08-21-2014 06:43 AM
I have to move a data source from one folder to another.
But there are a considerable number for reports and explorations already using it.
What is the easiest way to do so?
08-21-2014 07:25 AM
The easiest: Not having the physical names having coded anywhere.
That should been done as a kind of autoexec_ application initialization managed separately.
The advantage is changing something in that central managed setting is effective for all exiting code.
A little bit old_fashioned quality assurance version-management releasemanagement.
Old cobol programmers were educated and trained that way, no program/code being accepted with hard coded physical names.
The JCL (z/OS) scripting is required to solve that.
As you have an operational environment, why aren't you doing that in a similar way?
08-22-2014 03:00 AM
Depends on what you classify as easy...
You may already be familiar with the Change the Data Source option that is available in Visual Analytics Explorer - Change Data Source - User Guide or alternatively you could move the table and then next time you open the exploration you will be prompted to update your data source. For your reports, you can use the Repair Reports Window - User's Guide to fix them.
While this approach is manual it is easy in the sense that it is done via point and click using the Visual Analytics interface. The alternative might be to update the metadata and that would require knowledge of the SAS metadata APIs and development time.
08-22-2014 03:11 AM
Yes i am familiar with the Change Data Source option and the Repair Reports Prompt.
But I was looking for at way around those options.
08-21-2014 09:19 AM
Patching after the fact is often messy. Steps I would consider ...
1. Move the data
2. Put a VIEW in the old location, replacing the data. The VIEW merely copies the data from its new location.
3. Modify the autoexec to (a) define the libname(s) necessary to use the view, and (b) write a prominent message about getting faster service by modifying the program to utilize the new location. This is not foolproof, because different programs might use different LIBNAMEs to refer to the same source folder.
4. Impose a deadline. After a certain date, the old data/view disappears and programs will fail if not updated.
I've seen the analogous functionality carried out with macros, rather than data sets. When 2 complex macros perform the same task, how do you simplify maintenance? Both macros are in use in many programs, so you can't just get rid of one. The solution was to have one macro simply call the other. Then all maintenance could take place on a single macro.
08-21-2014 09:34 AM
Agree Astounding.... by your post of inventing by-passes. There are a lot of them, all dependent on the environment.
On Linux you can work with softlinks. Windows has (W7) also a link command. In SAS metadata a feature like that does not exist, but as it are met metadata-objects (known as those id's) sometimes you can move them to another location without causing broken associations. Renaming can also be possible and seeing all references using the new name. IT is more checking the impact.
08-22-2014 03:08 AM
That was clear but SAS-VA is running on top of EIP using the SAS metadata structures. It is possible to use SMC Eguide and AMO to access part on that.
You can load/prepare data into SAS-VA using DI. The difference is the LASR part for in memory analytics and some of the web-content.
You asked it as a very generic question, you got the answers in a very generic way.
I think when you knew the details of the technical impacts of your questionm you would not have asked anything as seeing the answers.
Another very generic approach is: Do not develop and test something in a production environment. (The technical changes as your question).
Having a dedicated separated environment for that, you can do that there. evaluating the risk impact etc.