[xsd-users] XSD uses a lot of memory

[Maxim Maslennikov]

Boris,

Enclosed are sources of the program based on libstudxml and xsd file.

The xsd program and xsd file are below:
#include <iostream>
#include <chrono>

#include "test.hxx"

using namespace std;

int main(int argc, const char * argv[]) {
    if (argc != 2)
    {
        cerr << "usage: " << argv[0] << " test.xml" << endl;
        return 1;
    }
    
    try
    {
        auto t = chrono::steady_clock::now();
        
        // Parsing and Binding
        unique_ptr<test::root> root (test::root_ (argv[1]));
        
        cerr << "Number of elements: " << root->record().size() << endl;
        cerr << "Binding Duration: " << chrono::duration_cast<chrono::milliseconds>(chrono::steady_clock::now()-t).count() << " ms\n”;
       
        // Serialize the modified object model to XML.
        //
        xml_schema::namespace_infomap map;
        map[""].name = "";
        map[""].schema = "test.xsd";
        
        t = chrono::steady_clock::now();
        
        test::root_ (cout, *root, map);
        
        cerr << "Serializing Duration: " << chrono::duration_cast<chrono::milliseconds>(chrono::steady_clock::now()-t).count() << " ms\n";
        
    }
    catch (const xml_schema::exception& e)
    {
        cerr << e << endl;
        return 1;
    }

}


To test the programs I generated a test file by gen program from libstudxml examples about 100Mb size.

Best Regards,
Maxim

> On Mar 10, 2015, at 3:05 PM, Boris Kolpackov <boris at codesynthesis.com> wrote:
> 
> Hi Maxim,
> 
> Maxim Maslennikov <maxim.maslennikov at gmail.com> writes:
> 
>> I meant my program uses an "object persistance" approach described
>> in your document - "XML Parsing and Serialization in C++ With
>> libstudxml” and also saves all data in memory.
> 
> You mean the data for the *entire* document is stored in memory?
> 
> 
>> So my question is why does xsd C++/tree use about ten time more
>> memory? I found it when generated a file of records about 100Mb
>> size.
> 
> If the answer to the above question is "yes", then I will need to
> see a test that reproduces this. C++/Tree definitely shouldn't
> use 10 times more memory.
> 
> Boris