I propose the ability to use patterns with several pattern-matching tokens in the  pattern rules.

This would enable fewer rules to cover more cases. One example where this might be (slightly) advantageous, is when compiling some files in both debug and non-debug mode. The pattern (%) is normally used to match the part of the file before postfix, and different rules are used for debug and non-debug cases.

%-dbg.o : %.c
   gcc -o$@ -DDEBUG $<

%.o : %.c
   gcc -o$@ -DNDEBUG $<

But the linking command also needs a command for each case. If it was possible to have two pattern-matching-tokens (% and ‰) a single rule would suffice.

%‰ : %‰.o
   gcc -o$@ $<

There are off-course a thousand ways to get around this restriction in practice - especially in this toy example.

But when both the number of versions of targets and the number of source files are large, the ability to have more than one pattern-matching-token per pattern rule, would be a huge improvement. I currently get the work done by using $(foreach), $(patsubst) and $(shell) extensively.

This makes writing and reading the Makefile pretty hard. I believe multiple pattern-matchers would appeal to many people. Especially those having experience with functional languages supporting pattern matching (ML variants or Haskell). 

I realize there is no quick fix to make this happen. It would just be a really nice feature.

Best regards,
Anders Borum