2 * \file output_xhtml.cpp
3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
8 * This code is based upon output_docbook.cpp
10 * Full author contact details are available in file CREDITS.
15 #include "output_xhtml.h"
18 #include "buffer_funcs.h"
19 #include "BufferParams.h"
22 #include "OutputParams.h"
23 #include "Paragraph.h"
24 #include "ParagraphList.h"
25 #include "ParagraphParameters.h"
28 #include "TextClass.h"
30 #include "support/lassert.h"
31 #include "support/debug.h"
32 #include "support/lstrings.h"
37 using namespace lyx::support;
43 docstring escapeChar(char_type c)
67 // escape what needs escaping
68 docstring htmlize(docstring const & str) {
70 docstring::const_iterator it = str.begin();
71 docstring::const_iterator en = str.end();
72 for (; it != en; ++it)
78 string escapeChar(char c)
102 // escape what needs escaping
103 string htmlize(string const & str) {
105 string::const_iterator it = str.begin();
106 string::const_iterator en = str.end();
107 for (; it != en; ++it)
108 d << escapeChar(*it);
113 string cleanAttr(string const & str)
116 string::const_iterator it = str.begin();
117 string::const_iterator en = str.end();
118 for (; it != en; ++it)
119 newname += isalnum(*it) ? *it : '_';
124 docstring cleanAttr(docstring const & str)
127 docstring::const_iterator it = str.begin();
128 docstring::const_iterator en = str.end();
129 for (; it != en; ++it)
138 bool isFontTag(string const & s)
140 return s == "em" || s == "strong"; // others?
145 docstring StartTag::asTag() const
147 string output = "<" + tag_;
149 output += " " + html::htmlize(attr_);
151 return from_utf8(output);
155 docstring StartTag::asEndTag() const
157 string output = "</" + tag_ + ">";
158 return from_utf8(output);
162 docstring EndTag::asEndTag() const
164 string output = "</" + tag_ + ">";
165 return from_utf8(output);
169 docstring CompTag::asTag() const
171 string output = "<" + tag_;
173 output += " " + html::htmlize(attr_);
175 return from_utf8(output);
179 ////////////////////////////////////////////////////////////////
183 ////////////////////////////////////////////////////////////////
185 XHTMLStream::XHTMLStream(odocstream & os)
186 : os_(os), nextraw_(false)
190 void XHTMLStream::cr()
193 os_ << from_ascii("\n");
197 void XHTMLStream::writeError(std::string const & s)
200 os_ << from_utf8("<!-- Output Error: " + s + " -->");
204 bool XHTMLStream::closeFontTags()
206 if (tag_stack_.empty())
208 // first, we close any open font tags we can close
209 StartTag curtag = tag_stack_.back();
210 while (html::isFontTag(curtag.tag_)) {
211 os_ << curtag.asEndTag();
212 tag_stack_.pop_back();
213 if (tag_stack_.empty())
214 // this probably shouldn't happen, since then the
215 // font tags weren't in any other tag. but that
216 // problem will likely be caught elsewhere.
218 curtag = tag_stack_.back();
220 // so we've hit a non-font tag. let's see if any of the
221 // remaining tags are font tags.
222 TagStack::const_iterator it = tag_stack_.begin();
223 TagStack::const_iterator en = tag_stack_.end();
224 bool noFontTags = true;
225 for (; it != en; ++it) {
226 if (html::isFontTag(it->tag_)) {
227 writeError("Font tag `" + it->tag_ + "' still open in closeFontTags().");
235 void XHTMLStream::clearTagDeque()
237 while (!pending_tags_.empty()) {
238 StartTag const & tag = pending_tags_.front();
241 tag_stack_.push_back(tag);
242 pending_tags_.pop_front();
247 XHTMLStream & XHTMLStream::operator<<(docstring const & d)
254 os_ << html::htmlize(d);
259 XHTMLStream & XHTMLStream::operator<<(const char * s)
262 docstring const d = from_ascii(s);
267 os_ << html::htmlize(d);
272 XHTMLStream & XHTMLStream::operator<<(char_type c)
279 os_ << html::escapeChar(c);
284 XHTMLStream & XHTMLStream::operator<<(NextRaw const &)
291 XHTMLStream & XHTMLStream::operator<<(StartTag const & tag)
293 if (tag.tag_.empty())
295 pending_tags_.push_back(tag);
302 XHTMLStream & XHTMLStream::operator<<(CompTag const & tag)
304 if (tag.tag_.empty())
313 bool XHTMLStream::isTagOpen(string const & stag)
315 TagStack::const_iterator sit = tag_stack_.begin();
316 TagStack::const_iterator const sen = tag_stack_.end();
317 for (; sit != sen; ++sit)
318 // we could check for the
319 if (sit->tag_ == stag)
325 // this is complicated, because we want to make sure that
326 // everything is properly nested. the code ought to make
327 // sure of that, but we won't assert (yet) if we run into
328 // a problem. we'll just output error messages and try our
329 // best to make things work.
330 XHTMLStream & XHTMLStream::operator<<(EndTag const & etag)
332 if (etag.tag_.empty())
334 // first make sure we're not closing an empty tag
335 if (!pending_tags_.empty()) {
336 StartTag const & stag = pending_tags_.back();
337 if (etag.tag_ == stag.tag_) {
338 // we have <tag></tag>, so we discard it and remove it
339 // from the pending_tags_.
340 pending_tags_.pop_back();
343 // there is a pending tag that isn't the one we are trying
345 // is this tag itself pending?
346 // non-const iterators because we may call erase().
347 TagDeque::iterator dit = pending_tags_.begin();
348 TagDeque::iterator const den = pending_tags_.end();
349 for (; dit != den; ++dit) {
350 if (dit->tag_ == etag.tag_) {
351 // it was pending, so we just erase it
352 writeError("Tried to close pending tag `" + etag.tag_
353 + "' when other tags were pending. Tag discarded.");
354 pending_tags_.erase(dit);
358 // so etag isn't itself pending. is it even open?
359 if (!isTagOpen(etag.tag_)) {
360 writeError("Tried to close `" + etag.tag_
361 + "' when tag was not open. Tag discarded.");
364 // ok, so etag is open.
365 // our strategy will be as below: we will do what we need to
366 // do to close this tag.
367 string estr = "Closing tag `" + etag.tag_
368 + "' when other tags are pending. Discarded pending tags:\n";
369 for (dit = pending_tags_.begin(); dit != den; ++dit)
370 estr += dit->tag_ + "\n";
372 // clear the pending tags...
373 pending_tags_.clear();
374 // ...and then just fall through.
377 // is the tag we are closing the last one we opened?
378 if (etag.tag_ == tag_stack_.back().tag_) {
380 os_ << etag.asEndTag();
381 // ...and forget about it
382 tag_stack_.pop_back();
386 // we are trying to close a tag other than the one last opened.
387 // let's first see if this particular tag is still open somehow.
388 if (!isTagOpen(etag.tag_)) {
389 writeError("Tried to close `" + etag.tag_
390 + "' when tag was not open. Tag discarded.");
394 // so the tag was opened, but other tags have been opened since
395 // and not yet closed.
396 // if it's a font tag, though...
397 if (html::isFontTag(etag.tag_)) {
398 // it won't be a problem if the other tags open since this one
399 // are also font tags.
400 TagStack::const_reverse_iterator rit = tag_stack_.rbegin();
401 TagStack::const_reverse_iterator ren = tag_stack_.rend();
402 for (; rit != ren; ++rit) {
403 if (rit->tag_ == etag.tag_)
405 if (!html::isFontTag(rit->tag_)) {
406 // we'll just leave it and, presumably, have to close it later.
407 writeError("Unable to close font tag `" + etag.tag_
408 + "' due to open non-font tag `" + rit->tag_ + "'.");
414 // <em>this is <strong>bold
415 // and are being asked to closed em. we want:
416 // <em>this is <strong>bold</strong></em><strong>
417 // first, we close the intervening tags...
418 StartTag curtag = tag_stack_.back();
419 // ...remembering them in a stack.
421 while (curtag.tag_ != etag.tag_) {
422 os_ << curtag.asEndTag();
423 fontstack.push_back(curtag);
424 tag_stack_.pop_back();
425 curtag = tag_stack_.back();
427 // now close our tag...
428 os_ << etag.asEndTag();
429 tag_stack_.pop_back();
431 // ...and restore the other tags.
432 rit = fontstack.rbegin();
433 ren = fontstack.rend();
434 for (; rit != ren; ++rit)
435 pending_tags_.push_back(*rit);
439 // it wasn't a font tag.
440 // so other tags were opened before this one and not properly closed.
441 // so we'll close them, too. that may cause other issues later, but it
442 // at least guarantees proper nesting.
443 writeError("Closing tag `" + etag.tag_
444 + "' when other tags are open, namely:");
445 StartTag curtag = tag_stack_.back();
446 while (curtag.tag_ != etag.tag_) {
447 writeError(curtag.tag_);
448 os_ << curtag.asEndTag();
449 tag_stack_.pop_back();
450 curtag = tag_stack_.back();
452 // curtag is now the one we actually want.
453 os_ << curtag.asEndTag();
454 tag_stack_.pop_back();
459 // End code for XHTMLStream
463 // convenience functions
465 inline void openTag(XHTMLStream & xs, Layout const & lay)
467 xs << StartTag(lay.htmltag(), lay.htmlattr());
471 inline void closeTag(XHTMLStream & xs, Layout const & lay)
473 xs << EndTag(lay.htmltag());
477 inline void openLabelTag(XHTMLStream & xs, Layout const & lay)
479 xs << StartTag(lay.htmllabeltag(), lay.htmllabelattr());
483 inline void closeLabelTag(XHTMLStream & xs, Layout const & lay)
485 xs << EndTag(lay.htmllabeltag());
489 inline void openItemTag(XHTMLStream & xs, Layout const & lay)
491 xs << StartTag(lay.htmlitemtag(), lay.htmlitemattr(), true);
495 inline void closeItemTag(XHTMLStream & xs, Layout const & lay)
497 xs << EndTag(lay.htmlitemtag());
500 // end of convenience functions
502 ParagraphList::const_iterator searchParagraphHtml(
503 ParagraphList::const_iterator p,
504 ParagraphList::const_iterator const & pend)
506 for (++p; p != pend && p->layout().latextype == LATEX_PARAGRAPH; ++p)
513 ParagraphList::const_iterator searchEnvironmentHtml(
514 ParagraphList::const_iterator const pstart,
515 ParagraphList::const_iterator const & pend)
517 ParagraphList::const_iterator p = pstart;
518 Layout const & bstyle = p->layout();
519 size_t const depth = p->params().depth();
520 for (++p; p != pend; ++p) {
521 Layout const & style = p->layout();
522 // It shouldn't happen that e.g. a section command occurs inside
523 // a quotation environment, at a higher depth, but as of 6/2009,
524 // it can happen. We pretend that it's just at lowest depth.
525 if (style.latextype == LATEX_COMMAND)
527 // If depth is down, we're done
528 if (p->params().depth() < depth)
530 // If depth is up, we're not done
531 if (p->params().depth() > depth)
533 // Now we know we are at the same depth
534 if (style.latextype == LATEX_PARAGRAPH
535 || style.latexname() != bstyle.latexname())
542 ParagraphList::const_iterator makeParagraphs(Buffer const & buf,
544 OutputParams const & runparams,
546 ParagraphList::const_iterator const & pbegin,
547 ParagraphList::const_iterator const & pend)
549 ParagraphList::const_iterator const begin = text.paragraphs().begin();
550 ParagraphList::const_iterator par = pbegin;
551 for (; par != pend; ++par) {
552 Layout const & lay = par->layout();
553 if (!lay.counter.empty())
554 buf.params().documentClass().counters().step(lay.counter);
555 // FIXME We should see if there's a label to be output and
556 // do something with it.
560 // If we are already in a paragraph, and this is the first one, then we
561 // do not want to open the paragraph tag.
562 // we also do not want to open it if the current layout does not permit
563 // multiple paragraphs.
564 bool const opened = runparams.html_make_pars &&
565 (par != pbegin || !runparams.html_in_par);
568 docstring const deferred =
569 par->simpleLyXHTMLOnePar(buf, xs, runparams, text.outerFont(distance(begin, par)));
571 // We want to issue the closing tag if either:
572 // (i) We opened it, and either html_in_par is false,
573 // or we're not in the last paragraph, anyway.
574 // (ii) We didn't open it and html_in_par is true,
575 // but we are in the first par, and there is a next par.
576 ParagraphList::const_iterator nextpar = par;
578 bool const needclose =
579 (opened && (!runparams.html_in_par || nextpar != pend))
580 || (!opened && runparams.html_in_par && par == pbegin && nextpar != pend);
585 if (!deferred.empty()) {
586 xs << XHTMLStream::NextRaw() << deferred;
594 ParagraphList::const_iterator makeBibliography(Buffer const & buf,
596 OutputParams const & runparams,
598 ParagraphList::const_iterator const & pbegin,
599 ParagraphList::const_iterator const & pend)
601 xs << StartTag("h2", "class='bibliography'");
602 xs << pbegin->layout().labelstring(false);
605 xs << StartTag("div", "class='bibliography'");
607 makeParagraphs(buf, xs, runparams, text, pbegin, pend);
613 bool isNormalEnv(Layout const & lay)
615 return lay.latextype == LATEX_ENVIRONMENT;
619 ParagraphList::const_iterator makeEnvironmentHtml(Buffer const & buf,
621 OutputParams const & runparams,
623 ParagraphList::const_iterator const & pbegin,
624 ParagraphList::const_iterator const & pend)
626 ParagraphList::const_iterator const begin = text.paragraphs().begin();
627 ParagraphList::const_iterator par = pbegin;
628 Layout const & bstyle = par->layout();
629 depth_type const origdepth = pbegin->params().depth();
631 // open tag for this environment
635 // we will on occasion need to remember a layout from before.
636 Layout const * lastlay = 0;
638 while (par != pend) {
639 Layout const & style = par->layout();
640 // the counter only gets stepped if we're in some kind of list,
641 // or if it's the first time through.
642 // note that enum, etc, are handled automatically.
643 // FIXME There may be a bug here about user defined enumeration
644 // types. If so, then we'll need to take the counter and add "i",
645 // "ii", etc, as with enum.
646 if (!style.counter.empty() &&
647 (par == pbegin || !isNormalEnv(style))
648 && style.latextype == LATEX_LIST_ENVIRONMENT)
649 buf.params().documentClass().counters().step(style.counter);
650 ParagraphList::const_iterator send;
651 // this will be positive, if we want to skip the initial word
652 // (if it's been taken for the label).
655 switch (style.latextype) {
656 case LATEX_ENVIRONMENT:
657 case LATEX_LIST_ENVIRONMENT:
658 case LATEX_ITEM_ENVIRONMENT: {
659 // There are two possiblities in this case.
660 // One is that we are still in the environment in which we
661 // started---which we will be if the depth is the same.
662 if (par->params().depth() == origdepth) {
663 LASSERT(bstyle == style, /* */);
665 closeItemTag(xs, *lastlay);
668 bool const labelfirst = style.htmllabelfirst();
669 if (isNormalEnv(style)) {
670 // in this case, we print the label only for the first
671 // paragraph (as in a theorem).
672 openItemTag(xs, style);
673 if (par == pbegin && style.htmllabeltag() != "NONE") {
674 docstring const lbl =
675 pbegin->expandLabel(style, buf.params(), false);
677 openLabelTag(xs, style);
679 closeLabelTag(xs, style);
683 } else { // some kind of list
685 openItemTag(xs, style);
686 if (style.labeltype == LABEL_MANUAL
687 && style.htmllabeltag() != "NONE") {
688 openLabelTag(xs, style);
689 // sep = par->firstWordLyXHTML(xs, runparams);
690 closeLabelTag(xs, style);
693 else if (style.labeltype != LABEL_NO_LABEL
694 && style.htmllabeltag() != "NONE") {
695 openLabelTag(xs, style);
696 xs << par->expandLabel(style, buf.params(), false);
697 closeLabelTag(xs, style);
701 openItemTag(xs, style);
703 par->simpleLyXHTMLOnePar(buf, xs, runparams,
704 text.outerFont(distance(begin, par)), sep);
706 // We may not want to close the tag yet, in particular,
707 // if we're not at the end...
709 // and are doing items...
710 && !isNormalEnv(style)
711 // and if the depth has changed...
712 && par->params().depth() != origdepth) {
713 // then we'll save this layout for later, and close it when
714 // we get another item.
717 closeItemTag(xs, style);
720 // The other possibility is that the depth has increased, in which
721 // case we need to recurse.
723 send = searchEnvironmentHtml(par, pend);
724 par = makeEnvironmentHtml(buf, xs, runparams, text, par, send);
728 case LATEX_PARAGRAPH:
729 send = searchParagraphHtml(par, pend);
730 par = makeParagraphs(buf, xs, runparams, text, par, send);
733 case LATEX_BIB_ENVIRONMENT:
736 par = makeParagraphs(buf, xs, runparams, text, par, send);
746 closeItemTag(xs, *lastlay);
747 closeTag(xs, bstyle);
753 void makeCommand(Buffer const & buf,
755 OutputParams const & runparams,
757 ParagraphList::const_iterator const & pbegin)
759 Layout const & style = pbegin->layout();
760 if (!style.counter.empty())
761 buf.params().documentClass().counters().step(style.counter);
765 // Label around sectioning number:
766 // FIXME Probably need to account for LABEL_MANUAL
767 if (style.labeltype != LABEL_NO_LABEL) {
768 openLabelTag(xs, style);
769 xs << pbegin->expandLabel(style, buf.params(), false);
770 closeLabelTag(xs, style);
771 // Otherwise the label might run together with the text
772 xs << from_ascii(" ");
775 ParagraphList::const_iterator const begin = text.paragraphs().begin();
776 pbegin->simpleLyXHTMLOnePar(buf, xs, runparams,
777 text.outerFont(distance(begin, pbegin)));
782 } // end anonymous namespace
785 void xhtmlParagraphs(Text const & text,
788 OutputParams const & runparams)
790 ParagraphList const & paragraphs = text.paragraphs();
791 ParagraphList::const_iterator par = paragraphs.begin();
792 ParagraphList::const_iterator pend = paragraphs.end();
794 OutputParams ourparams = runparams;
795 while (par != pend) {
796 Layout const & style = par->layout();
797 ParagraphList::const_iterator lastpar = par;
798 ParagraphList::const_iterator send;
800 switch (style.latextype) {
801 case LATEX_COMMAND: {
802 // The files with which we are working never have more than
803 // one paragraph in a command structure.
805 // if (ourparams.html_in_par)
806 // fix it so we don't get sections inside standard, e.g.
807 // note that we may then need to make runparams not const, so we
808 // can communicate that back.
809 // FIXME Maybe this fix should be in the routines themselves, in case
810 // they are called from elsewhere.
811 makeCommand(buf, xs, ourparams, text, par);
815 case LATEX_ENVIRONMENT:
816 case LATEX_LIST_ENVIRONMENT:
817 case LATEX_ITEM_ENVIRONMENT: {
818 // FIXME Same fix here.
819 send = searchEnvironmentHtml(par, pend);
820 par = makeEnvironmentHtml(buf, xs, ourparams, text, par, send);
823 case LATEX_BIB_ENVIRONMENT: {
824 // FIXME Same fix here.
825 send = searchEnvironmentHtml(par, pend);
826 par = makeBibliography(buf, xs, ourparams, text, par, send);
829 case LATEX_PARAGRAPH:
830 send = searchParagraphHtml(par, pend);
831 par = makeParagraphs(buf, xs, ourparams, text, par, send);
835 // makeEnvironment may process more than one paragraphs and bypass pend
836 if (distance(lastpar, par) >= distance(lastpar, pend))