path: root/poxml/parser.cpp

// #define POXML_DEBUG

#include "parser.h"
#include <iostream>
#include <stdlib.h>
#include <assert.h>
#include <tqregexp.h>

using namespace std;

static const char *singletags[] = {"beginpage","imagedata", "colspec", "spanspec",
                                   "anchor", "xref", "area",
                                   "footnoteref", "void", "inlinegraphic",
                                   "glosssee", "graphic", "xi:include",
                                   0};
static const char *cuttingtags[] = {"bridgehead", "trans_comment", "para", "title", "term",
                                    "entry", "contrib", "keyword", "example",
                                    "note", "footnote", "caution",
                                    "informalexample", "remark", "comment",
                                    "imageobject", "varlistentry", "thead",
                                    "tbody", "tgroup", "row", "screenshot", "screeninfo",
                                    "variablelist", "step", "procedure",
                                    "step", "holder", "listitem", "important",
                                    "author", "itemizedlist", "orderedlist",
                                    "caption", "textobject", "mediaobject",
                                    "tip", "glossdef", "inlinemediaobject",
                                    "simplelist", "member", "glossentry",
				    "areaspec", "corpauthor", "indexterm",
                                    "calloutlist", "callout", "subtitle",
				    "table", "part", "xi:fallback", "primary",
                                    "secondary", "chapter", "sect1", "sect2",
                                    "figure", "abstract", "sect3", "sect", "sect4",
                                    "warning", "preface", "authorgroup", "keywordset",
				    "informaltable", "qandaentry", "question", "answer",
				    "othercredit", "affiliation", "qandaset",
				    "cmdsynopsis", "funcsynopsis", "funcsynopsisinfo" ,
				    "epigraph", "attribution", "glossary", "chapterinfo",
				    "glossdiv", "blockingquote", "simplesect", "section",
				    "qandadiv", "refsect1", "refmeta", "formalpara",
				    "refentry", "refnamediv", "refpurpose", "refentrytitle",
                                    "refmiscinfo", "refsect2", "refsect3", "refsect1info",
                                    "refsect2info", "refsect3info", "refsection", "refsectioninfo",
                                    "refsynopsisdiv", "refsysnopsisdivinfo", "remark",
                                    "revdescription", "glossentry", "partinfo", 
				    "segmentedlist", "segtitle", "seg", "seglistitem", "screenco",
                                    0};
static const char *literaltags[] = {"literallayout", "synopsis", "screen",
				    "programlisting", 0};

bool StructureParser::fatalError ( const TQXmlParseException &e )
{
    cerr << "fatalError " << e.message().latin1() << " " << e.lineNumber() << " "
         << e.columnNumber() << endl;
    return false;
}

bool StructureParser::startDocument()
{
    infos_reg = TQRegExp("\\s*poxml_line=\"(\\d+)\" poxml_col=\"(\\d+)\"");
    do_not_split_reg = TQRegExp("\\s*condition=\"do-not-split\"");
    message = "";
    inside = 0;
    return true;
}

bool StructureParser::isCuttingTag(const TQString &qName)
{
    int index = 0;
    while (cuttingtags[index]) {
        if (cuttingtags[index] == qName)
            return true;
        index++;
    }
    return isLiteralTag(qName);
}

bool StructureParser::isSingleTag(const TQString &qName)
{
    int index = 0;
    while (singletags[index]) {
        if (singletags[index] == qName)
            return true;
        index++;
    }
    return false;
}

bool StructureParser::isLiteralTag(const TQString &qName)
{
    int index = 0;
    while (literaltags[index]) {
        if (literaltags[index] == qName)
            return true;
        index++;
    }
    return false;
}

bool StructureParser::skippedEntity ( const TQString & name )
{
    if (inside)
        message += TQString("&%1;").arg(name);
    return true;
}

bool StructureParser::startElement( const TQString& , const TQString& ,
                                    const TQString& qName,
                                    const TQXmlAttributes & attr )
{
    TQString tname = qName.lower();

    bool first = false;

    if (isCuttingTag(tname)) {
        if (!inside) {
            message = TQString();
            list.pc.increasePara();
            startline = locator->lineNumber();
            startcol = locator->columnNumber();
            first = true;
        }
        inside++;
    }

    if (inside)
    {
        TQString tmp = "<" + tname;
        for (int i = 0; i < attr.length(); i++) {
            tmp += TQString(" %1=\"%2\"").arg(attr.qName(i)).arg(attr.value(i));
        }
        tmp += TQString(" poxml_line=\"%1\"").arg(locator->lineNumber());
        tmp += TQString(" poxml_col=\"%1\"").arg(locator->columnNumber());

        if (isSingleTag(qName))
            tmp += "/>";
        else
            tmp += ">";
        message += tmp;
        if (first)
            startcol -= message.length();
    }

    if (tname == "anchor" || tname.left(4) == "sect" || tname == "chapter")
        if (!attr.value("id").isEmpty()) list.pc.addAnchor(attr.value("id"));

    return true;
}

bool StructureParser::startCDATA()
{
    if ( inside )
        message += "<![CDATA[";
    return true;
}

bool StructureParser::endCDATA()
{
    if ( inside )
        message += "]]>";
    return true;
}

bool StructureParser::isClosure(const TQString &message)
{
    assert(message.at(0) == '<');
    int endindex = 1;
    while (!message.at(endindex).isSpace() && message.at(endindex) != '>')
        endindex++;
    TQString tag = message.mid(1, endindex - 1);
    return closureTag(message, tag);
}

bool StructureParser::closureTag(const TQString& message, const TQString &tag)
{
#ifdef POXML_DEBUG
    tqDebug("closureTag %s %s", message.latin1(), tag.latin1());
#endif

    int inside = 0;
    uint index = 0;
    while (true)
    {
        int nextclose = message.find(TQRegExp(TQString::fromLatin1("</%1[\\s>]").arg(tag)), index);
        int nextstart = message.find(TQRegExp(TQString::fromLatin1("<%1[>\\s]").arg(tag)), index);
        //  tqDebug("finding %d %d %d %d", nextstart, nextclose, index, inside);
        if (nextclose == -1) {
#ifdef POXML_DEBUG
            tqDebug("ending on no close anymore %d %d %d %d", (!inside && index >= message.length()), inside, index, message.length());
#endif
            return !inside && index >= message.length();
        }
        if (nextstart == -1)
            nextstart = message.length() + 1;

        if (nextstart < nextclose) {
            inside++;
            index = nextstart + 1;
            while (message.at(index) != '>')
                index++;
            index++;
        } else {
            inside--;
            index = nextclose + 1;
            while (message.at(index) != '>')
                index++;
            index++;
            if (!inside) {
#ifdef POXML_DEBUG
                tqDebug("ending on exit %d", index >= message.length());
#endif
                return index >= message.length();
            }
        }
    }
}

void StructureParser::descape(TQString &message)
{
    uint index = 0;
    stripWhiteSpace( message );

    int inside = 0;
    bool lastws = false;

    while (index < message.length()) {
        switch (message.at(index).latin1()) {
            case '\n':
            case '\t':
            case '\r':
                if (!inside)
                    message[index] = ' ';
            case ' ':
                if (!inside && lastws)
                    message[index] = '\010';
                lastws = true;
                break;
            case '<': {
                uint endindex = index+1;
                while (endindex < message.length() && !message.at(endindex).isSpace() &&
                       message.at(endindex) != '>')
                    endindex++;
                TQString tag = message.mid(index + 1, endindex - index - 1);
                if (tag.at(0) == '/') {
                    if (isLiteralTag(tag.mid(1)))
                        inside--;
                } else
                    if (isLiteralTag(tag))
                        inside++;
                break;
            }
            default:
                lastws = false;
        }

        index++;
    }
    message.replace(TQRegExp("\010"), "");
}

bool StructureParser::formatMessage(MsgBlock &msg) const
{
#ifdef POXML_DEBUG
    tqDebug("formatMessage %s", msg.msgid.latin1());
#endif

    int offset = 0;
    bool changed = false;
    bool recurse = true;

    if (msg.msgid.isEmpty())
        return true;

    for (int index = 0; msg.msgid.at(index) == ' '; index++, offset++);
    stripWhiteSpace( msg.msgid );

    // removing starting single tags
    for (int index = 0; singletags[index]; index++)
    {
        int slen = strlen(singletags[index]);

        if (msg.msgid.left(slen + 1) == TQString::fromLatin1("<%1").arg(singletags[index]) &&
            !msg.msgid.at( slen + 1 ).isLetterOrNumber() )
        {
#ifdef POXML_DEBUG
            tqDebug("removing single tag %s", singletags[index]);
#endif
            int strindex = strlen(singletags[index]) + 1;
            while (msg.msgid.at(strindex) != '>')
                strindex++;
            msg.msgid = msg.msgid.mid(strindex + 1);
            changed = true;
            offset += strindex + 1;
            for (int index = 0; msg.msgid.at(index) == ' '; index++, offset++) ;
            stripWhiteSpace( msg.msgid );
        }
    }

    while (msg.msgid.right(2) == "/>")
    {
        int strindex = msg.msgid.length() - 2;
        while (msg.msgid.at(strindex) != '<')
            strindex--;
        msg.msgid = msg.msgid.left(strindex);
        stripWhiteSpace( msg.msgid ); // only removed space at the end
        changed = true;
    }

    for (int index = 0; msg.msgid.at(index) == ' '; index++, offset++) ;
    stripWhiteSpace( msg.msgid );

    while (true) {
        if (msg.msgid.at(0) != '<')
            break;
        if (msg.msgid.at(msg.msgid.length() - 1) != '>')
            break;
        int strindex = 1;
        while (msg.msgid.at(strindex) != ' ' && msg.msgid.at(strindex) != '>')
            strindex++;
        TQString starttag = msg.msgid.mid(1, strindex - 1);
        int endindex = msg.msgid.length() - 2;
        while (msg.msgid.at(endindex) != '<' && msg.msgid.at(endindex + 1) != '/')
            endindex--;
#ifdef POXML_DEBUG
        tqDebug("endIndex %d", endindex);
#endif
        strindex = endindex;
        TQString orig = msg.msgid;

        TQString endtag = msg.msgid.mid(endindex + 2, msg.msgid.length() - (endindex + 2) - 1);
        TQString endtag_attr = endtag.mid(endtag.find(' '), endtag.length());
        endtag.replace(infos_reg, "");
        if (endtag == starttag) {
            if (!closureTag(msg.msgid, starttag))
                break;

            // removing start/end tags
            msg.msgid = msg.msgid.left(endindex);
            strindex = 0;
            while (msg.msgid.at(strindex) != '>')
                strindex++;
            TQString attr = msg.msgid.left(strindex);
            msg.msgid = msg.msgid.mid(strindex + 1);
            offset += strindex + 1;
            for (int index = 0; msg.msgid.at(index) == ' '; index++, offset++) ;
            stripWhiteSpace( msg.msgid );
            msg.tag = starttag;

            if (infos_reg.search(attr) >= 0) {
                msg.lines.first().start_line = infos_reg.cap(1).toInt();
                msg.lines.first().start_col = infos_reg.cap(2).toInt();
#ifdef POXML_DEBUG
                tqDebug("col %s %s %d", attr.latin1(), msg.msgid.latin1(), msg.lines.first().start_col);
#endif
                offset = 0;

                if (infos_reg.search(endtag_attr) >= 0) {
                    msg.lines.first().end_line = infos_reg.cap(1).toInt();
                    msg.lines.first().end_col = infos_reg.cap(2).toInt() + 1;
                }
            }
            if (do_not_split_reg.search(attr) >= 0) {
                msg.do_not_split = true;
                break;
            }

            changed = true;
        } else
            break;
    }

#ifdef POXML_DEBUG
    tqDebug("formatMessage result %s %d %d", msg.msgid.latin1(), changed && recurse, msg.lines.first().start_col);
#endif

    msg.lines.first().offset += offset;
    if (msg.do_not_split)
        recurse = false;

    if (changed && recurse)
        formatMessage(msg);

    return !recurse; // indicates an abort
}

MsgList StructureParser::splitMessage(const MsgBlock &mb)
{
    MsgList result;

    MsgBlock msg1 = mb;
    MsgBlock msg2 = mb;

    TQString message = mb.msgid;

#ifdef POXML_DEBUG
    tqDebug("splitMessage %s", message.latin1());
#endif

    if (message.at(0) == '<') {
        int endindex = 1;
        while (!message.at(endindex).isSpace() && message.at(endindex) != '>')
            endindex++;
        TQString tag = message.mid(1, endindex - 1);

        if (closureTag(message, tag))
            goto error;

        if (isCuttingTag(tag))
        {
            // if the message starts with a cutting tag, this tag has to
            // end in between. We split both messages and format them
            int strindex = endindex;
            strindex++;

            int inside = 1;
            while (true) {
#ifdef POXML_DEBUG
                tqDebug("inside %s %d", message.mid(strindex, 35).latin1(), inside);
#endif

                // the exception for poxml_* attributes is made in the closing tag
                int closing_index = message.find(TQRegExp(TQString::fromLatin1("</%1[\\s>]").arg(tag)),
                                                 strindex);
                int starting_index = message.find(TQRegExp(TQString::fromLatin1("<%1[\\s>]").arg(tag)),
                                                  strindex);

#ifdef POXML_DEBUG
                tqDebug("index1 %d %d %d", closing_index, starting_index, strindex);
#endif

                // when a new start was found, we set the start_index after the next match
                // (and set strindex to it later - increasing inside)
                if (starting_index != -1) {
                    starting_index += tag.length() + 1;
                    while (message.at(starting_index) != '>')
                        starting_index++;
                    starting_index++;
                }

#ifdef POXML_DEBUG
                tqDebug("index %d %d %d", closing_index, starting_index, strindex);
#endif

                assert(closing_index != -1);
                closing_index += 3 + tag.length();
                while (message.at(closing_index - 1) != '>')
                    closing_index++;

                if (starting_index == -1) {
                    strindex = closing_index;
#ifdef POXML_DEBUG
                    tqDebug("set strindex %d", strindex);
#endif
                    inside--;
                    if (!inside)
                        break;
                    continue;
                }
                if (closing_index < starting_index)
                {
                    strindex = closing_index;
                    inside--;
                } else {
                    strindex = starting_index;
                    inside++;
                }

                if (!inside)
                    break;
            }

#ifdef POXML_DEBUG
            tqDebug("split into %s -AAAAAANNNNNNDDDDDD- %s", message.left(strindex).latin1(), message.mid(strindex).latin1());
#endif
            msg1.msgid = message.left(strindex);
            bool leave = formatMessage(msg1);

            msg2.msgid = message.mid(strindex);
            msg2.lines.first().offset += strindex;
            leave = leave & formatMessage(msg2);

            if (msg1.lines.first().end_line > msg2.lines.first().start_line ||
                (msg1.lines.first().end_line == msg2.lines.first().start_line &&
                 msg1.lines.first().end_col > msg2.lines.first().start_col))
            {
                msg2.lines.first().start_line = msg1.lines.first().end_line;
                msg2.lines.first().start_col = msg1.lines.first().end_col;
            }

#ifdef POXML_DEBUG
            tqDebug("splited %d-%d(%s) and %d-%d(%s)", msg1.lines.first().end_line,msg1.lines.first().end_col,
                   msg1.msgid.latin1(),
                   msg2.lines.first().start_line,msg2.lines.first().start_col, msg2.msgid.latin1());
#endif

            if (leave) {
                result.append(msg1);
                result.append(msg2);
                return result;
            }
            result = splitMessage(msg1);
            result += splitMessage(msg2);
            return result;
        }

    }

    if (message.at(message.length() - 1 ) == '>')
    {
        int endindex = message.length() - 1;
        while (endindex >= 0 && (message.at(endindex) != '<' || message.at(endindex + 1) != '/'))
            endindex--;
        TQString tag = message.mid(endindex + 2, message.length() - endindex - 3);
        if (tag.find(' ') > 0 ) {
            tag = tag.left(tag.find(' '));
        }
#ifdef POXML_DEBUG
        tqDebug("behind tag %s", tag.latin1());
#endif

        if (isCuttingTag(tag))
        {
            // if the message ends with a cutting tag, this tag has to
            // start in between. We split both messages and format them
            int strindex = endindex;

            int inside = 1;
            while (true) {
#ifdef POXML_DEBUG
                tqDebug("inside %s %d", message.mid(strindex, 35).latin1(), inside);
#endif

                int closing_index = message.findRev(TQRegExp(TQString::fromLatin1("</%1[\\s>]").arg(tag)),
                                                    strindex - 1);
                int starting_index = message.findRev(TQRegExp(TQString::fromLatin1("<%1[\\s>]").arg(tag)),
                                                     strindex - 1);

#ifdef POXML_DEBUG
                tqDebug("index1 %d %d %d", closing_index, starting_index, strindex);
#endif

                if (starting_index == -1) {
                    assert(inside == 1);
                    break;
                }

                if (closing_index > starting_index)
                {
                    strindex = closing_index;
                    inside++;
                } else {
                    strindex = starting_index;
                    inside--;
                }

                if (!inside)
                    break;
            }


#ifdef POXML_DEBUG
            tqDebug("split2 into \"%s\" -AAAAAANNNNNNNNNDDDDDDDDDDD- \"%s\"", message.left(strindex).latin1(), message.mid(strindex).latin1());
#endif

            msg1.msgid = message.left(strindex);
            formatMessage(msg1);

            msg2.msgid = message.mid(strindex);
            msg2.lines.first().offset += strindex;
            formatMessage(msg2);

            if (msg1.lines.first().end_line > msg2.lines.first().start_line ||
                (msg1.lines.first().end_line == msg2.lines.first().start_line &&
                 msg1.lines.first().end_col > msg2.lines.first().start_col))
            {
                msg1.lines.first().end_line = msg2.lines.first().start_line;
                msg1.lines.first().end_col = msg2.lines.first().start_col - 1;
            }

#ifdef POXML_DEBUG
            tqDebug("splited %d-%d(%s) and %d-%d(%s)", msg1.lines.first().end_line,msg1.lines.first().end_col,
                   msg1.msgid.latin1(),
                   msg2.lines.first().start_line,msg2.lines.first().start_col, msg2.msgid.latin1());
#endif

            result = splitMessage(msg1);
            result += splitMessage(msg2);

            return result;
        }
    }
error:
    result.append(mb);
    return result;
}

bool StructureParser::endElement( const TQString& , const TQString&, const TQString& qName)
{
    TQString tname = qName.lower();

    // tqDebug("endElement %s - %s %d", tname.latin1(), message.latin1(), inside);

    if (inside) {
        if (!isSingleTag(qName)) {
            message += TQString("</%1").arg(tname);
            message += TQString(" poxml_line=\"%1\"").arg(locator->lineNumber());
            message += TQString(" poxml_col=\"%1\"").arg(locator->columnNumber());
            message += ">";
        }
    }

    if (isCuttingTag(tname)) {
        inside--;
        if (!inside) {
            MsgBlock m;
            descape(message);
            m.msgid = message;

            BlockInfo bi;
            bi.start_line = startline;
            bi.start_col = startcol;
            bi.end_line = locator->lineNumber();
            bi.end_col = locator->columnNumber() + 1;
            bi.offset = m.lines.first().offset;
            m.lines.append(bi);
            formatMessage(m);

            MsgList messages = splitMessage(m);
            for (MsgList::Iterator it = messages.begin();
                 it != messages.end(); ++it)
            {
#ifdef POXML_DEBUG
                tqDebug("parser '%s' %d '%s' %d:%d", (*it).msgid.latin1(), (*it).lines.first().offset, message.mid((*it).lines.first().offset, 15).latin1(), (*it).lines.first().start_line, (*it).lines.first().start_col);
#endif
                // if the remaining text still starts with a tag, the poxml_ info
                // is most probably more correct
                if ((*it).msgid.at(0) == '<' && isClosure((*it).msgid)) {
                    if (infos_reg.search((*it).msgid) >= 0) {
                        (*it).lines.first().start_line = infos_reg.cap(1).toInt();
                        (*it).lines.first().start_col =  infos_reg.cap(2).toInt();;
                        (*it).lines.first().offset = 0;
                    }
                }
                (*it).msgid.replace(infos_reg, TQString());

                if (!(*it).msgid.isEmpty())
                    list.append(*it);
            }
        }
    }

    return true;
}

bool StructureParser::comment ( const TQString &c )
{
    if (c.left(7) != " TRANS:")
        return true;

    assert(false);
    return true;
}

TQString StructureParser::escapeLiterals( const TQString &_contents) {
    TQString contents = _contents;

    contents.replace(TQRegExp("\n"), "&POXML_LINEFEED;");
    contents.replace(TQRegExp("<"), "&POXML_LT;");
    contents.replace(TQRegExp(">"), "&POXML_GT;");
    contents.replace(TQRegExp("\t"), "        ");
    contents.replace(TQRegExp(" "), "&POXML_SPACE;");

    return contents;
}

TQString StructureParser::descapeLiterals( const TQString &_contents) {
    TQString contents = _contents;

    contents.replace(TQRegExp("&POXML_LINEFEED;"), "\n");
    contents.replace(TQRegExp("&POXML_LT;"), "<");
    contents.replace(TQRegExp("&POXML_GT;"), ">");
    contents.replace(TQRegExp("&POXML_SPACE;"), " ");
    contents.replace(TQRegExp("!POXML_AMP!"), "&");
    return contents;
}

void StructureParser::stripWhiteSpace( TQString &contents)
{
    contents = contents.stripWhiteSpace();
    bool changed;
    do {
        changed = false;
        if (contents.startsWith("&POXML_LINEFEED;")) {
            contents = contents.mid(strlen("&POXML_LINEFEED;"), contents.length());
            changed = true;
        }
        if (contents.startsWith("&POXML_SPACE;")) {
            contents = contents.mid(strlen("&POXML_SPACE;"), contents.length());
            changed = true;
        }
        if (contents.endsWith("&POXML_LINEFEED;")) {
            contents = contents.left(contents.length() - strlen("&POXML_LINEFEED;"));
            changed = true;
        }
        if (contents.endsWith("&POXML_SPACE;")) {
            contents = contents.left( contents.length() - strlen("&POXML_SPACE;"));
            changed = true;
        }
    } while (changed);
}

void StructureParser::cleanupTags( TQString &contents )
{
    contents.replace(TQRegExp("&"), "!POXML_AMP!");

    for (int index = 0; literaltags[index]; index++) {
        TQRegExp start(TQString("<%1[\\s>]").arg(literaltags[index]));
        TQRegExp end(TQString("</%1[\\s>]").arg(literaltags[index]));
        int strindex = 0;
        while (true) {
            strindex = contents.find(start, strindex);
            if (strindex < 0)
                break;
            while (contents.at(strindex) != '>')
                strindex++;
            strindex++; // one more
            int endindex = contents.find(end, strindex);
            TQString part = contents.mid(strindex, endindex - strindex);
            TQString newpart = escapeLiterals(part);
            contents.replace(strindex, part.length(), newpart);
            // this assumes that literal tags to not overlap
            strindex = strindex + newpart.length();
        }
    }

    TQRegExp unclosed("</(\\w*)\\s\\s*>");
    int index = -1;
    while (true) {
        index = unclosed.search(contents, index + 1);
        if (index < 0)
            break;
        TQString tag = unclosed.cap(1);
        contents.replace(index, unclosed.matchedLength(), TQString("</%1>").arg(tag));
    }

    TQRegExp start("<((\\s*[^<>\\s])*)\\s\\s*(/*)>");
    start.setMinimal(true);

    index = -1;
    while (true) {
        index = start.search(contents, index + 1);
        if (index < 0)
            break;
        TQString tag = start.cap(1);
	TQString cut = start.capturedTexts().last();
        // tqDebug("UNCLO %s %d -%s- -%s-", start.cap(0).latin1(), index, tag.latin1(), cut.latin1());
        contents.replace(index, start.matchedLength(), TQString("<%1%2>").arg(tag).arg(cut));
    }
    TQRegExp singletag("<(\\w*)\\s([^><]*)/>");

    index = -1;
    while (true) {
        index = singletag.search(contents, index + 1);
        if (index < 0)
            break;
        TQString tag = singletag.cap(1);
        if (!StructureParser::isSingleTag(tag)) {
            contents.replace(index, singletag.matchedLength(), TQString("<%1 %2></%3>").arg(tag).arg(singletag.cap(2)).arg(tag));
        }
    }

    TQRegExp trans_comment("<!-- TRANS:([^<>]*)-->");
    index = -1;
    while (true) {
        index = trans_comment.search(contents, index + 1);
        if (index < 0)
            break;
        TQString msgid = trans_comment.cap(1);
        contents.replace(index, trans_comment.matchedLength(), TQString("<trans_comment>%1</trans_comment>").arg(msgid));
    }

#ifdef POXML_DEBUG
    tqDebug("final  %s", contents.latin1());
#endif

}

static bool removeEmptyTag( TQString &contents, const TQString & tag)
{
//    tqDebug("cont %s %s", contents.latin1(), tag.latin1());

    TQRegExp empty(TQString("<%1[^>]*>[\\s\n][\\s\n]*</%2\\s*>").arg(tag).arg(tag));
    int strindex = 0;
    while (true) {
        strindex = contents.find(empty, strindex);
        if (strindex < 0)
            break;
        tqDebug("found empty tag %s", tag.latin1());
        contents.replace(strindex, empty.matchedLength(), " ");
        strindex++;
        return true;
    }
    return false;
}

void StructureParser::removeEmptyTags( TQString &contents )
{
    bool removed;
    do {
        removed = false;

        for (int index = 0; cuttingtags[index]; index++) {
            if (removeEmptyTag(contents, cuttingtags[index])) {
                removed = true;
                break;
            }
        }
        // as glossterm has two different semantics, it's likely
        // to break something when it's cuttingtag
        if (removeEmptyTag(contents, "glossterm"))
            removed = true;

    } while (removed);
}

bool StructureParser::characters(const TQString &ch)
{
    if (inside && !ch.isEmpty())
        message += ch;
    return true;
}

TQString escape(TQString message)
{
    message.replace(TQRegExp("\\\\"), "\\\\");
    message.replace(TQRegExp("\""), "\\\"");
    return message;
}

void outputMsg(const char *prefix, const TQString &message)
{
    TQStringList list = TQStringList::split('\n', message, true);
    TQString line;

    if (list.count() == 1) {
        line = list.first();
        if (line.isEmpty())
            cout << prefix << " \"\"\n";
        else
            cout << prefix << " \"" << escape(line).utf8().data() << "\"\n";
    } else {
        cout << prefix << " \"\"\n";
        for (TQStringList::ConstIterator it = list.begin(); it != list.end(); it++) {
            line = *it;
            if (!line.isEmpty()) {
                cout << "      \"" << escape(line).utf8().data();
                if (it == list.fromLast())
                    cout << "\"\n";
                else
                    cout << "\\n\"\n";
            } else {
                cout << "      \"";
                if (it != list.fromLast())
                    cout << "\\n";
                cout << "\"\n";
            }
        }
    }
}

TQString escapePO(TQString msgid)
{
    int index = 0;
    while (true) {
        index = msgid.find("\\n", index);
        if (index == -1)
            break;
        if (index >= 1 && msgid.at(index - 1) == '\\' && msgid.at(index - 2) != '\\') {
            msgid.replace(index - 1, 3, "&POXML_LITERALLINEFEED;");
            index += 3;
        } else
            msgid.replace(index, 2, "\n");
    }
    index = 0;
    while (true) {
        index = msgid.find("\\\"", index);
        if (index == -1)
            break;
        if (index > 1 && msgid.at(index - 1) == '\\' && msgid.at(index - 2) != '\\')
            msgid.replace(index - 1, 3, "&POXML_LITERALQUOTE;");
        else
            msgid.replace(index, 2, "\"");
    }
    index = 0;
    while (true) {
        index = msgid.find("\\t", index);
        if (index == -1)
            break;
        if (msgid.at(index - 1) == '\\')
            msgid.replace(index - 1, 3, "\\t");
        else
            msgid.replace(index, 2, "\t");
    }
    index = 0;
    while (true) {
        index = msgid.find("\\\\", index);
        if (index == -1)
            break;
        msgid.replace(index, 2, "\\");
        index += 1;
    }

    msgid.replace(TQRegExp("&POXML_LITERALLINEFEED;"), "\\n");
    msgid.replace(TQRegExp("&POXML_LITERALQUOTE;"), "\\");
    return msgid;
}


MsgList parseXML(const char *filename)
{
    StructureParser handler;
    TQFile xmlFile( filename );
    xmlFile.open(IO_ReadOnly);

    TQCString ccontents;
    ccontents.fill(0, xmlFile.size() + 1);
    memcpy(ccontents.data(), xmlFile.readAll().data(), xmlFile.size());
    xmlFile.close();

    TQString contents = TQString::fromUtf8( ccontents );
    StructureParser::cleanupTags(contents);

    while (true) {
        int index = contents.find("<!ENTITY");
        if (index < 0)
            break;
        int inside = 0;
        int endindex = index + 1;
        TQString replacement = "";
        while (contents.at(endindex) != '>' || inside)
        {
            switch (contents.at(endindex).latin1()) {
                case '<':
                    inside++; break;
                case '>':
                    inside--; break;
                case '\n':
                    replacement += '\n';
                    break;
                default:
                    break;
            }
            endindex++;
        }
        endindex++;
        contents.replace(index, endindex - index, replacement);
    }

    TQTextStream ts(contents.utf8(), IO_ReadOnly);
    TQXmlInputSource source( ts );
    TQXmlSimpleReader reader;
    reader.setFeature( "http://trolltech.com/xml/features/report-start-end-entity", true);
    reader.setContentHandler( &handler );
    reader.setLexicalHandler( &handler );
    reader.setDTDHandler( &handler );
    // reader.setErrorHandler( &handler );
    reader.parse( source );
    MsgList english = handler.getList();

    bool changed = false;

    do {
        changed = false;
        TQMap<TQString, TQString> msgids;

        for (MsgList::Iterator it = english.begin();
             it != english.end(); it++)
        {
            TQMap<TQString,TQString>::Iterator found = msgids.find((*it).msgid);
            if ((*it).msgid.length() < 4) {
                (*it).msgid = TQString("<%1>").arg((*it).tag) + (*it).msgid +
                              TQString("</%1>").arg((*it).tag);
                changed = true;
                break;
            }
            if (found != msgids.end()) {
                if (found.data() != (*it).tag) {
#ifdef POXML_DEBUG
                    tqDebug("same msgid for '%s' and '%s'", found.data().latin1(), (*it).tag.latin1());
#endif
                    changed = true;
                    TQString msgid = (*it).msgid;
                    for (MsgList::Iterator it2 = english.begin();
                         it2 != english.end(); it2++)
                    {
                        if ((*it2).msgid == msgid)
                            (*it2).msgid = TQString("<%1>").arg((*it2).tag) + msgid + TQString("</%1>").arg((*it2).tag);
                    }
                    break;
                }
            } else {
                msgids.insert((*it).msgid, (*it).tag);
            }
        }
    } while (changed);

    return english;
}