put _all_; datalines; Class Age Sex Survived First Adult Male Yes First Adult Male Yes First Adult Male Yes ; HTH Patrick ... View more

I assume that there is some confusion between the metadata definitions and the physical dataset/library path. ...metadata server and went to open up the permanent dataset that i attempted to create. I opened it via base SAS, and the dataset opened fine..." Does that mean you've run a base sas session on the physical box where the sas metadata server runs as well? Because: This way you accessed the physical sas file directly without using the metadata definitions. It doesn't matter that the metadata server is also running on this box. I assume that in some runs you use a libname statement (without meta) to access the data and in other cases you try to access the data via metadata definitions. Is the metadata definition for the library pre-assigned or not? Or how do you assign the library? I would first check the metadata definitions: Is the table you want to write to defined correctly (all the fields mapped)? Is the table part of the library you use? Does the library point to the expected physical location? It could be that you write your data to one location (using a libname statement without meta keyword) but that your looking somewhere else for this file when you're using the metadata definitions. I don't yet understand how the EG log can tell you that it wrote 0 columns and there are still all the data written - are you sure you're not looking at old data? But first: Check the metadata definitions. HTH Patrick ... View more

Hi yy1234 Something like that: data _null_; infile datalines missover; input String $50.; put String=; array StringPart {100} $100.; i=1; do while(scan(String,i,') /') ne ''); StringPart{i}=scan(String,i,') /'); StringPart{i}=compress(StringPart{i},'('); put StringPart{i}=; i+1; end; datalines; (F1) 172/233 (M20) 141 (M279) 256 (F125) 9/87 M(603)90 F(666)10/170 M(850)84 F(832)9/358 ; You could also first pass the data set and determine the necessary number of array elements before reading the "words" into separate variables. I've chosen 100 array elements. What could work to do it more dynamic is: length() of string - length() of string after compression of all delimiters - and then keeping the max value of this difference (call symput). Using this value (in macrovar) in array definition..... just an idea. And yes: There would also be some Perl RegEx which could be helpful - but I think in your case they wouldn't provide a simpler solution. Cheers Patrick ... View more

To 1: Just make sure to create test data to cover all possible cases - and then define the rules to build episodes before you even start with coding. To 2: Atomic is a term used in data organisation (not SAS specific). Out of a theory book: "Granularity refers to the level of detail represented by the values stored in a table's row. Data stored at their lowest level of granularity are said to be atomic data." Let us know the rules when you've got them and provide also the test data. Patrick ... View more

Hi bethcook That's a nice one! I see 2 issues here: 1. How to identify episodes. 2. How to organise the data. Question to 1: id fromdate todate 1 1 4 1 3 6 1 5 8 How many episodes would that be and which record would belong to which episode? Thoughts about 2: Your dataset can either be aggregated or on atomic level. To have both you can: A: have 2 datasets; generate a primary key for the aggregated dataset and add this key to the atomic dataset (the one you have already) as foreign key. B: create only one aggregated dataset and store all atomic information de-normalised into "arrays". C: keep the atomic dataset (number of rows) but add additional variables with aggregated values which are repeated (also "de-normalised"). I would go for variant A or C. Variant B will give you a lot of headache (and cause some macro coding). The number of variables to store the atomic information (=size of the array) would vary (maximum number of rows which make one episode). Let me know what you're heading for and answer question 1 please. Regards Patrick ... View more

You could address SAS Inst directly with your “best practice” question. The architecture using a citrix server together with SAS I have seen couldn’t compete with a “traditional” SAS-client/SAS-server architecture (and it was a fast WAN). All point-and-click operations had a tiny little bit of delay – it wasn’t a lot but users weren’t too happy with it. I assume this delays become much worse on a slow WAN. IT support was of course happy with citrix - the network guys didn’t like it too much. What about having the most common clients (i.e. SAS EG) as local installation and specialty software like DI on citrix? The point is: A lot of the work with SAS EG doesn’t cause any network traffic (most of the “code generation” work) and is therefore fast and smooth. The code is then sent to the server (that’s only a bit of text) and most of the results are kept on the server. It’s only when people look at results that they have to be download – so just make sure that the EG settings are in a way that result tables and the like are not opened automatically but only on user request. HTH Patrick ... View more

Hi Like some other SAS clients (i.e. SAS EG) SAS DI uses the object spawner to start a SAS session on the workspace server. There are several ways how to save your log - and it depends a bit in what environment you're working and what rights you have. http://support.sas.com/onlinedoc/913/getDoc/en/hostwin.hlp/win-sysop-altlog.htm Besides of proc printto there would also be the option ALTLOG (but you need access to the .cfg file). As this is processing jobs in DI I assume it is about unit testing: I would use the printto options - or just "manually" save the log from within DI. Another way would be to process the DI generated SAS code in batch - and all the logs are created as permanent files (you could even define where to save this log files). HTH Patrick ... View more

I just joined a project with a similar problem: A lot of observations and variables and performance problems (also a Star model). The solution was: Horizontal and vertical fragmentation. So: Split the biggest SAS tables in several tables and create views. For your case take a view with as little as possible fields (or access the appropriate SAS table directly). I'm aware that this is a design change as the ETL jobs have to be adopted - but it might be worth it if the problems persist. Regards Patrick ... View more

Only a partial answer: - You need a SAS OLAP Server to surface SAS OLAP cubes (you can build them without this server). - Instead of programing everything by yourself you could try to use MS Excel together with the SAS Add-in for MS Office (AMO). Create a pivot table and you can surface the SAS cubes. I don't have too much experience with .net but I believe to remember that I've seen Excel "objects" in Visual Studio. No idea how this works together with a SAS Add-in - but I would try to go down this road (with a lot of Internet research). Besides of AMO have also a look at the SAS OLAP viewer - there should be a possibility to implement one of the two as kind of portlet (or something similar in .net). There might be other and better ways how to do it - keep us informed if you learn more. HTH Patrick ... View more

Just in case you don't know this site already: http://robslink.com/SAS/Home.htm The examples there are often a great starting point for an own solution. Hope this helps at least a bit Patrick ... View more

Hi Myao You have to talk to your SAS admin so that he changes the SAS system option from NOXCMD to XCMD. http://support.sas.com/onlinedoc/913/getDoc/en/hostunx.hlp/a000663635.htm HTH Patrick ... View more

Hi Chris@SAS There were more posts in this tread! I think it is very disrespectful to just delete other peoples contributions only because one thinks he gave the one and final answers. The support page you point us to wants us to copy a dll into the „program file“ directory. I know a lot of organisations where a normal user wouldn’t be allowed to write to this directory. Regards Patrick ... View more

Hi David Like you I'm an experienced coder and a bit resistive to use EG. So this was a nice challenge. What worked was righ clicking the SAS table in the project explorer window and then use „Filter and Query...“. I „pointed-and-clicked“ the following code: %_eg_conditional_dropds(WORK.Query_for_HAVE); PROC SQL OUTOBS=50; CREATE TABLE WORK.Query_for_HAVE AS SELECT HAVE.i, (COUNT(HAVE.i)) AS COUNT_OF_i FROM WORK.HAVE AS HAVE GROUP BY HAVE.i ORDER BY COUNT_OF_i DESCENDING; QUIT; Regards Patrick ... View more