处理好与Boost的program_options复杂的选项 [英] Handle complex options with Boost's program_options

问题描述

我有使用不同的多层次模型生成图表的程序。每个多级模型包括生成一个较小的种子图形（比方说，50个节点），其可从几种模式中创建的。（例如 - 对于每个可能的边缘，选择以概率p以包括它）

I have a program that generates graphs using different multi-level models. Each multi-level model consists of a generation of a smaller seed graph (say, 50 nodes) which can be created from several models (for example - for each possible edge, choose to include it with probability p).

种子图形生成后，将图形展开成一个较大的（例如1000个节点），利用另一组模型之一

After the seed graph generation, the graph is expanded into a larger one (say 1000 nodes), using one of another set of models.

在每两个阶段，每一个模型需要不同数目的参数。

In each of the two stages, each model require a different number of parameters.

我想是有program_options解析不同的可能的参数，根据该模型的名称。

I would like to be have program_options parse the different possible parameters, according to the names of the models.

举例来说，假设我有两个种子图形模型：SA，它具有1的参数，以及SB，它有两个。也为扩展的一部分，我有两种型号：A和B，再以1和2的参数，分别为。我希望能够做一些事情，如：

For example, say I have two seed graphs models: SA, which has 1 parameters, and SB, which has two. Also for the expansion part, I have two models: A and B, again with 1 and 2 parameters, respectively. I would like to be able do something like:

./graph_generator --seed=SA 0.1 --expansion=A 0.2
./graph_generator --seed=SB 0.1 3 --expansion=A 0.2
./graph_generator --seed=SA 0.1 --expansion=B 10 20
./graph_generator --seed=SB 0.1 3 --expansion=B 10 20

和正确分析的参数。是，即使可能吗？

and have the parameters parsed correctly. Is that even possible?

推荐答案

通过使用的自定义验证和<一个href=\"http://www.boost.org/doc/libs/release/doc/html/boost/program_options/typed_value.html#id941480-bb\">boost::program_options::value::multitoken,你可以达到预期的效果：

By using a custom validator and boost::program_options::value::multitoken, you can achieve the desired result:

#include <iostream>
#include <boost/lexical_cast.hpp>
#include <boost/optional.hpp>
#include <boost/program_options.hpp>

// Holds parameters for seed/expansion model
struct Model
{
    std::string type;
    boost::optional<float> param1;
    boost::optional<float> param2;
};

// Called by program_options to parse a set of Model arguments
void validate(boost::any& v, const std::vector<std::string>& values,
              Model*, int)
{
    Model model;
    // Extract tokens from values string vector and populate Model struct.
    if (values.size() == 0)
    {
        throw boost::program_options::validation_error(
            "Invalid model specification");
    }
    model.type = values.at(0); // Should validate for A/B
    if (values.size() >= 2)
        model.param1 = boost::lexical_cast<float>(values.at(1));
    if (values.size() >= 3)
        model.param2 = boost::lexical_cast<float>(values.at(2));

    v = model;
}

int main(int argc, char* argv[])
{
    Model seedModel, expansionModel;

    namespace po = boost::program_options;
    po::options_description options("Generic options");
    options.add_options()
        ("seed",
             po::value<Model>(&seedModel)->multitoken(),
             "seed graph model")
        ("expansion",
             po::value<Model>(&expansionModel)->multitoken(),
             "expansion model")
        ;

    po::variables_map vm;
    po::store(po::parse_command_line(argc, argv, options), vm);
    po::notify(vm);

    std::cout << "Seed type: " << seedModel.type << "\n";
    if (seedModel.param1)
        std::cout << "Seed param1: " << *(seedModel.param1) << "\n";
    if (seedModel.param2)
        std::cout << "Seed param2: " << *(seedModel.param2) << "\n";

    std::cout << "Expansion type: " << expansionModel.type << "\n";
    if (expansionModel.param1)
        std::cout << "Expansion param1: " << *(expansionModel.param1) << "\n";
    if (expansionModel.param2)
        std::cout << "Expansion param2: " << *(expansionModel.param2) << "\n";

    return 0;
}

的验证功能可能需要更多的严谨性，但你的想法。

The validate function probably needs more rigor, but you get the idea.

这编译并为我工作。

这篇关于处理好与Boost的program_options复杂的选项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！