regc_color.c

从一个开源软件中摘取的正则表达式模块

			return COLORLESS;
		}
		cm->cd[co].sub = sco;
		cm->cd[sco].sub = sco;	/* open subcolor points to self */
	}
	assert(sco != NOSUB);

	return sco;
}

/*
 * subrange - allocate new subcolors to this range of chrs, fill in arcs
 */
static void
subrange(struct vars * v,
		 chr from,
		 chr to,
		 struct state * lp,
		 struct state * rp)
{
	uchr		uf;
	int			i;

	assert(from <= to);

	/* first, align "from" on a tree-block boundary */
	uf = (uchr) from;
	i = (int) (((uf + BYTTAB - 1) & (uchr) ~BYTMASK) - uf);
	for (; from <= to && i > 0; i--, from++)
		newarc(v->nfa, PLAIN, subcolor(v->cm, from), lp, rp);
	if (from > to)				/* didn't reach a boundary */
		return;

	/* deal with whole blocks */
	for (; to - from >= BYTTAB; from += BYTTAB)
		subblock(v, from, lp, rp);

	/* clean up any remaining partial table */
	for (; from <= to; from++)
		newarc(v->nfa, PLAIN, subcolor(v->cm, from), lp, rp);
}

/*
 * subblock - allocate new subcolors for one tree block of chrs, fill in arcs
 */
static void
subblock(struct vars * v,
		 chr start,				/* first of BYTTAB chrs */
		 struct state * lp,
		 struct state * rp)
{
	uchr		uc = start;
	struct colormap *cm = v->cm;
	int			shift;
	int			level;
	int			i;
	int			b;
	union tree *t;
	union tree *cb;
	union tree *fillt;
	union tree *lastt;
	int			previ;
	int			ndone;
	color		co;
	color		sco;

	assert((uc % BYTTAB) == 0);

	/* find its color block, making new pointer blocks as needed */
	t = cm->tree;
	fillt = NULL;
	for (level = 0, shift = BYTBITS * (NBYTS - 1); shift > 0;
		 level++, shift -= BYTBITS)
	{
		b = (uc >> shift) & BYTMASK;
		lastt = t;
		t = lastt->tptr[b];
		assert(t != NULL);
		fillt = &cm->tree[level + 1];
		if (t == fillt && shift > BYTBITS)
		{						/* need new ptr block */
			t = (union tree *) MALLOC(sizeof(struct ptrs));
			if (t == NULL)
			{
				CERR(REG_ESPACE);
				return;
			}
			memcpy(VS(t->tptr), VS(fillt->tptr),
				   BYTTAB * sizeof(union tree *));
			lastt->tptr[b] = t;
		}
	}

	/* special cases:  fill block or solid block */
	co = t->tcolor[0];
	cb = cm->cd[co].block;
	if (t == fillt || t == cb)
	{
		/* either way, we want a subcolor solid block */
		sco = newsub(cm, co);
		t = cm->cd[sco].block;
		if (t == NULL)
		{						/* must set it up */
			t = (union tree *) MALLOC(sizeof(struct colors));
			if (t == NULL)
			{
				CERR(REG_ESPACE);
				return;
			}
			for (i = 0; i < BYTTAB; i++)
				t->tcolor[i] = sco;
			cm->cd[sco].block = t;
		}
		/* find loop must have run at least once */
		lastt->tptr[b] = t;
		newarc(v->nfa, PLAIN, sco, lp, rp);
		cm->cd[co].nchrs -= BYTTAB;
		cm->cd[sco].nchrs += BYTTAB;
		return;
	}

	/* general case, a mixed block to be altered */
	i = 0;
	while (i < BYTTAB)
	{
		co = t->tcolor[i];
		sco = newsub(cm, co);
		newarc(v->nfa, PLAIN, sco, lp, rp);
		previ = i;
		do
		{
			t->tcolor[i++] = sco;
		} while (i < BYTTAB && t->tcolor[i] == co);
		ndone = i - previ;
		cm->cd[co].nchrs -= ndone;
		cm->cd[sco].nchrs += ndone;
	}
}

/*
 * okcolors - promote subcolors to full colors
 */
static void
okcolors(struct nfa * nfa,
		 struct colormap * cm)
{
	struct colordesc *cd;
	struct colordesc *end = CDEND(cm);
	struct colordesc *scd;
	struct arc *a;
	color		co;
	color		sco;

	for (cd = cm->cd, co = 0; cd < end; cd++, co++)
	{
		sco = cd->sub;
		if (UNUSEDCOLOR(cd) || sco == NOSUB)
		{
			/* has no subcolor, no further action */
		}
		else if (sco == co)
		{
			/* is subcolor, let parent deal with it */
		}
		else if (cd->nchrs == 0)
		{
			/* parent empty, its arcs change color to subcolor */
			cd->sub = NOSUB;
			scd = &cm->cd[sco];
			assert(scd->nchrs > 0);
			assert(scd->sub == sco);
			scd->sub = NOSUB;
			while ((a = cd->arcs) != NULL)
			{
				assert(a->co == co);
				uncolorchain(cm, a);
				a->co = sco;
				colorchain(cm, a);
			}
			freecolor(cm, co);
		}
		else
		{
			/* parent's arcs must gain parallel subcolor arcs */
			cd->sub = NOSUB;
			scd = &cm->cd[sco];
			assert(scd->nchrs > 0);
			assert(scd->sub == sco);
			scd->sub = NOSUB;
			for (a = cd->arcs; a != NULL; a = a->colorchain)
			{
				assert(a->co == co);
				newarc(nfa, a->type, sco, a->from, a->to);
			}
		}
	}
}

/*
 * colorchain - add this arc to the color chain of its color
 */
static void
colorchain(struct colormap * cm,
		   struct arc * a)
{
	struct colordesc *cd = &cm->cd[a->co];

	if (cd->arcs != NULL)
		cd->arcs->colorchainRev = a;
	a->colorchain = cd->arcs;
	a->colorchainRev = NULL;
	cd->arcs = a;
}

/*
 * uncolorchain - delete this arc from the color chain of its color
 */
static void
uncolorchain(struct colormap * cm,
			 struct arc * a)
{
	struct colordesc *cd = &cm->cd[a->co];
	struct arc *aa = a->colorchainRev;

	if (aa == NULL)
	{
		assert(cd->arcs == a);
		cd->arcs = a->colorchain;
	}
	else
	{
		assert(aa->colorchain == a);
		aa->colorchain = a->colorchain;
	}
	if (a->colorchain != NULL)
		a->colorchain->colorchainRev = aa;
	a->colorchain = NULL;		/* paranoia */
	a->colorchainRev = NULL;
}

/*
 * singleton - is this character in its own color?
 */
static int						/* predicate */
singleton(struct colormap * cm,
		  chr c)
{
	color		co;				/* color of c */

	co = GETCOLOR(cm, c);
	if (cm->cd[co].nchrs == 1 && cm->cd[co].sub == NOSUB)
		return 1;
	return 0;
}

/*
 * rainbow - add arcs of all full colors (but one) between specified states
 */
static void
rainbow(struct nfa * nfa,
		struct colormap * cm,
		int type,
		pcolor but,				/* COLORLESS if no exceptions */
		struct state * from,
		struct state * to)
{
	struct colordesc *cd;
	struct colordesc *end = CDEND(cm);
	color		co;

	for (cd = cm->cd, co = 0; cd < end && !CISERR(); cd++, co++)
		if (!UNUSEDCOLOR(cd) && cd->sub != co && co != but &&
			!(cd->flags & PSEUDO))
			newarc(nfa, type, co, from, to);
}

/*
 * colorcomplement - add arcs of complementary colors
 *
 * The calling sequence ought to be reconciled with cloneouts().
 */
static void
colorcomplement(struct nfa * nfa,
				struct colormap * cm,
				int type,
				struct state * of,		/* complements of this guy's PLAIN
										 * outarcs */
				struct state * from,
				struct state * to)
{
	struct colordesc *cd;
	struct colordesc *end = CDEND(cm);
	color		co;

	assert(of != from);
	for (cd = cm->cd, co = 0; cd < end && !CISERR(); cd++, co++)
		if (!UNUSEDCOLOR(cd) && !(cd->flags & PSEUDO))
			if (findarc(of, PLAIN, co) == NULL)
				newarc(nfa, type, co, from, to);
}


#ifdef REG_DEBUG

/*
 * dumpcolors - debugging output
 */
static void
dumpcolors(struct colormap * cm,
		   FILE *f)
{
	struct colordesc *cd;
	struct colordesc *end;
	color		co;
	chr			c;
	char	   *has;

	fprintf(f, "max %ld\n", (long) cm->max);
	if (NBYTS > 1)
		fillcheck(cm, cm->tree, 0, f);
	end = CDEND(cm);
	for (cd = cm->cd + 1, co = 1; cd < end; cd++, co++) /* skip 0 */
		if (!UNUSEDCOLOR(cd))
		{
			assert(cd->nchrs > 0);
			has = (cd->block != NULL) ? "#" : "";
			if (cd->flags & PSEUDO)
				fprintf(f, "#%2ld%s(ps): ", (long) co, has);
			else
				fprintf(f, "#%2ld%s(%2d): ", (long) co,
						has, cd->nchrs);

			/*
			 * Unfortunately, it's hard to do this next bit more efficiently.
			 *
			 * Spencer's original coding has the loop iterating from CHR_MIN
			 * to CHR_MAX, but that's utterly unusable for 32-bit chr. For
			 * debugging purposes it seems fine to print only chr codes up to
			 * 1000 or so.
			 */
			for (c = CHR_MIN; c < 1000; c++)
				if (GETCOLOR(cm, c) == co)
					dumpchr(c, f);
			fprintf(f, "\n");
		}
}

/*
 * fillcheck - check proper filling of a tree
 */
static void
fillcheck(struct colormap * cm,
		  union tree * tree,
		  int level,			/* level number (top == 0) of this block */
		  FILE *f)
{
	int			i;
	union tree *t;
	union tree *fillt = &cm->tree[level + 1];

	assert(level < NBYTS - 1);	/* this level has pointers */
	for (i = BYTTAB - 1; i >= 0; i--)
	{
		t = tree->tptr[i];
		if (t == NULL)
			fprintf(f, "NULL found in filled tree!\n");
		else if (t == fillt)
		{
		}
		else if (level < NBYTS - 2)		/* more pointer blocks below */
			fillcheck(cm, t, level + 1, f);
	}
}

/*
 * dumpchr - print a chr
 *
 * Kind of char-centric but works well enough for debug use.
 */
static void
dumpchr(chr c,
		FILE *f)
{
	if (c == '\\')
		fprintf(f, "\\\\");
	else if (c > ' ' && c <= '~')
		putc((char) c, f);
	else
		fprintf(f, "\\u%04lx", (long) c);
}

#endif   /* REG_DEBUG */