Yup, journalists at it again.
Yup, journalists at it again.
http://xover.mud.at/~marty/iug2/p243-hopper.pdf
I'm not sure what your objection is but she wrote this in 1952 and specifically used the words compiling routine.
If you look at that paper, she gives an example of the language on figure 9 of page 248:
0 b0i I(1,2,3,4,5) 1,2,3,10,6
1 apn 1,10 4
2 x-e 4 5
3 amc 6,1 7
4 ts0 7 8
5 am0 5,8 9
6 yrs 1,9 0(1,2)
7 aaL 1,2,3 1 8,1
8 ust
(I hope I copied the above correctly, the table is a bit hard to interpret.)
It looks rather like assembly language. Except that the mnemonics in column 2 aren't actually assembly opcodes, they identify subroutines written in machine code. Variables are represented by numbers. The third column is used for input arguments, the fourth for output arguments, the fifth for control flow.
It was a "compiler" in the sense that it read the program and wrote out machine code for each instruction to (a) move input arguments from the variables according to the calling convention (b) make the subroutine call (c) move the output arguments back to the memory locations for the variables.
Very primitive, but I don't think anyone else in 1952 was doing any better. (Other 1950s languages look far more modern – such as LISP and FORTRAN – but they were developed 4-5 years later.)
Mostly kidding, but this is an important step. Assembly, as it came to be known, was one of the first "high level" languages because you could multiply two numbers without having to know the exact opcode for multiplying.