/*
 *  sherlock.c - written by Loki from Rob Pike's sig and comp programs.
 *
 *  This program takes filenames given on the command line,
 *  and reads those files into memory, then compares them
 *  all pairwise to find those which are most similar.
 *
 *  It uses a digital signature generation scheme to randomly
 *  discard information, thus allowing a better match.
 *  Essentially it hashes up N adjacent 'words' of input,
 *  and semi-randomly throws away many of the hashed values
 *  so that it become hard to hide the plagiarised text.
 */

#include <stdlib.h>
#include <string.h>
#include <stdio.h>

char *		Progname = "sherlock";
int		Ntoken = 3;
int		Zerobits = 4;
unsigned long	zeromask;
int		ntoken = 0;
char **		token;
FILE *		Outfile;
int		Thresh = 20;

/* characters to ignore at start and end of each word */
char *		Ignore = " \t\n";

/* characters to treat as word-separators or words on their own */
char *		Punct_full = ",.<>/?;:'\"`~[]{}\\|!@#$%^&*()-+_=";
char *		Punct = "";

typedef struct Sig Sig;
struct Sig
{
	int		nval;
	unsigned long	*val;
};

void	init_token_array(void);
Sig *	signature(FILE *);
int	compare(Sig *, Sig *);

void usage(void)
{
	fprintf(stderr, "%s: find similar files\n", Progname);

	fprintf(stderr, "usage: %s", Progname);
	fprintf(stderr, " [options] file1 file2 ...\n");

	fprintf(stderr, "options:");
	fprintf(stderr, " [-t threshold%%]");
	fprintf(stderr, " [-z zerobits]");
	fprintf(stderr, " [-n chainlength]");
	fprintf(stderr, " [-o outfile]");
	fprintf(stderr, "\n");

	fprintf(stderr, "defaults:");
	fprintf(stderr, " threshold=20%%");
	fprintf(stderr, " zerobits=3");
	fprintf(stderr, " chainlength=4");
	fprintf(stderr, " outfile=the screen");
	fprintf(stderr, "\n");
	exit(2);
}

int main(int argc, char *argv[])
{
	FILE *f;
	int i, j, nfiles, start, percent;
	char *s, *outname;
	Sig **sig;

	Outfile = stdout;
	outname = NULL;

	/* handle options */
	for (start=1; start < argc; start++) {
		if (argv[start][0] != '-')
			break;
		switch (argv[start][1]) {
		case 't':
			s = argv[++start];
			if (s == NULL)
				usage();
			Thresh = atoi(s);
			if (Thresh < 0 || Thresh > 100)
				usage();
			break;
		case 'z':
			s = argv[++start];
			if (s == NULL)
				usage();
			Zerobits = atoi(s);
			if (Zerobits < 0 || Zerobits > 31)
				usage();
			break;
		case 'n':
			s = argv[++start];
			if (s == NULL)
				usage();
			Ntoken = atoi(s);
			if (Ntoken <= 0)
				usage();
			break;
		case 'o':
			s = argv[++start];
			if (s == NULL)
				usage();
			outname = s;
			break;
		default:
			usage();
		}
	}

	nfiles = argc - start;
	if (nfiles < 2)
		usage();

	/* initialise */
	if (outname != NULL)
		Outfile = fopen(outname, "w");
	init_token_array();
	zeromask = (1<<Zerobits)-1;
	sig = malloc(nfiles * sizeof(Sig *));

	/* generate signatures for each file */
	for (i=0; i < nfiles; i++) {
		/* fprintf(stderr, "%s: Reading %s\n", Progname, argv[i+start]); */
		f = fopen(argv[i+start], "r");
		if (f == NULL) {
			fprintf(stderr, "%s: can't open %s:",
				Progname, argv[i+start]);
			perror(NULL);
			continue;
		}
		sig[i] = signature(f);
		fclose(f);
	}

	/* compare each signature pair-wise */
	for (i=0; i < nfiles; i++)
		for (j=i+1; j < nfiles; j++) {
			percent = compare(sig[i], sig[j]);
			if (percent >= Thresh)
				fprintf(Outfile, "%s and %s: %d%%\n",
					argv[i+start], argv[j+start], percent);
		}

	return 0;
}

/* read_word: read a 'word' from the input, ignoring leading characters
   which are inside the 'ignore' string, and stopping if one of
   the 'ignore' or 'punct' characters is found.
   Uses memory allocation to avoid buffer overflow problems.
*/

char * read_word(FILE *f, int *length, char *ignore, char *punct)
{
	long max;
        char *word;
        long pos;
        char *c;
        int ch, is_ignore, is_punct;

        /* check for EOF first */
        if (feof(f)) {
                length = 0;
                return NULL;
        }

        /* allocate a buffer to hold the string */
        pos = 0;
	max = 128;
        word = malloc(sizeof(char) * max);
        c = & word[pos];

	/* initialise some defaults */
	if (ignore == NULL)
		ignore = "";
	if (punct == NULL)
		punct = "";

        /* read characters into the buffer, resizing it if necessary */
        while ((ch = getc(f)) != EOF) {
		is_ignore = (strchr(ignore, ch) != NULL);
		if (pos == 0) {
			if (is_ignore)
				/* ignorable char found at start, skip it */
				continue;
		}
		if (is_ignore)
			/* ignorable char found after start, stop */
			break;
		is_punct = (strchr(punct, ch) != NULL);
		if (is_punct && (pos > 0)) {
			ungetc(ch, f);
			break;
		}
                *c = ch;
                c++;
                pos++;
		if (is_punct)
			break;
                if (pos == max) {
                        /* realloc buffer twice the size */
                        max += max;
                        word = realloc(word, max);
                        c = & word[pos];
                }
        }

        /* set length and check for EOF condition */
        *length = pos;
        if (pos == 0) {
                free(word);
                return NULL;
        }

        /* terminate the string and shrink to smallest required space */
        *c = '\0';
        word = realloc(word, pos+1);
        return word;
}

/* ulcmp:  compare *p1 and *p2 */
int ulcmp(const void *p1, const void *p2)
{
	unsigned long v1, v2;

	v1 = *(unsigned long *) p1;
	v2 = *(unsigned long *) p2;
	if (v1 < v2)
		return -1;
	else if (v1 == v2)
		return 0;
	else
		return 1;
}

/* hash:  hash an array of char* into an unsigned long hash code */
unsigned long hash(char *tok[])
{
	unsigned long h;
	unsigned char *s;
	int i;

	h = 0;
	for (i=0; i < Ntoken; i++)
		for (s=(unsigned char*)tok[i]; *s; s++)
			h = h*31 + *s;
	return h;
}

void init_token_array(void)
{
	int i;

	/* create global array of char* and initialise all to NULL */
	token = malloc(Ntoken * sizeof(char*));
	for (i=0; i < Ntoken; i++)
		token[i] = NULL;
}

Sig * signature(FILE *f)
{
	int nv, na;
	unsigned long *v, h;
	char *str;
	int i, ntoken;
	Sig *sig;

	/* start loading hash values, after we have Ntoken of them */
	v = NULL;
	na = 0;
	nv = 0;
	ntoken = 0;
	while ((str = read_word(f, &i, Ignore, Punct)) != NULL)
	{
		/* step words down by one */
		free(token[0]);
		for (i=0; i < Ntoken-1; i++)
			token[i] = token[i+1];
		/* add new word into array */
		token[Ntoken-1] = str;

		/* if we don't yet have enough words in the array continue */
		ntoken++;
		if (ntoken < Ntoken)
			continue;

		/* hash the array of words */
		h = hash(token);
		if ((h & zeromask) != 0)
			continue;

		/* discard zeros from end of hash value */
		h = h >> Zerobits;

		/* add value into the signature array, resizing if needed */
		if (nv == na) {
			na += 100;
			v = realloc(v, na*sizeof(unsigned long));
		}
		v[nv++] = h;
	}

	/* sort the array of hash values for speed */
	qsort(v, nv, sizeof(v[0]), ulcmp);

	/* allocate and return the Sig structure for this file */
	sig = malloc(sizeof(Sig));
	sig->nval = nv;
	sig->val = v;
	return sig;
}

int compare(Sig *s0, Sig *s1)
{
	int i0, i1, nboth, nsimilar;
	unsigned long v;

	i0 = 0;
	i1 = 0;
	nboth = 0;
	while (i0 < s0->nval && i1 < s1->nval) {
		if (s0->val[i0] == s1->val[i1]) {
			v = s0->val[i0];
			while (i0 < s0->nval && v == s0->val[i0]) {
				i0++;
				nboth++;
			}
			while (i1 < s1->nval && v == s1->val[i1]) {
				i1++;
				nboth++;
			}
			continue;
		}
		if (s0->val[i0] < s1->val[i1])
			i0++;
		else
			i1++;
	}

	if (s0->nval + s1->nval == 0)
		return 0;	/* ignore if both are empty files */

	if (s0->nval + s1->nval == nboth)
		return 100;	/* perfect match if all hash codes match */

	nsimilar = nboth / 2;
	return 100 * nsimilar / (s0->nval + s1->nval - nsimilar);
}

/*
 *  Let f1 == filesize(file1) == A+B
 *  and f2 == filesize(file2) == A+C
 *  where A is the similar section and B or C are dissimilar
 *
 *  Similarity = 100 * A / (f1 + f2 - A)
 *             = 100 * A / (A+B + A+C - A)
 *             = 100 * A / (A+B+C)
 *
 *  Thus if A==B==C==n the similarity will be 33% (one third)
 *  This is desireable since we are finding the ratio of similarities
 *  as a fraction of (similarities+dissimilarities).
 *
 *  The other way of doing things would be to find the ratio of
 *  the sum of similarities as a fraction of total file size:
 *  Similarity = 100 * (A+A) / (A+B + A+C)
 *  This produces higher percentages and more false matches.
 */