Download and regex parse an url source code

[Gojir4]

@Mr-Gisa If you are still interested, here is the full code of my application parsing Doxygen html files using QXmlQuery. The code is not well commented but feel free to ask me if you need some details. I will try to remember, because I did that one year ago... You will see that sometimes I have mixed both QXmlQuery and QRegularExpression to achieve my goal.

That's not really the same that you are trying to achieve, but maybe it could help you to understand how QXmlQuery works.
I cannot provide the Doxygen HTML files because of confidentiality, but you can easily generate them using an arbitrary projet or C++ header file.

#include <QBuffer>
#include <QDebug>
#include <QCommandLineParser>
#include <QCoreApplication>
#include <QFileInfo>
#include <QTextStream>
#include <QXmlQuery>
#include <QJsonDocument>
#include <QJsonObject>
#include <QRegularExpression>
#include "iostream"

#ifdef WIN32
#include "windows.h"
HANDLE hOut = NULL;
HANDLE hErr = NULL;
#endif

QTextStream o(stdout);
QTextStream e(stderr);

//const QString INDEX_QUERY = "for $a in doc($index)/html/body/div/div/table/tr/td/a\n"
//                            "where $a/@class = 'el'\n"
//                            "order by $a/text()\n"
//                            "return concat($a/text(), '-', $a/@href, ';')";

const QString INDEX_QUERY = "for $item in doc($root)/html/body/div/div/table/tr/td/a\n"
                            "where $item[@target]\n"
                            "return concat($item/text(), '-', $item/@href, '\n')";


const QString CLASS_QUERY = "for $item in doc($root)/html/body/div/table/tr\n"
                            "where contains($item/@class, 'memitem:')\n"
                            "return $item";

const QString MEMBER_QUERY = "for $item in doc($root)/html/tr/td/a\n"
                             "return concat($item/text(), \'-\', $item/@href,'\n')";

const QString GLOBAL_INDEX_QUERY = "for $item in doc($root)/html/body/div/table/tr/td/a\n"
                                   "where $item/@class = 'el' and $item[@href] and $item[not (@title)]\n"
                                   "return concat($item/text(), '-', $item/@href, '\n')";

const QString ENUM_QUERY = "for $item in doc($root)/html/body/div/table/tr\n"
                           "where $item/td/text()[contains(., 'enum' )]\n"
                           "return concat($item/td[@class = 'memItemRight'], ';')";

const QString ENUM_QUERY_URL_FILTER = "for $item in doc($root)/html/body/div/table/tr\n"
                                      "where $item/td/text()[contains(., 'enum' )]\n"
                                      "return $item/td[@class = 'memItemRight']";

const QString ENUM_URL_QUERY = "for $item in doc($root)/html/td\n"
                               "return $item/a[1]/@href/string()";



const QString CLASS_HEADER_METHODS = "pub-methods";
const QString CLASS_HEADER_ATTRIBS = "pub-attribs";
const QString CLASS_HEADER_CONSTANTS = "pub-types";
const QString CLASS_HEADER_SIGNALS = "signals";

enum ConsoleColors{
    BLACK = 0,
    DARK_BLUE = 1,
    DARK_GREEN = 2,
    DARK_CYAN = 3,
    DARK_RED = 4,
    DARK_MAGENTA = 5,
    DARK_YELLOW = 6,
    LIGHT_GREY = 7,
    DARK_GREY = 8,
    BLUE = 9,
    GREEN = 10,
    CYAN = 11,
    RED = 12,
    MAGENTA = 13,
    YELLOW = 14,
    WHITE = 15

};


void msg(const QString &msg, int color = WHITE){
#ifdef WIN32
    SetConsoleTextAttribute(hOut, color);
#endif
    o << msg << endl;
    o.flush();
#ifdef WIN32
    SetConsoleTextAttribute(hOut, WHITE);
#endif

}
int err(const QString &msg, int color = RED){
#ifdef WIN32
    SetConsoleTextAttribute(hErr, color);
#endif
    e << msg << endl;
    e.flush();
#ifdef WIN32
    SetConsoleTextAttribute(hOut, WHITE);
#endif
    return 1;
}

bool evaluateFile(const QString &filepath, const QString &query, QString &result, const QString &varName = "root"){
    QString str;
    QBuffer buf;
    QFile f(filepath);
    if(!f.open(QFile::ReadOnly)){
        err(QString("Cannot open %1").arg(f.fileName()));
        return false;
    }

    str = QString(f.readAll());
    f.close();
    
    str.replace("<html xmlns=\"http://www.w3.org/1999/xhtml\">", "<html>");
    buf.setData(str.toUtf8());
    buf.open(QIODevice::ReadOnly);

    QXmlQuery indexQuery;
    indexQuery.bindVariable(varName, &buf);
    indexQuery.setQuery(query);

    if(!indexQuery.evaluateTo(&result)){
        return false;
    }
    return true;
}

bool evaluateString(const QString &s, const QString &query, QString &result, const QString &varName = "root", bool wrapInHtml = true){
    QString str = s;
    QBuffer buf;
    if(wrapInHtml){
        str = "<html>\n" + str + "\n</html>";
    } else {
        str.replace("<html xmlns=\"http://www.w3.org/1999/xhtml\">", "<html>");
    }
    buf.setData(str.toUtf8());
    buf.open(QIODevice::ReadOnly);

    QXmlQuery indexQuery;
    indexQuery.bindVariable(varName, &buf);
    indexQuery.setQuery(query);

    if(!indexQuery.evaluateTo(&result)){
        return false;
    }
    result = result.trimmed();
    return true;
}

struct ClassFile{
    QString file;
    QString className;

    bool operator ==(const ClassFile &other){
        return this->file == other.file && this->className == other.className;
    }
};

int main(int argc, char *argv[])
{
    QCoreApplication a(argc, argv);

#ifdef WIN32
    hOut = GetStdHandle(STD_OUTPUT_HANDLE);
    hErr = GetStdHandle(STD_ERROR_HANDLE);
#endif

    QCommandLineParser parser;
    parser.setApplicationDescription("EM4315 Test Bench - Html Help Parser Helper");
    parser.addHelpOption();
    QCommandLineOption outputOption(QStringList() << "o" << "output", "Output file name, default is 'doc_links.json'", "file");
    parser.addOption(outputOption);
    parser.addPositionalArgument("annotated", QCoreApplication::translate("main", "Path to annotated.html file from doxygen"));
    parser.addPositionalArgument("global", QCoreApplication::translate("main", "Path to group__gglobal.html file from doxygen"));

    // Process the actual command line arguments given by the user
    parser.process(a.arguments());

    const QStringList posArgs = parser.positionalArguments();
    if(posArgs.isEmpty()){
        return err("Error: annotated.html path missing");
    } else if(!QFileInfo::exists(posArgs.first())){
        return err("Error: annotated.html file not found");
    }

    QString path = posArgs.first();

    //Get object and types
    //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

    //Index file
    QString result;
    if(!evaluateFile(path, INDEX_QUERY, result)){
        return err(QString("Index evaluation failed"));
    }
    msg(QString("Result: %1").arg(result));

    //Make list of each files for each class (object and types)
    QList<ClassFile> classes;
    for(const QString &s : result.split("\n", QString::SkipEmptyParts)){
        int pos = s.indexOf("-");
        if((pos < 0) || (pos >= (s.size() -1)))
            continue;
        ClassFile cf;
        cf.className = s.mid(0, pos).trimmed();
        cf.file = QFileInfo(path).absolutePath() + "/" + s.mid(pos + 1).trimmed();
        if(classes.contains(cf))
            continue;
        classes << cf;
        msg(QString("Class found: %1 - %2").arg(cf.className, cf.file), GREEN);
    }

    //Process each class file and extract members
    QVariantMap helpMap;
    for(const ClassFile &cf : classes){

        helpMap.insert(cf.className, QFileInfo(cf.file).fileName());
       
        QString classRes;
        if(!evaluateFile(cf.file, CLASS_QUERY, classRes)){
            err(QString("Parsing error class: %1").arg(cf.className), YELLOW);
            continue;
        }

        QString enumRes;
        if(!evaluateFile(cf.file, ENUM_QUERY, enumRes)){
            err(QString("Enum parsing error class: %1").arg(cf.className), YELLOW);
            enumRes = "";
        } else {

            QString enumUrlFiltered;
            if(!evaluateFile(cf.file, ENUM_QUERY_URL_FILTER, enumUrlFiltered)){
                err(QString("Enum parsing error class: %1").arg(cf.className), YELLOW);
                enumUrlFiltered = "";
            }

            QString enumUrlRes;
            //msg(enumUrlFiltered);
            if(!evaluateString(enumUrlFiltered, ENUM_URL_QUERY, enumUrlRes)){
                err(QString("Enum url parsing error class: %1").arg(cf.className), YELLOW);
                enumUrlRes = "";
            }


            //Parse enum result
            QStringList enums = enumRes.split(";", QString::SkipEmptyParts);
            int i = 0 ;
            QStringList enumUrls = enumUrlRes.split(" ", QString::SkipEmptyParts);
            //QString currentUrl;
            for(const QString &e : enums){
                if(i >= enumUrls.size())
                    break;
                
                QRegularExpression htmlReplace("<br/>|<[/]{0,1}b>");
                QRegularExpression getEnumsRegEx("\\{([\\w\\W]+?)\\}");

                QRegularExpressionMatch getEnumMatch = getEnumsRegEx.match(e);
                if(getEnumMatch.hasMatch()){
                    QString content = getEnumMatch.captured();
                    content = content.replace(htmlReplace, "");
                    QStringList constants = content.split(",", QString::SkipEmptyParts);
                    for(QString konst: constants){
                        konst = konst.replace(QRegularExpression("^\\{[\\s\\\\n]*"), "").trimmed();
                        int pos = konst.indexOf("=");
                        const QString fullName =  cf.className + "." + (pos > 0 ? konst.left(pos).trimmed() : konst);

                        helpMap[fullName] = enumUrls.at(i).trimmed();
                        //currentUrl = enumUrls.at(i).trimmed();
                    }

                }
                i++;
            }
        }



        //Parse class members        
        QString memberRes;
        if(!evaluateString(classRes, MEMBER_QUERY, memberRes)){
            err(QString("Parsing error member: %1").arg(classRes), YELLOW);
            continue;
        }

        QStringList members = memberRes.split("\n", QString::SkipEmptyParts);
        if(members.isEmpty())
            continue;

        for(const QString &m : members){
            int pos = m.indexOf("-");
            if((pos < 1 )|| (pos >= (m.size()-1)))
                continue;

            if((cf.className == m.mid(0, pos).trimmed()) && cf.className.at(0).isUpper())
                helpMap.insert(cf.className, m.mid(pos + 1).trimmed());
            else
                helpMap.insert(cf.className + "." + m.mid(0, pos).trimmed(), m.mid(pos + 1).trimmed());
        }

    }



    //Get globals items
    //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    if(posArgs.size() >= 2){

        path = posArgs.at(1);
        //Index file
        QString globalRes;
        if(!evaluateFile(path, GLOBAL_INDEX_QUERY, globalRes)){
            return err(QString("Global index evaluation failed"));
        }

        //Parse query result
        //Format "name-url"
        QStringList globals = globalRes.split("\n", QString::SkipEmptyParts);
        if(!globals.isEmpty()){
            for(const QString &m : globals){
                int pos = m.indexOf("-");
                if((pos < 1 )|| (pos >= (m.size()-1)))
                    continue;
                helpMap.insert(m.mid(0, pos).trimmed(), m.mid(pos + 1).trimmed());
            }
        }

        msg(QString("Global result: %1").arg(globalRes));
    }

    //qDebug() << "Help map: " << helpMap;
    QString output = parser.value(outputOption);
    if(output.isEmpty())
        output = "doc_links.json";

    QJsonObject json = QJsonObject::fromVariantMap(helpMap);
    QFile fOut(output);
    QJsonDocument jdoc(json);
    if(!fOut.open(QFile::WriteOnly))
        return err(QString("Cannot open output file %1").arg(fOut.fileName()));

    fOut.write(jdoc.toJson());
    fOut.close();

    return 0;
}

[JonB]

@Mr-Gisa
Just to be clear: web page source code is written in HTML + whatever (e.g. JavaScript, etc.). Qt does not have an HTML parser. You cannot reliably parse any HTML using regular expressions. You cannot reliably recognise what bits are really URLs in the web page with regular expressions.

What you can do is use regular expressions to "guess" what bits might be genuine URLs. You might find some spurious, "extra" ones, and you might miss some "genuine" ones. It might do better or worse on some HTML sources (e.g. Doxygen) than others.

So long as you are happy with this "approximation" that's OK. But you will never guarantee to get them all correctly.

I don't really know what @Gojir4 's code is doing. HTML is not XML, so if it's relying on HTML input being parseable as well-formed XML it will fail. XHTML is XML, and so will parse, but not many sites produce XHTML. Maybe Doxygen does, others do not.

[Gojir4]

@JonB You are right, Doxygen produce XHTML, I didn't notice that sorry. I assumed naively that HTML and XML can be parsed the same way.

[Gojir4]

From the doc of QXmlQuery
"QXmlQuery is typically used to query XML data, but it can also query non-XML data that has been modeled to look like XML."

and then the code example parse an HTML file:

  QXmlQuery query;
  query.setQuery("doc('index.html')/html/body/p[1]");

I'm a little bit confused about this right now.

[JonB]

@Gojir4
Yes, note the

non-XML data that has been modeled to look like XML

and the page's further:

The example uses QXmlQuery to match the first paragraph of an XML document and then output the result to a device as XML.

So (bearing in mind I know nothing about this!), what exactly does the doc('index.html') deliver? In http://doc.qt.io/qt-5/xmlprocessing.html I can see it mentions:

When Qt XML Patterns loads an XML resource, e.g., using the fn:doc() function

but I can't click on that. Where is fn:doc() documented?

EDIT
OK, fn:doc() is just an XQuery function for accessing the document object.

So that assumes that you already have a parsed document. All the examples I can see anywhere other than that example access a .xml file, not a .html one, which is as I would expect.

So I assume this will only work for you if the particular HTML file you pass happens to parse as XML, i.e. it's either XHTML in the first place, or the HTML it contains does not have anything HTML-but-not-XML in it (which may be the case for some HTML documents but not others).

Try putting, say, precisely <br> (and no </br>) somewhere in your HTML and see if it still parses? <br> is a common example of legal HTML, but is not legal in XHTML or XML...?

[Gojir4]

@JonB I think the fn:doc is part of the XQuery/XPath specification

[JonB]

@Gojir4
See my EDIT above.

[Gojir4]

@JonB You are right, tags without corresponding closing tag, as <br>, are not handled by XQuery, you got the error "Opening and ending tag mismatch".
But, depending of the input format, I think this could be easily handled by making some replacement in the html code before to evaluate it with XQuery. That's what I did when I have used XQuery.

[JonB]

@Gojir4

But, depending of the input format, I think this could be easily handled by making some replacement in the html code before to evaluate it with XQuery. That's what I did when I have used XQuery.

And I do not think that is "easy", precisely because as I said you don't have a parser for HTML, and regular expressions are a hack which at best work "approximately" and at worst get it all wrong! That's all I was trying to warn the OP about --- it won't be robust for his random HTML pages. If it works for you/him, good luck!

[Gojir4]

@JonB said in Download and regex parse an url source code:

and regular expressions are a hack which at best work "approximately"

I dont agree about that, in my opinion, regex are extremely powerful and work as expected if used correctly. I agree that's not designed to make code parsing, but combined with other "search algorithm", like XQuery, or simple string manipulation, it can achieve almost everything. I'm using regex from years now, and I never see any "approximative" result, except, of course, when the regular expression itself was badly defined.
But, that's only my opinion.

[JonB]

@Gojir4
I never said regular expressions themselves are "approximative"! Of course they work. But if you do not know/cannot correctly parse the input (HTML in this case), then what they recognise/do can, and often is, simply faulty. Your regular expression for recognising a URL might, for example, pick one up from inside a commented out fragment without knowing it has done so. That may or may not matter to you/the OP, I don't know.

There are plenty of posts on, say, stackoverflow explaining why HTML cannot be correctly parsed/recognised via regular expressions.

[Mr Gisa]

I solved the problem by using the myhtml library, it's fast and did the trick really nicely.

[JonB]

@Mr-Gisa
Then you should make that myhtml in your post a link to wherever it is, to help others. Thanks.

[Mr Gisa]

@JonB I was going to do that but due the heavy amount of things I forgot, thanks for pointing it out.

    QString html = "<html><head></head><body><div><span>HTML</span><a href=\"http://www.google.com\">a</a><a href=\"ohyeah.com\">b</a></div></body></html>";
    QByteArray chtml = html.toUtf8().constData();

    // basic init
    myhtml_t* myhtml = myhtml_create();
    myhtml_init(myhtml, MyHTML_OPTIONS_DEFAULT, 1, 0);

    // first tree init
    myhtml_tree_t* tree = myhtml_tree_create();
    myhtml_tree_init(tree, myhtml);

    // parse html
    myhtml_parse(tree, MyENCODING_UTF_8, chtml, strlen(chtml));

    // get the A collection
    myhtml_collection_t *collection = myhtml_get_nodes_by_tag_id(tree, NULL, MyHTML_TAG_A, NULL);

    for(size_t i = 0; i < collection->length; i++) {
        myhtml_tree_attr_t *gets_attr = myhtml_attribute_by_key(collection->list[i], "href", 4);

        if (gets_attr) {
            const char *attr_char = myhtml_attribute_value(gets_attr, NULL);
            qDebug() << attr_char;
        }
    }

    // release resources
    myhtml_collection_destroy(collection);
    myhtml_tree_destroy(tree);
    myhtml_destroy(myhtml);

[JonB]

@Mr-Gisa
No, you misunderstand! I want to know: what is this "myhtml" thing? Is it a package? Source code? I want the hyperlink to wherever it is, so that I can look at/download it like you have done!

[Mr Gisa]

MyHTML is a fast HTML Parser using Threads implemented as a pure C99 library with no outside dependencies. https://github.com/lexborisov/myhtml

[Gojir4]

@JonB Ok, I see, sorry for my misunderstood. I agree with that. My point was that if you know in advance which format you will have to parse (as for Doxygen), regex and xquery can becomes a solution. Anyway, the problem has been solved :).

[Mr Gisa]

That is okay, you helped a lot