Files in this tarball
 README: file thatyou're reading
 c.range-counting-tree: directory with range counting tree
    |
    |-- rct.h: interface to range tree
    |-- rct.c: range tree implementation
 point: directory with definition of 2-d point
    |
    |-- point.h: interface
    |-- point.c: implementation
 peacock: directory with Peacock's test
    |- c.bf: brute force implmentation  iof Peacock's test
    |- c.rct: range counting tree implementation of Peacock's test
        |
        |- peacock.h: interface of Peacock's test
        |- peacock.c: implementation of Peacock's test using range trees
        |- statistic.c: test program
 ff: directory with Fasano and Franceschini's test
    |- c.rct: range counting tree implementation of the test
        |
        |- ff.h: interface to the test
        |- ff.c: implementation of the test using range trees
        |- statistic.c: test program


Compilation:

 From peacock/c.rct:

    gcc -std=c99 -Wall rct.c peacock.c point.c statistic.c -o p

 From ff/c.rct:
     gcc -std=c99 -Wall rct.c ff.c point.c statistic.c -o ff

Execution:
    the test programs share the same interface:
      they read the points from STDIN
        one point per line (x-coordinate, followed by space, followed by y-coordinate)
      they assume assume a unique input stream with all points from both samples
      they receive two command-line arguments: numeber of points in each sample

    From ff/c.rct, assuming that s0 contains 1024 and s1 contains 1024 points

    cat s0 | cat _ s1 | ./ff 1024 1024

    both programs output time for computing and distance

IMPORTANT:
 - time is computed using clock. not reliable for long runs
 - distance is the value returned. NOT probability
 - See Peacock's or Fasano and Franceschini's papers (also Press et ali's "Numerical recipes") for trivial algorithm 
 for probability computation
 

# vi:sw=4 ts=4 sm et autoindent backspace=eol,start,indent lines=35 columns=120