lpogoda <lpogoda@HOTMAIL.COM> wrote:

>50 year old literary allusions aside, you seem to be saying >that a low proportion of practicing programmers have >read the documentation, and by implication very few >know what they're doing. That is as may be, >what doesn't follow is that it's the language that's to >blame for the situation.

It's not the language, it's the documentation.

In version 5, the documentation for NODUPLICATES stated with "checks for and eliminates duplicate records". That's pretty clear and unambiguous.

The next paragraph describes how the process is performed, and says "This option causes PROC SORT to compare all variable values for each observation to the previous one written to the output data set." That is misleading, as it implies that each output observation is compared with the previous output observation. That's not how it works, as the version 9 documentaiton makes clear (and there's a new system option, SORTDUP, to make it work in the way a reasonable reader would think it ought to work).

But the version 5 documentation didn't make that clear, and the only way to find out how it actually worked was to use a "properly" constucted set of input data. I had used NODUPLICATES for years before I (accidentally) encountered the problem.

-- JackHamilton@FirstHealth.com Manager, Technical Development Metrics Department, First Health West Sacramento, California USA