迭代与升压mmaped gzip文件 [英] Iterating over mmaped gzip file with boost

查看:131
本文介绍了迭代与升压mmaped gzip文件的处理方法,对大家解决问题具有一定的参考价值,需要的朋友们下面随着小编来一起学习吧!

问题描述

我想学习升压和C ++中的一些模板编程,但我真的有这样一个困难时期实现一个简单的类使用 mapped_file_source 在gzip文件的迭代。我基本上是在TSV格式的边列表使得在gzip文件每一行的格式为:<源:INT><标签>< D​​ST:INT> 。我想是实施 gz_file 类,它暴露了一个开始和结束迭代器,我可以得到一个边缘(的std ::对< INT, INT方式> )我每次查询迭代器

问题是拷贝构造函数是坏,因为我不知道我在哪里放置在gzip文件。

下面是code我到目前为止有:

 类gz_graph {
上市:
    gz_graph(为const char *文件名)
    {
        m_file.open(文件名);
        如果(!m_file.is_open()){
            抛出std :: runtime_error(错误打开文件);
        }        M_DATA = m_file.data();
        m_data_size = m_file.size()/的sizeof(M_DATA [0]);        汽车RET = posix_madvise((无效*)M_DATA,m_data_size,POSIX_MADV_SEQUENTIAL);
    }    一流的迭代器;    迭代器开始()const的
    {
        返回的迭代器(这一点,假);
    }    迭代结束()const的
    {
        返回的迭代器(这一点,真正的);
    }    一流的迭代器:公众的std ::迭代器<的std :: forward_iterator_tag,边缘> {
        上市:
            迭代器(gz_graph常量*裁判,布尔消耗)
                :m_ref(REF)
                  m_cur_edge(-1,-1),
                  m_consumed(消耗)
            {
                如果(!消耗){
                    初始化();
                    提前();
                }
            }            迭代器(常量迭代器和放大器; X)
                :m_ref(x.m_ref)
                  m_cur_edge(x.m_cur_edge)
            {
                如果(!x.m_consumed){
                    初始化();
                    提前();
                }                性病::法院LT&;< 复制构造函数<<的std :: ENDL;
            }            VALUE_TYPE常量和放大器;运算符*()const的
            {
                返回m_cur_edge;
            }            VALUE_TYPE常量*操作符>()const的
            {
                返回&安培; m_cur_edge;
            }            迭代器和放大器;符++()
            {
                提前();
                返回*这一点;
            }            布尔运算符==(迭代器常量和放大器;其他)常量
            {
                断言(m_ref == other.m_ref);
                返回m_cur_edge == other.m_cur_edge;
            }            布尔运算符=(const的迭代器放大器和;除外)!常量
            {
                返回(*此==其他)!;
            }        私人的:
            void初始化()
            {
                提高::输入输出流:: array_source源(m_ref-> M_DATA,m_ref-> m_data_size);
                m_in.push(升压::输入输出流:: gzip_decom pressor());
                m_in.push(源​​);
            }            无效提前()
            {
                标准::字符串line_str;
                如果(!函数getline(m_in,line_str)){
                    m_consumed = TRUE;
                    m_cur_edge =边缘(-1,-1);
                    返回;
                }                的std ::矢量<标准::字符串>可疑交易报告;
                提高::分裂序列(STR,line_str,提振:: is_any_of(\\ t的));                如果(strs.size()!= 2)
                    抛出std :: runtime_error(必须每行2场);                INT SRC =的boost :: lexical_cast的< INT>(strs.at(0));
                INT DST =的boost :: lexical_cast的< INT>(strs.at(1));                m_cur_edge =边缘(SRC,DST);//性病::法院LT&;< 读行<< line_str<<的std :: ENDL;
            }            gz_graph常量* m_ref;
            边缘m_cur_edge;
            提高::输入输出流:: filtering_istream m_in;
            布尔m_consumed;
        };私人的:
    提高::输入输出流:: mapped_file_source m_file;
    字符常量* M_DATA;
    为size_t m_data_size;
};


解决方案

我只想用一个的std :: istream_iterator 在这里。我不知道究竟如何跨preT你的伪输入code,所以让我的幽默,你和做了复杂解析:

 结构边缘:性病::对< INT,INT> {};的std :: istream的&安培;运营商的GT;>(的std :: istream的&放大器是,边及放大器;边)
{
    使用空间boost ::精神::补气;
    返回的是>>匹配(SRC:> int_>'\\ T'>中的DST:> int_>> EOL,edge.first,edge.second);
}

我希望你会很乐意把它简单得多,但更简单更容易,对吧?

现在主程序看起来像

 为(
        的std :: istream_iterator<边及GT;它(FS>>的std :: noskipws),结束;
        !它=结束;
        它++)
{
    性病::法院LT&;<它 - >首先<< 到&所述;&下; IT->第二个<< \\ n;
}

其中, FS filtering_istream 具有 gzip_decom pressor ​​。看到它的 住在Coliru


全code

 的#include<升压/输入输出流/设备/ mapped_file.hpp>
#包括LT&;升压/输入输出流/ filtering_stream.hpp>
#包括LT&;升压/输入输出流/过滤/ gzip.hpp>#包括LT&;升压/精神/有/ qi.hpp>
#包括LT&;升压/精神/有/ qi_match.hpp>
#包括LT&;&迭代器GT;结构边缘:性病::对< INT,INT> {
};的std :: istream的&安培;运营商的GT;>(的std :: istream的&放大器是,边及放大器;边)
{
    使用空间boost ::精神::补气;
    返回的是>>匹配(SRC:> int_>'\\ T'>中的DST:> int_>> EOL,edge.first,edge.second);
}命名空间IO =的boost ::输入输出流;诠释的main()
{
    IO :: mapped_file_source CSV(csv.txt.gz);    IO ::流< IO :: mapped_file_source>文本流(CSV);
    IO :: filtering_istream FS;
    fs.push(IO :: gzip_decom pressor ​​{});
    fs.push(文本流);    对于(
            的std :: istream_iterator<边及GT;它(FS>>的std :: noskipws),最后的;
            它=上!;
            它++)
    {
        性病::法院LT&;<它 - >首先<< 到&所述;&下; IT->第二个<< \\ n;
    }
}

I am trying to learn boost and some template programming in C++ but I am really having such an hard time to implement a simple class for iterating over Gzip files using mapped_file_source. I essentially have an edge list in TSV format such that each line in the gzip file is of the format: <src:int><tab><dst:int>. What I want is to implement a gz_file class that exposes a begin and end iterator over which I can get an edge (std::pair<int,int>) each time I query the iterator.

The problem is the copy constructor which is broken since I cannot known where I am positioned in the gzip file.

Here is the code I have so far:

class gz_graph {
public:
    gz_graph(const char * filename)
    {
        m_file.open(filename);
        if (!m_file.is_open()) {
            throw std::runtime_error("Error opening file");
        }

        m_data = m_file.data();
        m_data_size = m_file.size() / sizeof(m_data[0]);

        auto ret = posix_madvise((void*)m_data, m_data_size, POSIX_MADV_SEQUENTIAL);
    }

    class iterator;

    iterator begin() const
    {
        return iterator(this, false);
    }

    iterator end() const
    {
        return iterator(this, true);
    }

    class iterator : public std::iterator<std::forward_iterator_tag, Edge> {
        public:
            iterator(gz_graph const * ref, bool consumed)
                : m_ref(ref),
                  m_cur_edge(-1, -1),
                  m_consumed(consumed)
            {
                if (!consumed) {
                    initialize();
                    advance();
                }
            }

            iterator(const iterator& x)
                : m_ref(x.m_ref),
                  m_cur_edge(x.m_cur_edge)
            {
                if (!x.m_consumed) {
                    initialize();
                    advance();
                }

                std::cout << "Copy constructor" << std::endl;
            }

            value_type const& operator*() const
            {
                return m_cur_edge;
            }

            value_type const* operator->() const
            {
                return &m_cur_edge;
            }

            iterator& operator++()
            {
                advance();
                return *this;
            }

            bool operator==(iterator const& other) const
            {
                assert(m_ref == other.m_ref);
                return m_cur_edge == other.m_cur_edge;
            }

            bool operator!=(iterator const& other) const
            {
                return !(*this == other);
            }

        private:
            void initialize()
            {
                boost::iostreams::array_source source(m_ref->m_data, m_ref->m_data_size);
                m_in.push(boost::iostreams::gzip_decompressor());
                m_in.push(source);
            }

            void advance()
            {
                std::string line_str;
                if (!getline(m_in, line_str)) {
                    m_consumed = true;
                    m_cur_edge = Edge(-1, -1);
                    return;
                }

                std::vector<std::string> strs;
                boost::split(strs, line_str, boost::is_any_of("\t"));

                if (strs.size() != 2)
                    throw std::runtime_error("Required 2 fields per line");

                int src = boost::lexical_cast<int>(strs.at(0));
                int dst = boost::lexical_cast<int>(strs.at(1));

                m_cur_edge = Edge(src, dst);

//                std::cout << "Read line " << line_str << std::endl;
            }

            gz_graph const * m_ref;
            Edge m_cur_edge;
            boost::iostreams::filtering_istream m_in;
            bool m_consumed;
        };

private:
    boost::iostreams::mapped_file_source m_file;
    char const* m_data;
    size_t m_data_size;
};

解决方案

I would just use a std::istream_iterator here. I'm not sure how exactly to interpret your "input pseudo-code", so let me humor you and do the "complicated" parsing:

struct Edge : std::pair<int, int> { };

std::istream& operator>>(std::istream& is, Edge& edge)
{
    using namespace boost::spirit::qi;
    return is >> match("src:" > int_ > '\t' > "dst:" > int_ >> eol, edge.first, edge.second);
}

I expect you would be happy to have it much simpler, but simpler is easier, right?

Now the main program looks like

for (
        std::istream_iterator<Edge> it(fs >> std::noskipws), end;
        it != end;
        ++it)
{
    std::cout << it->first << " to " << it->second << "\n";
}

Where fs is the filtering_istream that has the gzip_decompressor. See it Live On Coliru


Full Code

#include <boost/iostreams/device/mapped_file.hpp>
#include <boost/iostreams/filtering_stream.hpp>
#include <boost/iostreams/filter/gzip.hpp>

#include <boost/spirit/include/qi.hpp>
#include <boost/spirit/include/qi_match.hpp>
#include <iterator>

struct Edge : std::pair<int, int> {
};

std::istream& operator>>(std::istream& is, Edge& edge)
{
    using namespace boost::spirit::qi;
    return is >> match("src:" > int_ > '\t' > "dst:" > int_ >> eol, edge.first, edge.second);
}

namespace io = boost::iostreams;

int main()
{
    io::mapped_file_source csv("csv.txt.gz");

    io::stream<io::mapped_file_source> textstream(csv);
    io::filtering_istream fs;
    fs.push(io::gzip_decompressor{});
    fs.push(textstream);

    for (
            std::istream_iterator<Edge> it(fs >> std::noskipws), last;
            it != last;
            ++it)
    {
        std::cout << it->first << " to " << it->second << "\n";
    }
}

这篇关于迭代与升压mmaped gzip文件的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!

查看全文
登录 关闭
扫码关注1秒登录
发送“验证码”获取 | 15天全站免登陆