build.c

来自「这是一个同样来自贝尔实验室的和UNIX有着渊源的操作系统, 其简洁的设计和实现易」· C语言 代码 · 共 2,670 行 · 第 1/5 页

					if(tr->align.halign != ALnone)
						c->align.halign = tr->align.halign;
					else if(tag == Tth)
						c->align.halign = ALcenter;
					else
						c->align.halign = ALleft;
				}
				if(c->align.valign == ALnone) {
					if(tr->align.valign != ALnone)
						c->align.valign = tr->align.valign;
					else
						c->align.valign = ALmiddle;
				}
				c->nextinrow = tr->cells;
				tr->cells = c;
				break;

			case Ttd+RBRA:
			case Tth+RBRA:
				if(curtab == nil || curtab->cells == nil) {
					if(warn)
						fprint(2, "unexpected %T\n", tok);
					continue;
				}
				ps = finishcell(curtab, ps);
				break;

			// <!ELEMENT TEXTAREA - - (	//PCDATA)>
			case Ttextarea:
				if(is->curform == nil) {
					if(warn)
						fprint(2, "<TEXTAREA> not inside <FORM>\n");
					continue;
				}
				field = newformfield(Ftextarea,
					++is->curform->nfields,
					is->curform,
					aval(tok, Aname),
					nil,
					0,
					0,
					is->curform->fields);
				is->curform->fields = field;
				field->rows = auintval(tok, Arows, 3);
				field->cols = auintval(tok, Acols, 50);
				field->value = getpcdata(toks, tokslen, &toki);
				if(warn && toki < tokslen - 1 && toks[toki + 1].tag != Ttextarea + RBRA)
					fprint(2, "warning: <TEXTAREA> data ended by %T\n", &toks[toki + 1]);
				ffit = newiformfield(field);
				additem(ps, ffit, tok);
				if(ffit->genattr != nil)
					field->events = ffit->genattr->events;
				break;

			// <!ELEMENT TITLE - - (	//PCDATA)* -(%head.misc)>
			case Ttitle:
				di->doctitle = getpcdata(toks, tokslen, &toki);
				if(warn && toki < tokslen - 1 && toks[toki + 1].tag != Ttitle + RBRA)
					fprint(2, "warning: <TITLE> data ended by %T\n", &toks[toki + 1]);
				break;

			// <!ELEMENT TR - O (TH|TD)+>
			// rows are accumulated in reverse order in curtab->rows
			case Ttr:
				if(curtab == nil) {
					if(warn)
						fprint(2, "warning: <TR> outside <TABLE>\n");
					continue;
				}
				if(ps->inpar) {
					popjust(ps);
					ps->inpar = 0;
				}
				ps = finishcell(curtab, ps);
				if(curtab->rows != nil)
					curtab->rows->flags = 0;
				curtab->rows = newtablerow(aalign(tok),
					makebackground(nil, acolorval(tok, Abgcolor, curtab->background.color)),
					TFparsing,
					curtab->rows);
				break;

			case Ttr+RBRA:
				if(curtab == nil || curtab->rows == nil) {
					if(warn)
						fprint(2, "warning: unexpected </TR>\n");
					continue;
				}
				ps = finishcell(curtab, ps);
				tr = curtab->rows;
				if(tr->cells == nil) {
					if(warn)
						fprint(2, "warning: empty row\n");
					curtab->rows = tr->next;
					tr->next = nil;
				}
				else
					tr->flags = 0;
				break;

			// <!ELEMENT (TT|CODE|KBD|SAMP) - - (%text)*>
			case Ttt:
			case Tcode:
			case Tkbd:
			case Tsamp:
				pushfontstyle(ps, FntT);
				break;

			// Tags that have empty action
			case Tabbr:
			case Tabbr+RBRA:
			case Tacronym:
			case Tacronym+RBRA:
			case Tarea+RBRA:
			case Tbase+RBRA:
			case Tbasefont+RBRA:
			case Tbr+RBRA:
			case Tdd+RBRA:
			case Tdt+RBRA:
			case Tframe+RBRA:
			case Thr+RBRA:
			case Thtml:
			case Thtml+RBRA:
			case Timg+RBRA:
			case Tinput+RBRA:
			case Tisindex+RBRA:
			case Tli+RBRA:
			case Tlink:
			case Tlink+RBRA:
			case Tmeta+RBRA:
			case Toption+RBRA:
			case Tparam+RBRA:
			case Ttextarea+RBRA:
			case Ttitle+RBRA:
				break;


			// Tags not implemented
			case Tbdo:
			case Tbdo+RBRA:
			case Tbutton:
			case Tbutton+RBRA:
			case Tdel:
			case Tdel+RBRA:
			case Tfieldset:
			case Tfieldset+RBRA:
			case Tiframe:
			case Tiframe+RBRA:
			case Tins:
			case Tins+RBRA:
			case Tlabel:
			case Tlabel+RBRA:
			case Tlegend:
			case Tlegend+RBRA:
			case Tobject:
			case Tobject+RBRA:
			case Toptgroup:
			case Toptgroup+RBRA:
			case Tspan:
			case Tspan+RBRA:
				if(warn) {
					if(tag > RBRA)
						tag -= RBRA;
					fprint(2, "warning: unimplemented HTML tag: %S\n", tagnames[tag]);
				}
				break;

			default:
				if(warn)
					fprint(2, "warning: unknown HTML tag: %S\n", tok->text);
				break;
			}
	}
	// some pages omit trailing </table>
	while(curtab != nil) {
		if(warn)
			fprint(2, "warning: <TABLE> not closed\n");
		if(curtab->cells != nil) {
			ps = finishcell(curtab, ps);
			if(curtab->cells == nil) {
				if(warn)
					fprint(2, "warning: empty table\n");
			}
			else {
				if(curtab->rows != nil)
					curtab->rows->flags = 0;
				finish_table(curtab);
				ps->skipping = 0;
				additem(ps, newitable(curtab), curtab->tabletok);
				addbrk(ps, 0, 0);
			}
		}
		if(is->tabstk != nil)
			is->tabstk = is->tabstk->next;
		curtab->next = di->tables;
		di->tables = curtab;
		curtab = is->tabstk;
	}
	outerps = lastps(ps);
	ans = outerps->items->next;
	freeitem(outerps->items);
	// note: ans may be nil and di->kids not nil, if there's a frameset!
	outerps->items = newispacer(ISPnull);
	outerps->lastit = outerps->items;
	is->psstk = ps;
	if(ans != nil && di->hasscripts) {
		// TODO evalscript(nil);
		;
	}

return_ans:
	if(dbgbuild) {
		assert(validitems(ans));
		if(ans == nil)
			fprint(2, "getitems returning nil\n");
		else
			printitems(ans, "getitems returning:");
	}
	return ans;
}

// Concatenate together maximal set of Data tokens, starting at toks[toki+1].
// Lexer has ensured that there will either be a following non-data token or
// we will be at eof.
// Return emallocd trimmed concatenation, and update *ptoki to last used toki
static Rune*
getpcdata(Token* toks, int tokslen, int* ptoki)
{
	Rune*	ans;
	Rune*	p;
	Rune*	trimans;
	int	anslen;
	int	trimanslen;
	int	toki;
	Token*	tok;

	ans = nil;
	anslen = 0;
	// first find length of answer
	toki = (*ptoki) + 1;
	while(toki < tokslen) {
		tok = &toks[toki];
		if(tok->tag == Data) {
			toki++;
			anslen += _Strlen(tok->text);
		}
		else
			break;
	}
	// now make up the initial answer
	if(anslen > 0) {
		ans = _newstr(anslen);
		p = ans;
		toki = (*ptoki) + 1;
		while(toki < tokslen) {
			tok = &toks[toki];
			if(tok->tag == Data) {
				toki++;
				p = _Stradd(p, tok->text, _Strlen(tok->text));
			}
			else
				break;
		}
		*p = 0;
		_trimwhite(ans, anslen, &trimans, &trimanslen);
		if(trimanslen != anslen) {
			p = ans;
			ans = _Strndup(trimans, trimanslen);
			free(p);
		}
	}
	*ptoki = toki-1;
	return ans;
}

// If still parsing head of curtab->cells list, finish it off
// by transferring the items on the head of psstk to the cell.
// Then pop the psstk and return the new psstk.
static Pstate*
finishcell(Table* curtab, Pstate* psstk)
{
	Tablecell*	c;
	Pstate* psstknext;

	c = curtab->cells;
	if(c != nil) {
		if((c->flags&TFparsing)) {
			psstknext = psstk->next;
			if(psstknext == nil) {
				if(warn)
					fprint(2, "warning: parse state stack is wrong\n");
			}
			else {
				c->content = psstk->items->next;
				c->flags &= ~TFparsing;
				freepstate(psstk);
				psstk = psstknext;
			}
		}
	}
	return psstk;
}

// Make a new Pstate for a cell, based on the old pstate, oldps.
// Also, put the new ps on the head of the oldps stack.
static Pstate*
cell_pstate(Pstate* oldps, int ishead)
{
	Pstate*	ps;
	int	sty;

	ps = newpstate(oldps);
	ps->skipwhite = 1;
	ps->curanchor = oldps->curanchor;
	copystack(&ps->fntstylestk, &oldps->fntstylestk);
	copystack(&ps->fntsizestk, &oldps->fntsizestk);
	ps->curfont = oldps->curfont;
	ps->curfg = oldps->curfg;
	ps->curbg = oldps->curbg;
	copystack(&ps->fgstk, &oldps->fgstk);
	ps->adjsize = oldps->adjsize;
	if(ishead) {
		sty = ps->curfont%NumSize;
		ps->curfont = FntB*NumSize + sty;
	}
	return ps;
}

// Return a new Pstate with default starting state.
// Use link to add it to head of a list, if any.
static Pstate*
newpstate(Pstate* link)
{
	Pstate*	ps;

	ps = (Pstate*)emalloc(sizeof(Pstate));
	ps->curfont = DefFnt;
	ps->curfg = Black;
	ps->curbg.image = nil;
	ps->curbg.color = White;
	ps->curul = ULnone;
	ps->curjust = ALleft;
	ps->curstate = IFwrap;
	ps->items = newispacer(ISPnull);
	ps->lastit = ps->items;
	ps->prelastit = nil;
	ps->next = link;
	return ps;
}

// Return last Pstate on psl list
static Pstate*
lastps(Pstate* psl)
{
	assert(psl != nil);
	while(psl->next != nil)
		psl = psl->next;
	return psl;
}

// Add it to end of ps item chain, adding in current state from ps.
// Also, if tok is not nil, scan it for generic attributes and assign
// the genattr field of the item accordingly.
static void
additem(Pstate* ps, Item* it, Token* tok)
{
	int	aid;
	int	any;
	Rune*	i;
	Rune*	c;
	Rune*	s;
	Rune*	t;
	Attr*	a;
	SEvent*	e;

	if(ps->skipping) {
		if(warn)
			fprint(2, "warning: skipping item: %I\n", it);
		return;
	}
	it->anchorid = ps->curanchor;
	it->state |= ps->curstate;
	if(tok != nil) {
		any = 0;
		i = nil;
		c = nil;
		s = nil;
		t = nil;
		e = nil;
		for(a = tok->attr; a != nil; a = a->next) {
			aid = a->attid;
			if(!attrinfo[aid])
				continue;
			switch(aid) {
			case Aid:
				i = a->value;
				break;

			case Aclass:
				c = a->value;
				break;

			case Astyle:
				s = a->value;
				break;

			case Atitle:
				t = a->value;
				break;

			default:
				assert(aid >= Aonblur && aid <= Aonunload);
				e = newscriptevent(scriptev[a->attid], a->value, e);
				break;
			}
			a->value = nil;
			any = 1;
		}
		if(any)
			it->genattr = newgenattr(i, c, s, t, e);
	}
	ps->curstate &= ~(IFbrk|IFbrksp|IFnobrk|IFcleft|IFcright);
	ps->prelastit = ps->lastit;
	ps->lastit->next = it;
	ps->lastit = it;
}

// Make a text item out of s,
// using current font, foreground, vertical offset and underline state.
static Item*
textit(Pstate* ps, Rune* s)
{
	assert(s != nil);
	return newitext(s, ps->curfont, ps->curfg, ps->curvoff + Voffbias, ps->curul);
}

// Add text item or items for s, paying attention to
// current font, foreground, baseline offset, underline state,
// and literal mode.  Unless we're in literal mode, compress
// whitespace to single blank, and, if curstate has a break,
// trim any leading whitespace.  Whether in literal mode or not,
// turn nonbreaking spaces into spacer items with IFnobrk set.
//
// In literal mode, break up s at newlines and add breaks instead.
// Also replace tabs appropriate number of spaces.
// In nonliteral mode, break up the items every 100 or so characters
// just to make the layout algorithm not go quadratic.
//
// addtext assumes ownership of s.
static void
addtext(Pstate* ps, Rune* s)
{
	int	n;
	int	i;
	int	j;
	int	k;
	int	col;
	int	c;
	int	nsp;
	Item*	it;
	Rune*	ss;
	Rune*	p;
	Rune	buf[SMALLBUFSIZE];

	assert(s != nil);
	n = runestrlen(s);
	i = 0;
	j = 0;
	if(ps->literal) {
		col = 0;
		while(i < n) {
			if(s[i] == '\n') {
				if(i > j) {
					// trim trailing blanks from line
					for(k = i; k > j; k--)
						if(s[k - 1] != ' ')
							break;
					if(k > j)
						additem(ps, textit(ps, _Strndup(s+j, k-j)), nil);
				}
				addlinebrk(ps, 0);
				j = i + 1;
				col = 0;
			}
			else {
				if(s[i] == '\t') {
					col += i - j;
					nsp = 8 - (col%8);
					// make ss = s[j:i] + nsp spaces
					ss = _newstr(i-j+nsp);
					p = _Stradd(ss, s+j, i-j);
					p = _Stradd(p, L"        ", nsp);
					*p = 0;
					additem(ps, textit(ps, ss), nil);
					col += nsp;
					j = i + 1;
				}
				else if(s[i] == NBSP) {
					if(i > j)
						additem(ps, textit(ps, _Strndup(s+j, i-j)), nil);
					addnbsp(ps);
					col += (i - j) + 1;
					j = i + 1;
				}
			}
			i++;
		}
		if(i > j) {
			if(j == 0 && i == n) {
				// just transfer s over
				additem(ps, textit(ps, s), nil);
			}
			else {
				additem(ps, textit(ps, _Strndup(s+j, i-j)), nil);
				free(s);
			}
		}
	}
	else {	// not literal mode
		if((ps->curstate&IFbrk) || ps->lastit == ps->items)
			while(i < n) {
				c = s[i];
				if(c >= 256 || !isspace(c))
					break;
				i++;
			}
		p = buf;
		for(j = i; i < n; i++) {
			assert(p+i-j < buf+SMALLBUFSIZE-1);
			c = s[i];
			if(c == NBSP) {
				if(i > j)
					p = _Stradd(p, s+j, i-j);
				if(p > buf)