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:
50 case XMLStream::ESCAPE_ALL:
54 } else if (c == '>') {
59 case XMLStream::ESCAPE_AND:
70 // escape what needs escaping
71 docstring xmlize(docstring const &str, XMLStream::EscapeSettings e) {
73 docstring::const_iterator it = str.begin();
74 docstring::const_iterator en = str.end();
75 for (; it != en; ++it)
76 d << escapeChar(*it, e);
81 docstring escapeChar(char c, XMLStream::EscapeSettings e) {
82 LATTEST(static_cast<unsigned char>(c) < 0x80);
83 return escapeChar(static_cast<char_type>(c), e);
87 docstring cleanAttr(docstring const & str)
90 docstring::const_iterator it = str.begin();
91 docstring::const_iterator en = str.end();
92 for (; it != en; ++it) {
93 char_type const c = *it;
94 newname += isAlnumASCII(c) ? c : char_type('_');
100 docstring StartTag::writeTag() const {
101 docstring output = '<' + tag_;
102 if (!attr_.empty()) {
103 docstring attributes = xml::xmlize(attr_, XMLStream::ESCAPE_NONE);
104 attributes.erase(attributes.begin(), std::find_if(attributes.begin(), attributes.end(),
105 [](int c) {return !std::isspace(c);}));
106 if (!attributes.empty()) {
107 output += ' ' + attributes;
115 docstring StartTag::writeEndTag() const {
116 return from_utf8("</") + tag_ + from_utf8(">");
120 bool StartTag::operator==(FontTag const &rhs) const {
125 docstring EndTag::writeEndTag() const {
126 return from_utf8("</") + tag_ + from_utf8(">");
130 docstring CompTag::writeTag() const {
131 docstring output = '<' + from_utf8(tag_);
132 if (!attr_.empty()) {
133 // Erase the beginning of the attributes if it contains space characters: this function deals with that
135 docstring attributes = xmlize(from_utf8(attr_), XMLStream::ESCAPE_NONE);
136 attributes.erase(attributes.begin(), std::find_if(attributes.begin(), attributes.end(),
137 [](int c) {return !std::isspace(c);}));
138 if (!attributes.empty()) {
139 output += ' ' + attributes;
147 bool FontTag::operator==(StartTag const & tag) const
149 FontTag const * const ftag = tag.asFontTag();
152 return (font_type_ == ftag->font_type_);
158 void XMLStream::writeError(std::string const &s) const {
160 os_ << from_utf8("<!-- Output Error: " + s + " -->\n");
164 void XMLStream::writeError(docstring const &s) const {
166 os_ << from_utf8("<!-- Output Error: ") << s << from_utf8(" -->\n");
170 bool XMLStream::closeFontTags() {
171 if (isTagPending(xml::parsep_tag))
172 // we haven't had any content
175 // this may be a useless check, since we ought at least to have
176 // the parsep_tag. but it can't hurt too much to be careful.
177 if (tag_stack_.empty())
180 // first, we close any open font tags we can close
181 TagPtr *curtag = &tag_stack_.back();
182 while ((*curtag)->asFontTag()) {
183 if (**curtag != xml::parsep_tag)
184 os_ << (*curtag)->writeEndTag();
185 tag_stack_.pop_back();
186 // this shouldn't happen, since then the font tags
187 // weren't in any other tag.
188 // LASSERT(!tag_stack_.empty(), return true);
189 if (tag_stack_.empty())
191 curtag = &tag_stack_.back();
194 if (**curtag == xml::parsep_tag)
197 // so we've hit a non-font tag.
198 writeError("Tags still open in closeFontTags(). Probably not a problem,\n"
199 "but you might want to check these tags:");
200 TagDeque::const_reverse_iterator it = tag_stack_.rbegin();
201 TagDeque::const_reverse_iterator const en = tag_stack_.rend();
202 for (; it != en; ++it) {
203 if (**it == xml::parsep_tag)
205 writeError((*it)->tag_);
211 void XMLStream::startDivision(bool keep_empty) {
212 pending_tags_.push_back(makeTagPtr(xml::StartTag(xml::parsep_tag)));
218 void XMLStream::endDivision() {
219 if (isTagPending(xml::parsep_tag)) {
220 // this case is normal. it just means we didn't have content,
221 // so the parsep_tag never got moved onto the tag stack.
222 while (!pending_tags_.empty()) {
223 // clear all pending tags up to and including the parsep tag.
224 // note that we work from the back, because we want to get rid
225 // of everything that hasn't been used.
226 TagPtr const cur_tag = pending_tags_.back();
227 pending_tags_.pop_back();
228 if (*cur_tag == xml::parsep_tag)
233 dumpTagStack("EndDivision");
239 if (!isTagOpen(xml::parsep_tag)) {
240 writeError("No division separation tag found in endDivision().");
244 // this case is also normal, if the parsep tag is the last one
245 // on the stack. otherwise, it's an error.
246 while (!tag_stack_.empty()) {
247 TagPtr const cur_tag = tag_stack_.back();
248 tag_stack_.pop_back();
249 if (*cur_tag == xml::parsep_tag)
251 writeError("Tag `" + cur_tag->tag_ + "' still open at end of paragraph. Closing.");
252 os_ << cur_tag->writeEndTag();
256 dumpTagStack("EndDivision");
261 void XMLStream::clearTagDeque() {
262 while (!pending_tags_.empty()) {
263 TagPtr const & tag = pending_tags_.front();
264 if (*tag != xml::parsep_tag)
266 os_ << tag->writeTag();
267 tag_stack_.push_back(tag);
268 pending_tags_.pop_front();
273 XMLStream &XMLStream::operator<<(docstring const &d) {
275 os_ << xml::xmlize(d, escape_);
276 escape_ = ESCAPE_ALL;
281 XMLStream &XMLStream::operator<<(const char *s) {
283 docstring const d = from_ascii(s);
284 os_ << xml::xmlize(d, escape_);
285 escape_ = ESCAPE_ALL;
290 XMLStream &XMLStream::operator<<(char_type c) {
292 os_ << xml::escapeChar(c, escape_);
293 escape_ = ESCAPE_ALL;
298 XMLStream &XMLStream::operator<<(char c) {
300 os_ << xml::escapeChar(c, escape_);
301 escape_ = ESCAPE_ALL;
306 XMLStream &XMLStream::operator<<(int i) {
309 escape_ = ESCAPE_ALL;
314 XMLStream &XMLStream::operator<<(EscapeSettings e) {
320 XMLStream &XMLStream::operator<<(xml::StartTag const &tag) {
321 if (tag.tag_.empty())
323 pending_tags_.push_back(makeTagPtr(tag));
330 XMLStream &XMLStream::operator<<(xml::ParTag const &tag) {
331 if (tag.tag_.empty())
333 pending_tags_.push_back(makeTagPtr(tag));
338 XMLStream &XMLStream::operator<<(xml::CompTag const &tag) {
339 if (tag.tag_.empty())
342 os_ << tag.writeTag();
347 XMLStream &XMLStream::operator<<(xml::FontTag const &tag) {
348 if (tag.tag_.empty())
350 pending_tags_.push_back(makeTagPtr(tag));
355 XMLStream &XMLStream::operator<<(xml::CR const &) {
357 os_ << from_ascii("\n");
362 bool XMLStream::isTagOpen(xml::StartTag const &stag, int maxdepth) const {
363 auto sit = tag_stack_.begin();
364 auto sen = tag_stack_.cend();
365 for (; sit != sen && maxdepth != 0; ++sit) {
374 bool XMLStream::isTagOpen(xml::EndTag const &etag, int maxdepth) const {
375 auto sit = tag_stack_.begin();
376 auto sen = tag_stack_.cend();
377 for (; sit != sen && maxdepth != 0; ++sit) {
386 bool XMLStream::isTagPending(xml::StartTag const &stag, int maxdepth) const {
387 auto sit = pending_tags_.begin();
388 auto sen = pending_tags_.cend();
389 for (; sit != sen && maxdepth != 0; ++sit) {
398 // this is complicated, because we want to make sure that
399 // everything is properly nested. the code ought to make
400 // sure of that, but we won't assert (yet) if we run into
401 // a problem. we'll just output error messages and try our
402 // best to make things work.
403 XMLStream &XMLStream::operator<<(xml::EndTag const &etag) {
404 if (etag.tag_.empty())
407 // if this tag is pending, we can simply discard it.
408 if (!pending_tags_.empty()) {
409 if (etag == *pending_tags_.back()) {
410 // we have <tag></tag>, so we discard it and remove it
411 // from the pending_tags_.
412 pending_tags_.pop_back();
416 // there is a pending tag that isn't the one we are trying
419 // is this tag itself pending?
420 // non-const iterators because we may call erase().
421 TagDeque::iterator dit = pending_tags_.begin();
422 TagDeque::iterator const den = pending_tags_.end();
423 for (; dit != den; ++dit) {
425 // it was pending, so we just erase it
426 writeError("Tried to close pending tag `" + to_utf8(etag.tag_)
427 + "' when other tags were pending. Last pending tag is `"
428 + to_utf8(pending_tags_.back()->writeTag())
429 + "'. Tag discarded.");
430 pending_tags_.erase(dit);
434 // so etag isn't itself pending. is it even open?
435 if (!isTagOpen(etag)) {
436 writeError("Tried to close `" + to_utf8(etag.tag_)
437 + "' when tag was not open. Tag discarded.");
440 // ok, so etag is open.
441 // our strategy will be as below: we will do what we need to
442 // do to close this tag.
443 string estr = "Closing tag `" + to_utf8(etag.tag_)
444 + "' when other tags are pending. Discarded pending tags:\n";
445 for (dit = pending_tags_.begin(); dit != den; ++dit)
446 estr += to_utf8(xml::xmlize((*dit)->writeTag(), XMLStream::ESCAPE_ALL)) + "\n";
448 // clear the pending tags...
449 pending_tags_.clear();
450 // ...and then just fall through.
453 // make sure there are tags to be closed
454 if (tag_stack_.empty()) {
455 writeError("Tried to close `" + etag.tag_
456 + "' when no tags were open!");
460 // is the tag we are closing the last one we opened?
461 if (etag == *tag_stack_.back()) {
463 os_ << etag.writeEndTag();
464 // ...and forget about it
465 tag_stack_.pop_back();
469 // we are trying to close a tag other than the one last opened.
470 // let's first see if this particular tag is still open somehow.
471 if (!isTagOpen(etag)) {
472 writeError("Tried to close `" + etag.tag_
473 + "' when tag was not open. Tag discarded.");
477 // so the tag was opened, but other tags have been opened since
478 // and not yet closed.
479 // if it's a font tag, though...
480 if (etag.asFontTag()) {
481 // it won't be a problem if the other tags open since this one
482 // are also font tags.
483 TagDeque::const_reverse_iterator rit = tag_stack_.rbegin();
484 TagDeque::const_reverse_iterator ren = tag_stack_.rend();
485 for (; rit != ren; ++rit) {
488 if (!(*rit)->asFontTag()) {
489 // we'll just leave it and, presumably, have to close it later.
490 writeError("Unable to close font tag `" + etag.tag_
491 + "' due to open non-font tag `" + (*rit)->tag_ + "'.");
497 // <em>this is <strong>bold
498 // and are being asked to closed em. we want:
499 // <em>this is <strong>bold</strong></em><strong>
500 // first, we close the intervening tags...
501 TagPtr *curtag = &tag_stack_.back();
502 // ...remembering them in a stack.
504 while (etag != **curtag) {
505 os_ << (*curtag)->writeEndTag();
506 fontstack.push_back(*curtag);
507 tag_stack_.pop_back();
508 curtag = &tag_stack_.back();
510 os_ << etag.writeEndTag();
511 tag_stack_.pop_back();
513 // ...and restore the other tags.
514 rit = fontstack.rbegin();
515 ren = fontstack.rend();
516 for (; rit != ren; ++rit)
517 pending_tags_.push_back(*rit);
521 // it wasn't a font tag.
522 // so other tags were opened before this one and not properly closed.
523 // so we'll close them, too. that may cause other issues later, but it
524 // at least guarantees proper nesting.
525 writeError("Closing tag `" + etag.tag_
526 + "' when other tags are open, namely:");
527 TagPtr *curtag = &tag_stack_.back();
528 while (etag != **curtag) {
529 writeError((*curtag)->tag_);
530 if (**curtag != xml::parsep_tag)
531 os_ << (*curtag)->writeEndTag();
532 tag_stack_.pop_back();
533 curtag = &tag_stack_.back();
535 // curtag is now the one we actually want.
536 os_ << (*curtag)->writeEndTag();
537 tag_stack_.pop_back();
543 docstring xml::escapeString(docstring const & raw, XMLStream::EscapeSettings e)
546 bin.reserve(raw.size() * 2); // crude approximation is sufficient
547 for (size_t i = 0; i != raw.size(); ++i)
548 bin += xml::escapeChar(raw[i], e);
554 docstring const xml::uniqueID(docstring const & label)
557 static atomic_uint seed(1000);
558 return label + convert<docstring>(++seed);
562 docstring xml::cleanID(docstring const & orig)
564 // The standard xml:id only allows letters,
565 // digits, '-' and '.' in a name.
566 // This routine replaces illegal characters by '-' or '.'
567 // and adds a number for uniqueness if need be.
568 docstring const allowed = from_ascii(".-_");
570 // Use a cache of already mangled names: the alterations may merge several IDs as one. This ensures that the IDs
571 // are not mixed up in the document.
572 typedef map<docstring, docstring> MangledMap;
573 static QThreadStorage<MangledMap> tMangledNames;
574 static QThreadStorage<int> tMangleID;
576 MangledMap & mangledNames = tMangledNames.localData();
578 // If the name is already known, just return it.
579 MangledMap::const_iterator const known = mangledNames.find(orig);
580 if (known != mangledNames.end())
581 return known->second;
583 // Start creating the mangled name by iterating over the characters.
585 docstring::const_iterator it = orig.begin();
586 docstring::const_iterator end = orig.end();
588 // Make sure it starts with a letter.
589 if (!isAlphaASCII(*it) && allowed.find(*it) >= allowed.size())
593 bool mangle = false; // Indicates whether the ID had to be changed, i.e. if ID no more ensured to be unique.
594 for (; it != end; ++it) {
596 if (isAlphaASCII(c) || isDigitASCII(c) || c == '-' || c == '.'
597 || allowed.find(c) < allowed.size())
599 else if (c == '_' || c == ' ') {
603 else if (c == ':' || c == ',' || c == ';' || c == '!') {
614 int & mangleID = tMangleID.localData();
615 content += "-" + convert<docstring>(mangleID++);
618 mangledNames[orig] = content;
624 void xml::openTag(odocstream & os, string const & name, string const & attribute)
627 // This should be fixed in layout files later.
628 string param = subst(attribute, "<", "\"");
629 param = subst(param, ">", "\"");
631 // Note: we ignore the name if it empty or if it is a comment "<!-- -->" or
632 // if the name is *dummy*.
633 // We ignore dummy because dummy is not a valid docbook element and it is
634 // the internal name given to single paragraphs in the latex output.
635 // This allow us to simplify the code a lot and is a reasonable compromise.
636 if (!name.empty() && name != "!-- --" && name != "dummy") {
637 os << '<' << from_ascii(name);
639 os << ' ' << from_ascii(param);
645 void xml::closeTag(odocstream & os, string const & name)
647 if (!name.empty() && name != "!-- --" && name != "dummy")
648 os << "</" << from_ascii(name) << '>';
652 void xml::openTag(Buffer const & buf, odocstream & os,
653 OutputParams const & runparams, Paragraph const & par)
655 Layout const & style = par.layout();
656 string const & name = style.latexname();
657 string param = style.latexparam();
658 Counters & counters = buf.params().documentClass().counters();
660 string id = par.getID(buf, runparams);
664 if (param.find('#') != string::npos) {
665 string::size_type pos = param.find("id=<");
666 string::size_type end = param.find(">");
667 if( pos != string::npos && end != string::npos)
668 param.erase(pos, end-pos + 1);
670 attribute = id + ' ' + param;
672 if (param.find('#') != string::npos) {
674 if (!style.counter.empty())
675 // This uses InternalUpdate at the moment becuase xml output
676 // does not do anything with tracked counters, and it would need
677 // to track layouts if it did want to use them.
678 counters.step(style.counter, InternalUpdate);
680 counters.step(from_ascii(name), InternalUpdate);
681 int i = counters.value(from_ascii(name));
682 attribute = subst(param, "#", convert<string>(i));
687 openTag(os, name, attribute);
691 void xml::closeTag(odocstream & os, Paragraph const & par)
693 Layout const & style = par.layout();
694 closeTag(os, style.latexname());