Thursday 31 January 2013

NOTE: Ampersands Again

My recent post on double and triple ampersands in macros ended with some doubt. Laurie and Dave immediately took up the challenge to resolve the doubt. You can see the whole painful saga in the comments of the original post, but I've repeated Dave's summary of the conclusions here and I offer my thanks to Laurie and Dave for their tenacious work!
The macro parser performs two basic processes on each pass. If it finds a double ampersand , it converts it into a single one and then proceeds with the rest of the statement; and if it finds a string &xyz that it recognises as a macro variable, it resolves it and then proceeds with the rest of the statement. At the end of the statement, it checks to see if it has been left with any ampersands in the parsed output. If there were, it runs through the process again.

That's exactly how we understood it to work.

The detail we had missed is in what happens if it finds a string &abc that it DOESN'T recognise as a macro variable name. If on the same pass, in the head of the same text string, it also performs a successful double ampersand resolution, it does not regard the unrecognised variable as an error, but allows the string &abc to go through unchanged to the next iteration. But if there were no double ampersand resolutions, it knows that it has had its last chance at a successful resolution and so it will throw a warning - I think it's not an error but I can't check as I don't have SAS installed here.