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 = '<' + from_utf8(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(from_utf8(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) const
173 os_ << from_utf8("<!-- Output Error: " + s + " -->\n");
177 void XMLStream::writeError(docstring const &s) const
180 os_ << from_utf8("<!-- Output Error: ") << s << from_utf8(" -->\n");
184 bool XMLStream::closeFontTags()
186 if (isTagPending(xml::parsep_tag))
187 // we haven't had any content
190 // this may be a useless check, since we ought at least to have
191 // the parsep_tag. but it can't hurt too much to be careful.
192 if (tag_stack_.empty())
195 // first, we close any open font tags we can close
196 TagPtr *curtag = &tag_stack_.back();
197 while ((*curtag)->asFontTag()) {
198 if (**curtag != xml::parsep_tag)
199 os_ << (*curtag)->writeEndTag();
200 tag_stack_.pop_back();
201 // this shouldn't happen, since then the font tags
202 // weren't in any other tag.
203 LASSERT(!tag_stack_.empty(), return true);
204 if (tag_stack_.empty())
206 curtag = &tag_stack_.back();
209 if (**curtag == xml::parsep_tag)
212 // so we've hit a non-font tag.
213 writeError("Tags still open in closeFontTags(). Probably not a problem,\n"
214 "but you might want to check these tags:");
215 TagDeque::const_reverse_iterator it = tag_stack_.rbegin();
216 TagDeque::const_reverse_iterator const en = tag_stack_.rend();
217 for (; it != en; ++it) {
218 if (**it == xml::parsep_tag)
220 writeError((*it)->tag_);
226 void XMLStream::startDivision(bool keep_empty)
228 pending_tags_.push_back(makeTagPtr(xml::StartTag(xml::parsep_tag)));
234 void XMLStream::endDivision()
236 if (isTagPending(xml::parsep_tag)) {
237 // this case is normal. it just means we didn't have content,
238 // so the parsep_tag never got moved onto the tag stack.
239 while (!pending_tags_.empty()) {
240 // clear all pending tags up to and including the parsep tag.
241 // note that we work from the back, because we want to get rid
242 // of everything that hasn't been used.
243 TagPtr const cur_tag = pending_tags_.back();
244 pending_tags_.pop_back();
245 if (*cur_tag == xml::parsep_tag)
250 dumpTagStack("EndDivision");
256 if (!isTagOpen(xml::parsep_tag)) {
257 writeError("No division separation tag found in endDivision().");
261 // this case is also normal, if the parsep tag is the last one
262 // on the stack. otherwise, it's an error.
263 while (!tag_stack_.empty()) {
264 TagPtr const cur_tag = tag_stack_.back();
265 tag_stack_.pop_back();
266 if (*cur_tag == xml::parsep_tag)
268 writeError("Tag `" + cur_tag->tag_ + "' still open at end of paragraph. Closing.");
269 os_ << cur_tag->writeEndTag();
273 dumpTagStack("EndDivision");
278 void XMLStream::clearTagDeque()
280 while (!pending_tags_.empty()) {
281 TagPtr const & tag = pending_tags_.front();
282 if (*tag != xml::parsep_tag)
284 os_ << tag->writeTag();
285 tag_stack_.push_back(tag);
286 pending_tags_.pop_front();
291 XMLStream &XMLStream::operator<<(docstring const &d)
294 os_ << xml::escapeString(d, escape_);
295 escape_ = ESCAPE_ALL;
300 XMLStream &XMLStream::operator<<(const char *s)
303 docstring const d = from_ascii(s);
304 os_ << xml::escapeString(d, escape_);
305 escape_ = ESCAPE_ALL;
310 XMLStream &XMLStream::operator<<(char_type c)
313 os_ << xml::escapeChar(c, escape_);
314 escape_ = ESCAPE_ALL;
319 XMLStream &XMLStream::operator<<(char c)
322 os_ << xml::escapeChar(c, escape_);
323 escape_ = ESCAPE_ALL;
328 XMLStream &XMLStream::operator<<(int i)
332 escape_ = ESCAPE_ALL;
337 XMLStream &XMLStream::operator<<(EscapeSettings e)
344 XMLStream &XMLStream::operator<<(xml::StartTag const &tag)
346 if (tag.tag_.empty())
348 pending_tags_.push_back(makeTagPtr(tag));
355 XMLStream &XMLStream::operator<<(xml::ParTag const &tag)
357 if (tag.tag_.empty())
359 pending_tags_.push_back(makeTagPtr(tag));
364 XMLStream &XMLStream::operator<<(xml::CompTag const &tag)
366 if (tag.tag_.empty())
369 os_ << tag.writeTag();
374 XMLStream &XMLStream::operator<<(xml::FontTag const &tag)
376 if (tag.tag_.empty())
378 pending_tags_.push_back(makeTagPtr(tag));
383 XMLStream &XMLStream::operator<<(xml::CR const &)
386 os_ << from_ascii("\n");
391 bool XMLStream::isTagOpen(xml::StartTag const &stag, int maxdepth) const
393 auto sit = tag_stack_.begin();
394 auto sen = tag_stack_.cend();
395 for (; sit != sen && maxdepth != 0; ++sit) {
404 bool XMLStream::isTagOpen(xml::EndTag const &etag, int maxdepth) const
406 auto sit = tag_stack_.begin();
407 auto sen = tag_stack_.cend();
408 for (; sit != sen && maxdepth != 0; ++sit) {
417 bool XMLStream::isTagPending(xml::StartTag const &stag, int maxdepth) const
419 auto sit = pending_tags_.begin();
420 auto sen = pending_tags_.cend();
421 for (; sit != sen && maxdepth != 0; ++sit) {
430 // this is complicated, because we want to make sure that
431 // everything is properly nested. the code ought to make
432 // sure of that, but we won't assert (yet) if we run into
433 // a problem. we'll just output error messages and try our
434 // best to make things work.
435 XMLStream &XMLStream::operator<<(xml::EndTag const &etag)
437 if (etag.tag_.empty())
440 // if this tag is pending, we can simply discard it.
441 if (!pending_tags_.empty()) {
442 if (etag == *pending_tags_.back()) {
443 // we have <tag></tag>, so we discard it and remove it
444 // from the pending_tags_.
445 pending_tags_.pop_back();
449 // there is a pending tag that isn't the one we are trying
452 // is this tag itself pending?
453 // non-const iterators because we may call erase().
454 TagDeque::iterator dit = pending_tags_.begin();
455 TagDeque::iterator const den = pending_tags_.end();
456 for (; dit != den; ++dit) {
458 // it was pending, so we just erase it
459 writeError("Tried to close pending tag `" + to_utf8(etag.tag_)
460 + "' when other tags were pending. Last pending tag is `"
461 + to_utf8(pending_tags_.back()->writeTag())
462 + "'. Tag discarded.");
463 pending_tags_.erase(dit);
467 // so etag isn't itself pending. is it even open?
468 if (!isTagOpen(etag)) {
469 writeError("Tried to close `" + to_utf8(etag.tag_)
470 + "' when tag was not open. Tag discarded.");
473 // ok, so etag is open.
474 // our strategy will be as below: we will do what we need to
475 // do to close this tag.
476 string estr = "Closing tag `" + to_utf8(etag.tag_)
477 + "' when other tags are pending. Discarded pending tags:\n";
478 for (dit = pending_tags_.begin(); dit != den; ++dit)
479 estr += to_utf8(xml::escapeString((*dit)->writeTag(), XMLStream::ESCAPE_ALL)) + "\n";
481 // clear the pending tags...
482 pending_tags_.clear();
483 // ...and then just fall through.
486 // make sure there are tags to be closed
487 if (tag_stack_.empty()) {
488 writeError("Tried to close `" + etag.tag_
489 + "' when no tags were open!");
493 // is the tag we are closing the last one we opened?
494 if (etag == *tag_stack_.back()) {
496 os_ << etag.writeEndTag();
497 // ...and forget about it
498 tag_stack_.pop_back();
502 // we are trying to close a tag other than the one last opened.
503 // let's first see if this particular tag is still open somehow.
504 if (!isTagOpen(etag)) {
505 writeError("Tried to close `" + etag.tag_
506 + "' when tag was not open. Tag discarded.");
510 // so the tag was opened, but other tags have been opened since
511 // and not yet closed.
512 // if it's a font tag, though...
513 if (etag.asFontTag()) {
514 // it won't be a problem if the other tags open since this one
515 // are also font tags.
516 TagDeque::const_reverse_iterator rit = tag_stack_.rbegin();
517 TagDeque::const_reverse_iterator ren = tag_stack_.rend();
518 for (; rit != ren; ++rit) {
521 if (!(*rit)->asFontTag()) {
522 // we'll just leave it and, presumably, have to close it later.
523 writeError("Unable to close font tag `" + etag.tag_
524 + "' due to open non-font tag `" + (*rit)->tag_ + "'.");
530 // <em>this is <strong>bold
531 // and are being asked to closed em. we want:
532 // <em>this is <strong>bold</strong></em><strong>
533 // first, we close the intervening tags...
534 TagPtr *curtag = &tag_stack_.back();
535 // ...remembering them in a stack.
537 while (etag != **curtag) {
538 os_ << (*curtag)->writeEndTag();
539 fontstack.push_back(*curtag);
540 tag_stack_.pop_back();
541 curtag = &tag_stack_.back();
543 os_ << etag.writeEndTag();
544 tag_stack_.pop_back();
546 // ...and restore the other tags.
547 rit = fontstack.rbegin();
548 ren = fontstack.rend();
549 for (; rit != ren; ++rit)
550 pending_tags_.push_back(*rit);
554 // it wasn't a font tag.
555 // so other tags were opened before this one and not properly closed.
556 // so we'll close them, too. that may cause other issues later, but it
557 // at least guarantees proper nesting.
558 writeError("Closing tag `" + etag.tag_
559 + "' when other tags are open, namely:");
560 TagPtr *curtag = &tag_stack_.back();
561 while (etag != **curtag) {
562 writeError((*curtag)->tag_);
563 if (**curtag != xml::parsep_tag)
564 os_ << (*curtag)->writeEndTag();
565 tag_stack_.pop_back();
566 curtag = &tag_stack_.back();
568 // curtag is now the one we actually want.
569 os_ << (*curtag)->writeEndTag();
570 tag_stack_.pop_back();
576 docstring xml::uniqueID(docstring const & label)
579 static atomic_uint seed(1000);
580 return label + convert<docstring>(++seed);
584 docstring xml::cleanID(docstring const & orig)
586 // The standard xml:id only allows letters, digits, '-' and '.' in a name.
587 // This routine replaces illegal characters by '-' or '.' and adds a number for uniqueness if need be.
589 // Use a cache of already mangled names: the alterations may merge several IDs as one. This ensures that the IDs
590 // are not mixed up in the document.
591 // This code could be improved: it uses Qt outside the GUI part. Any TLS implementation could do the trick.
592 typedef map<docstring, docstring> MangledMap;
593 static QThreadStorage<MangledMap> tMangledNames;
594 static QThreadStorage<int> tMangleID;
596 // If the name is already known, just return it.
597 MangledMap & mangledNames = tMangledNames.localData();
598 auto const known = mangledNames.find(orig);
599 if (known != mangledNames.end())
600 return known->second;
602 // Start creating the mangled name by iterating over the characters.
604 auto it = orig.cbegin();
605 auto end = orig.cend();
607 // Make sure it starts with a letter.
608 if (!isAlphaASCII(*it))
611 // Parse the ID character by character and change what needs to.
612 bool mangle = false; // Indicates whether the ID had to be changed, i.e. if ID no more ensured to be unique.
613 for (; it != end; ++it) {
615 if (isAlphaASCII(c) || isDigitASCII(c) || c == '-' || c == '.' || c == '_') {
617 } else if (c == ':' || c == ',' || c == ';' || c == '!') {
620 } else { // Other invalid characters, such as ' '.
626 // If there had to be a change, check if ID unicity is still guaranteed.
627 // This avoids having a clash if satisfying XML requirements for ID makes two IDs identical, like "a:b" and "a!b",
628 // as both of them would be transformed as "a.b". With this procedure, one will become "a.b" and the other "a.b-1".
629 if (mangle && mangledNames.find(content) != mangledNames.end()) {
630 int & mangleID = tMangleID.localData();
631 content += "-" + convert<docstring>(mangleID);
635 // Save the new ID to avoid recomputing it afterwards and to ensure stability over the document.
636 mangledNames[orig] = content;
641 void xml::openTag(odocstream & os, string const & name, string const & attribute)
644 // This should be fixed in layout files later.
645 string param = subst(attribute, "<", "\"");
646 param = subst(param, ">", "\"");
648 // Note: we ignore the name if it empty or if it is a comment "<!-- -->" or
649 // if the name is *dummy*.
650 // We ignore dummy because dummy is not a valid docbook element and it is
651 // the internal name given to single paragraphs in the latex output.
652 // This allow us to simplify the code a lot and is a reasonable compromise.
653 if (!name.empty() && name != "!-- --" && name != "dummy") {
654 os << '<' << from_ascii(name);
656 os << ' ' << from_ascii(param);
662 void xml::closeTag(odocstream & os, string const & name)
664 if (!name.empty() && name != "!-- --" && name != "dummy")
665 os << "</" << from_ascii(name) << '>';
669 void xml::openTag(Buffer const & buf, odocstream & os,
670 OutputParams const & runparams, Paragraph const & par)
672 Layout const & style = par.layout();
673 string const & name = style.latexname();
674 string param = style.latexparam();
675 Counters & counters = buf.params().documentClass().counters();
677 string id = par.getID(buf, runparams);
681 if (param.find('#') != string::npos) {
682 string::size_type pos = param.find("id=<");
683 string::size_type end = param.find(">");
684 if( pos != string::npos && end != string::npos)
685 param.erase(pos, end-pos + 1);
687 attribute = id + ' ' + param;
689 if (param.find('#') != string::npos) {
691 if (!style.counter.empty())
692 // This uses InternalUpdate at the moment becuase xml output
693 // does not do anything with tracked counters, and it would need
694 // to track layouts if it did want to use them.
695 counters.step(style.counter, InternalUpdate);
697 counters.step(from_ascii(name), InternalUpdate);
698 int i = counters.value(from_ascii(name));
699 attribute = subst(param, "#", convert<string>(i));
704 openTag(os, name, attribute);
708 void xml::closeTag(odocstream & os, Paragraph const & par)
710 Layout const & style = par.layout();
711 closeTag(os, style.latexname());