3 * This file is part of LyX, the document processor.
4 * License details can be found in the file COPYING.
9 * Full author contact details are available in file CREDITS.
17 #include "BufferParams.h"
20 #include "OutputParams.h"
21 #include "Paragraph.h"
23 #include "TextClass.h"
25 #include "support/convert.h"
26 #include "support/docstream.h"
27 #include "support/lassert.h"
28 #include "support/lstrings.h"
29 #include "support/textutils.h"
34 #include <QThreadStorage>
37 using namespace lyx::support;
43 docstring escapeChar(char_type c, XMLStream::EscapeSettings e)
46 switch (e) { // For HTML: always ESCAPE_NONE. For XML: it depends, hence the parameter.
47 case XMLStream::ESCAPE_NONE:
48 case XMLStream::ESCAPE_COMMENTS:
51 case XMLStream::ESCAPE_ALL:
55 } else if (c == '>') {
60 case XMLStream::ESCAPE_AND:
71 docstring escapeChar(char c, XMLStream::EscapeSettings e)
73 LATTEST(static_cast<unsigned char>(c) < 0x80);
74 return escapeChar(static_cast<char_type>(c), e);
78 docstring escapeString(docstring const & raw, XMLStream::EscapeSettings e)
81 bin.reserve(raw.size() * 2); // crude approximation is sufficient
82 for (size_t i = 0; i != raw.size(); ++i) {
84 if (e == XMLStream::ESCAPE_COMMENTS && c == '-' && i > 0 && raw[i - 1] == '-')
87 bin += xml::escapeChar(c, e);
94 docstring cleanAttr(docstring const & str)
97 docstring::const_iterator it = str.begin();
98 docstring::const_iterator en = str.end();
99 for (; it != en; ++it) {
100 char_type const c = *it;
101 newname += isAlnumASCII(c) ? c : char_type('_');
107 docstring StartTag::writeTag() const
109 docstring output = '<' + tag_;
110 if (!attr_.empty()) {
111 docstring attributes = xml::escapeString(attr_, XMLStream::ESCAPE_NONE);
112 attributes.erase(attributes.begin(), std::find_if(attributes.begin(), attributes.end(),
113 [](int c) {return !std::isspace(c);}));
114 if (!attributes.empty()) {
115 output += ' ' + attributes;
123 docstring StartTag::writeEndTag() const
125 return from_utf8("</") + tag_ + from_utf8(">");
129 bool StartTag::operator==(FontTag const &rhs) const
135 docstring EndTag::writeEndTag() const
137 return from_utf8("</") + tag_ + from_utf8(">");
141 docstring CompTag::writeTag() const
143 docstring output = '<' + tag_;
144 if (!attr_.empty()) {
145 // Erase the beginning of the attributes if it contains space characters: this function deals with that
147 docstring attributes = escapeString(attr_, XMLStream::ESCAPE_NONE);
148 attributes.erase(attributes.begin(), std::find_if(attributes.begin(), attributes.end(),
149 [](int c) {return !std::isspace(c);}));
150 if (!attributes.empty()) {
151 output += ' ' + attributes;
159 bool FontTag::operator==(StartTag const & tag) const
161 FontTag const * const ftag = tag.asFontTag();
164 return (font_type_ == ftag->font_type_);
170 void XMLStream::writeError(std::string const &s)
173 *this << ESCAPE_NONE << from_utf8("<!-- Output Error: " + s + " -->");
178 void XMLStream::writeError(docstring const &s)
181 *this << ESCAPE_NONE << from_utf8("<!-- Output Error: ");
183 *this << ESCAPE_NONE << from_utf8(" -->");
188 XMLStream::TagPtr XMLStream::getLastStackTag()
190 return tag_stack_.back();
194 bool XMLStream::closeFontTags()
196 if (isTagPending(xml::parsep_tag))
197 // we haven't had any content
200 // this may be a useless check, since we ought at least to have
201 // the parsep_tag. but it can't hurt too much to be careful.
202 if (tag_stack_.empty())
205 // first, we close any open font tags we can close
206 TagPtr *curtag = &tag_stack_.back();
207 while ((*curtag)->asFontTag()) {
208 if (**curtag != xml::parsep_tag)
209 os_ << (*curtag)->writeEndTag();
210 tag_stack_.pop_back();
211 if (tag_stack_.empty())
213 curtag = &tag_stack_.back();
216 if (**curtag == xml::parsep_tag)
219 // so we've hit a non-font tag.
220 writeError("Tags still open in closeFontTags(). Probably not a problem,\n"
221 "but you might want to check these tags:");
222 TagDeque::const_reverse_iterator it = tag_stack_.rbegin();
223 TagDeque::const_reverse_iterator const en = tag_stack_.rend();
224 for (; it != en; ++it) {
225 if (**it == xml::parsep_tag)
227 writeError((*it)->tag_);
233 void XMLStream::startDivision(bool keep_empty)
235 pending_tags_.push_back(makeTagPtr(xml::StartTag(xml::parsep_tag)));
241 void XMLStream::endDivision()
243 if (isTagPending(xml::parsep_tag)) {
244 // this case is normal. it just means we didn't have content,
245 // so the parsep_tag never got moved onto the tag stack.
246 while (!pending_tags_.empty()) {
247 // clear all pending tags up to and including the parsep tag.
248 // note that we work from the back, because we want to get rid
249 // of everything that hasn't been used.
250 TagPtr const cur_tag = pending_tags_.back();
251 pending_tags_.pop_back();
252 if (*cur_tag == xml::parsep_tag)
257 dumpTagStack("EndDivision");
263 if (!isTagOpen(xml::parsep_tag)) {
264 writeError("No division separation tag found in endDivision().");
268 // this case is also normal, if the parsep tag is the last one
269 // on the stack. otherwise, it's an error.
270 while (!tag_stack_.empty()) {
271 TagPtr const cur_tag = tag_stack_.back();
272 tag_stack_.pop_back();
273 if (*cur_tag == xml::parsep_tag)
275 writeError("Tag `" + cur_tag->tag_ + "' still open at end of paragraph. Closing.");
276 os_ << cur_tag->writeEndTag();
280 dumpTagStack("EndDivision");
285 void XMLStream::clearTagDeque()
287 while (!pending_tags_.empty()) {
288 TagPtr const & tag = pending_tags_.front();
289 if (*tag != xml::parsep_tag)
291 os_ << tag->writeTag();
292 tag_stack_.push_back(tag);
293 pending_tags_.pop_front();
298 XMLStream &XMLStream::operator<<(docstring const &d)
300 is_last_tag_cr_ = false;
302 os_ << xml::escapeString(d, escape_);
303 escape_ = ESCAPE_ALL;
308 XMLStream &XMLStream::operator<<(const char *s)
310 is_last_tag_cr_ = false;
312 docstring const d = from_ascii(s);
313 os_ << xml::escapeString(d, escape_);
314 escape_ = ESCAPE_ALL;
319 XMLStream &XMLStream::operator<<(char_type c)
321 is_last_tag_cr_ = false;
323 os_ << xml::escapeChar(c, escape_);
324 escape_ = ESCAPE_ALL;
329 XMLStream &XMLStream::operator<<(char c)
331 is_last_tag_cr_ = false;
333 os_ << xml::escapeChar(c, escape_);
334 escape_ = ESCAPE_ALL;
339 XMLStream &XMLStream::operator<<(int i)
341 is_last_tag_cr_ = false;
344 escape_ = ESCAPE_ALL;
349 XMLStream &XMLStream::operator<<(EscapeSettings e)
351 // Don't update is_last_tag_cr_ here, as this does not output anything.
357 XMLStream &XMLStream::operator<<(xml::StartTag const &tag)
359 is_last_tag_cr_ = false;
360 if (tag.tag_.empty())
362 pending_tags_.push_back(makeTagPtr(tag));
369 XMLStream &XMLStream::operator<<(xml::ParTag const &tag)
371 is_last_tag_cr_ = false;
372 if (tag.tag_.empty())
374 pending_tags_.push_back(makeTagPtr(tag));
379 XMLStream &XMLStream::operator<<(xml::CompTag const &tag)
381 is_last_tag_cr_ = false;
382 if (tag.tag_.empty())
385 os_ << tag.writeTag();
390 XMLStream &XMLStream::operator<<(xml::FontTag const &tag)
392 is_last_tag_cr_ = false;
393 if (tag.tag_.empty())
395 pending_tags_.push_back(makeTagPtr(tag));
400 XMLStream &XMLStream::operator<<(xml::CR const &)
402 is_last_tag_cr_ = true;
404 os_ << from_ascii("\n");
409 bool XMLStream::isTagOpen(xml::StartTag const &stag, int maxdepth) const
411 auto sit = tag_stack_.begin();
412 auto sen = tag_stack_.cend();
413 for (; sit != sen && maxdepth != 0; ++sit) {
422 bool XMLStream::isTagOpen(xml::EndTag const &etag, int maxdepth) const
424 auto sit = tag_stack_.begin();
425 auto sen = tag_stack_.cend();
426 for (; sit != sen && maxdepth != 0; ++sit) {
435 bool XMLStream::isTagPending(xml::StartTag const &stag, int maxdepth) const
437 auto sit = pending_tags_.begin();
438 auto sen = pending_tags_.cend();
439 for (; sit != sen && maxdepth != 0; ++sit) {
448 // this is complicated, because we want to make sure that
449 // everything is properly nested. the code ought to make
450 // sure of that, but we won't assert (yet) if we run into
451 // a problem. we'll just output error messages and try our
452 // best to make things work.
453 XMLStream &XMLStream::operator<<(xml::EndTag const &etag)
455 is_last_tag_cr_ = false;
457 if (etag.tag_.empty())
460 // if this tag is pending, we can simply discard it.
461 if (!pending_tags_.empty()) {
462 if (etag == *pending_tags_.back()) {
463 // we have <tag></tag>, so we discard it and remove it
464 // from the pending_tags_.
465 pending_tags_.pop_back();
469 // there is a pending tag that isn't the one we are trying
472 // is this tag itself pending?
473 // non-const iterators because we may call erase().
474 TagDeque::iterator dit = pending_tags_.begin();
475 TagDeque::iterator const den = pending_tags_.end();
476 for (; dit != den; ++dit) {
478 // it was pending, so we just erase it
479 writeError("Tried to close pending tag `" + to_utf8(etag.tag_)
480 + "' when other tags were pending. Last pending tag is `"
481 + to_utf8(pending_tags_.back()->writeTag())
482 + "'. Tag discarded.");
483 pending_tags_.erase(dit);
487 // so etag isn't itself pending. is it even open?
488 if (!isTagOpen(etag)) {
489 writeError("Tried to close `" + to_utf8(etag.tag_)
490 + "' when tag was not open. Tag discarded.");
493 // ok, so etag is open.
494 // our strategy will be as below: we will do what we need to
495 // do to close this tag.
496 string estr = "Closing tag `" + to_utf8(etag.tag_)
497 + "' when other tags are pending. Discarded pending tags:\n";
498 for (dit = pending_tags_.begin(); dit != den; ++dit)
499 estr += to_utf8(xml::escapeString((*dit)->writeTag(), XMLStream::ESCAPE_ALL)) + "\n";
501 // clear the pending tags...
502 pending_tags_.clear();
503 // ...and then just fall through.
506 // make sure there are tags to be closed
507 if (tag_stack_.empty()) {
508 writeError("Tried to close `" + etag.tag_
509 + "' when no tags were open!");
513 // is the tag we are closing the last one we opened?
514 if (etag == *tag_stack_.back()) {
516 os_ << etag.writeEndTag();
517 // ...and forget about it
518 tag_stack_.pop_back();
522 // we are trying to close a tag other than the one last opened.
523 // let's first see if this particular tag is still open somehow.
524 if (!isTagOpen(etag)) {
525 writeError("Tried to close `" + etag.tag_
526 + "' when tag was not open. Tag discarded.");
530 // so the tag was opened, but other tags have been opened since
531 // and not yet closed.
532 // if it's a font tag, though...
533 if (etag.asFontTag()) {
534 // it won't be a problem if the other tags open since this one
535 // are also font tags.
536 TagDeque::const_reverse_iterator rit = tag_stack_.rbegin();
537 TagDeque::const_reverse_iterator ren = tag_stack_.rend();
538 for (; rit != ren; ++rit) {
541 if (!(*rit)->asFontTag()) {
542 // we'll just leave it and, presumably, have to close it later.
543 writeError("Unable to close font tag `" + etag.tag_
544 + "' due to open non-font tag `" + (*rit)->tag_ + "'.");
550 // <em>this is <strong>bold
551 // and are being asked to closed em. we want:
552 // <em>this is <strong>bold</strong></em><strong>
553 // first, we close the intervening tags...
554 TagPtr *curtag = &tag_stack_.back();
555 // ...remembering them in a stack.
557 while (etag != **curtag) {
558 os_ << (*curtag)->writeEndTag();
559 fontstack.push_back(*curtag);
560 tag_stack_.pop_back();
561 curtag = &tag_stack_.back();
563 os_ << etag.writeEndTag();
564 tag_stack_.pop_back();
566 // ...and restore the other tags.
567 rit = fontstack.rbegin();
568 ren = fontstack.rend();
569 for (; rit != ren; ++rit)
570 pending_tags_.push_back(*rit);
574 // it wasn't a font tag.
575 // so other tags were opened before this one and not properly closed.
576 // so we'll close them, too. that may cause other issues later, but it
577 // at least guarantees proper nesting.
578 writeError("Closing tag `" + etag.tag_
579 + "' when other tags are open, namely:");
580 TagPtr *curtag = &tag_stack_.back();
581 while (etag != **curtag) {
582 writeError((*curtag)->tag_);
583 if (**curtag != xml::parsep_tag)
584 os_ << (*curtag)->writeEndTag();
585 tag_stack_.pop_back();
586 curtag = &tag_stack_.back();
588 // curtag is now the one we actually want.
589 os_ << (*curtag)->writeEndTag();
590 tag_stack_.pop_back();
596 docstring xml::uniqueID(docstring const & label)
599 static atomic_uint seed(1000);
600 return label + convert<docstring>(++seed);
604 docstring xml::cleanID(docstring const & orig)
606 // The standard xml:id only allows letters, digits, '-' and '.' in a name.
607 // This routine replaces illegal characters by '-' or '.' and adds a number for uniqueness if need be.
609 // Use a cache of already mangled names: the alterations may merge several IDs as one. This ensures that the IDs
610 // are not mixed up in the document.
611 // This code could be improved: it uses Qt outside the GUI part. Any TLS implementation could do the trick.
612 typedef map<docstring, docstring> MangledMap;
613 static QThreadStorage<MangledMap> tMangledNames;
614 static QThreadStorage<int> tMangleID;
616 // If the name is already known, just return it.
617 MangledMap & mangledNames = tMangledNames.localData();
618 auto const known = mangledNames.find(orig);
619 if (known != mangledNames.end())
620 return known->second;
622 // Start creating the mangled name by iterating over the characters.
624 auto it = orig.cbegin();
625 auto end = orig.cend();
627 // Make sure it starts with a letter.
628 if (!isAlphaASCII(*it))
631 // Parse the ID character by character and change what needs to.
632 bool mangle = false; // Indicates whether the ID had to be changed, i.e. if ID no more ensured to be unique.
633 for (; it != end; ++it) {
635 if (isAlphaASCII(c) || isDigitASCII(c) || c == '-' || c == '.' || c == '_') {
637 } else if (c == ':' || c == ',' || c == ';' || c == '!') {
640 } else { // Other invalid characters, such as ' '.
646 // If there had to be a change, check if ID unicity is still guaranteed.
647 // This avoids having a clash if satisfying XML requirements for ID makes two IDs identical, like "a:b" and "a!b",
648 // as both of them would be transformed as "a.b". With this procedure, one will become "a.b" and the other "a.b-1".
649 if (mangle && mangledNames.find(content) != mangledNames.end()) {
650 int & mangleID = tMangleID.localData();
652 content += "-" + convert<docstring>(mangleID);
656 // Save the new ID to avoid recomputing it afterwards and to ensure stability over the document.
657 mangledNames[orig] = content;
662 void xml::openTag(odocstream & os, string const & name, string const & attribute)
665 // This should be fixed in layout files later.
666 string param = subst(attribute, "<", "\"");
667 param = subst(param, ">", "\"");
669 // Note: we ignore the name if it empty or if it is a comment "<!-- -->" or
670 // if the name is *dummy*.
671 // We ignore dummy because dummy is not a valid docbook element and it is
672 // the internal name given to single paragraphs in the latex output.
673 // This allow us to simplify the code a lot and is a reasonable compromise.
674 if (!name.empty() && name != "!-- --" && name != "dummy") {
675 os << '<' << from_ascii(name);
677 os << ' ' << from_ascii(param);
683 void xml::closeTag(odocstream & os, string const & name)
685 if (!name.empty() && name != "!-- --" && name != "dummy")
686 os << "</" << from_ascii(name) << '>';
690 void xml::openTag(Buffer const & buf, odocstream & os,
691 OutputParams const & runparams, Paragraph const & par)
693 Layout const & style = par.layout();
694 string const & name = style.latexname();
695 string param = style.latexparam();
696 Counters & counters = buf.params().documentClass().counters();
698 string id = par.getID(buf, runparams);
702 if (param.find('#') != string::npos) {
703 string::size_type pos = param.find("id=<");
704 string::size_type end = param.find(">");
705 if( pos != string::npos && end != string::npos)
706 param.erase(pos, end-pos + 1);
708 attribute = id + ' ' + param;
710 if (param.find('#') != string::npos) {
712 if (!style.counter.empty())
713 // This uses InternalUpdate at the moment becuase xml output
714 // does not do anything with tracked counters, and it would need
715 // to track layouts if it did want to use them.
716 counters.step(style.counter, InternalUpdate);
718 counters.step(from_ascii(name), InternalUpdate);
719 int i = counters.value(from_ascii(name));
720 attribute = subst(param, "#", convert<string>(i));
725 openTag(os, name, attribute);
729 void xml::closeTag(odocstream & os, Paragraph const & par)
731 Layout const & style = par.layout();
732 closeTag(os, style.latexname());