Spec
[ { "operation": "shift", "spec": { "data": { "ARINVT01": { "*": { // array index of ARINVT01 // // &0 is "firstName" // &1 is the current array index of ARINVT01 "firstName": "[&1].FNAME", "lastname": "[&1].LNAME", "INVLOC": { "data": { "*": { // array index of data array // // &0 is "mmm" // &1 is the current array index of the data array // &2 is "data" // &3 is "INVLOC" // &4 is the current array index of ARINVT01 // basically the "&" "number" logic is // how many levels to walk back up the document // to lookup the currently matched value "mmm": "[&4].INVBALANCES.[&1].user", "nnn": "[&4].INVBALANCES.[&1].CURBAL" } } } } } } } } ]

Just out of curiosity I've taken a look at what happens under the hood, and I've used [dtruss/strace][1] on each test.
C++
./a.out < in Saw 6512403 lines in 8 seconds. Crunch speed: 814050
syscalls `sudo dtruss -c ./a.out < in`
CALL COUNT __mac_syscall 1 <snip> open 6 pread 8 mprotect 17 mmap 22 stat64 30 read_nocancel 25958

Python
./a.py < in Read 6512402 lines in 1 seconds. LPS: 6512402
syscalls `sudo dtruss -c ./a.py < in`
CALL COUNT __mac_syscall 1 <snip> open 5 pread 8 mprotect 17 mmap 21 stat64 29
[1]: http://en.wikipedia.org/wiki/Strace