Ticket #4512: options_description.cpp

// Copyright Vladimir Prus 2002-2004.
// Copyright Bertolt Mildner 2004.
// Distributed under the Boost Software License, Version 1.0.
// (See accompanying file LICENSE_1_0.txt
// or copy at http://www.boost.org/LICENSE_1_0.txt)


#define BOOST_PROGRAM_OPTIONS_SOURCE
#include <boost/program_options/config.hpp>
#include <boost/program_options/options_description.hpp>
// FIXME: this is only to get multiple_occurences class
// should move that to a separate headers.
#include <boost/program_options/parsers.hpp>


#include <boost/lexical_cast.hpp>
#include <boost/tokenizer.hpp>
#include <boost/detail/workaround.hpp>
#include <boost/throw_exception.hpp>

#include <cassert>
#include <climits>
#include <cstring>
#include <cstdarg>
#include <sstream>
#include <iterator>
using namespace std;

namespace boost { namespace program_options {

   namespace {

       template< class charT >
       std::basic_string< charT >  tolower_(const std::basic_string< charT >& str)
       {
           std::basic_string< charT > result;
           for (typename std::basic_string< charT >::size_type i = 0; i < str.size(); ++i)
           {
               result.append(1, static_cast< charT >(std::tolower(str[i])));
           }
           return result;
       }

    }  // unnamed namespace


    option_description::option_description()
    {
    }

    option_description::
    option_description(const char* name,
                       const value_semantic* s)
    : m_value_semantic(s)
    {
        this->set_name(name);
    }


    option_description::
    option_description(const char* name,
                       const value_semantic* s,
                       const char* description)
    : m_description(description), m_value_semantic(s)
    {
        this->set_name(name);
    }

    option_description::~option_description()
    {
    }

    option_description::match_result
    option_description::match(const std::string& option,
                              bool approx,
                              bool long_ignore_case,
                              bool short_ignore_case) const
    {
        match_result result = no_match;

        std::string local_long_name((long_ignore_case ? tolower_(m_long_name) : m_long_name));

        if (!local_long_name.empty()) {

            std::string local_option = (long_ignore_case ? tolower_(option) : option);

            if (*local_long_name.rbegin() == '*')
            {
                // The name ends with '*'. Any specified name with the given
                // prefix is OK.
                if (local_option.find(local_long_name.substr(0, local_long_name.length()-1))
                    == 0)
                    result = approximate_match;
            }

            if (local_long_name == local_option)
            {
                result = full_match;
            }
            else if (approx)
            {
                if (local_long_name.find(local_option) == 0)
                {
                    result = approximate_match;
                }
            }
        }

        if (result != full_match)
        {
            std::string local_option(short_ignore_case ? tolower_(option) : option);
            std::string local_short_name(short_ignore_case ? tolower_(m_short_name) : m_short_name);

            if (local_short_name == local_option)
            {
                result = full_match;
            }
        }

        return result;
    }

    const std::string&
    option_description::key(const std::string& option) const
    {
        if (!m_long_name.empty())
            if (m_long_name.find('*') != string::npos)
                // The '*' character means we're long_name
                // matches only part of the input. So, returning
                // long name will remove some of the information,
                // and we have to return the option as specified
                // in the source.
                return option;
            else
                return m_long_name;
        else
            return m_short_name;
    }

    const std::string&
    option_description::long_name() const
    {
        return m_long_name;
    }

    option_description&
    option_description::set_name(const char* _name)
    {
        std::string name(_name);
        string::size_type n = name.find(',');
        if (n != string::npos) {
            assert(n == name.size()-2);
            m_long_name = name.substr(0, n);
            m_short_name = '-' + name.substr(n+1,1);
        } else {
            m_long_name = name;
        }
        return *this;
    }

    const std::string&
    option_description::description() const
    {
        return m_description;
    }

    shared_ptr<const value_semantic>
    option_description::semantic() const
    {
        return m_value_semantic;
    }

    std::string
    option_description::format_name() const
    {
        if (!m_short_name.empty())
            return string(m_short_name).append(" [ --").
            append(m_long_name).append(" ]");
        else
            return string("--").append(m_long_name);
    }

    std::string
    option_description::format_parameter() const
    {
        if (m_value_semantic->max_tokens() != 0)
            return m_value_semantic->name();
        else
            return "";
    }

    options_description_easy_init::
    options_description_easy_init(options_description* owner)
    : owner(owner)
    {}

    options_description_easy_init&
    options_description_easy_init::
    operator()(const char* name,
               const char* description)
    {
        // Create untypes semantic which accepts zero tokens: i.e. 
        // no value can be specified on command line.
        // FIXME: does not look exception-safe
        shared_ptr<option_description> d(
            new option_description(name, new untyped_value(true), description));

        owner->add(d);
        return *this;
    }

    options_description_easy_init&
    options_description_easy_init::
    operator()(const char* name,
               const value_semantic* s)
    {
        shared_ptr<option_description> d(new option_description(name, s));
        owner->add(d);
        return *this;
    }

    options_description_easy_init&
    options_description_easy_init::
    operator()(const char* name,
               const value_semantic* s,
               const char* description)
    {
        shared_ptr<option_description> d(new option_description(name, s, description));

        owner->add(d);
        return *this;
    }

    const unsigned options_description::m_default_line_length = 80;

    options_description::options_description(unsigned line_length,
                                             unsigned min_description_length)
    : m_line_length(line_length)
    , m_min_description_length(min_description_length)
    {
        // we require a space between the option and description parts, so add 1.
        assert(m_min_description_length < m_line_length - 1);
    }

    options_description::options_description(const std::string& caption,
                                             unsigned line_length,
                                             unsigned min_description_length)
    : m_caption(caption)
    , m_line_length(line_length)
    , m_min_description_length(min_description_length)
    {
        // we require a space between the option and description parts, so add 1.
        assert(m_min_description_length < m_line_length - 1);
    }

    void
    options_description::add(shared_ptr<option_description> desc)
    {
        m_options.push_back(desc);
        belong_to_group.push_back(false);
    }

    options_description&
    options_description::add(const options_description& desc)
    {
        shared_ptr<options_description> d(new options_description(desc));
        groups.push_back(d);

        for (size_t i = 0; i < desc.m_options.size(); ++i) {
            add(desc.m_options[i]);
            belong_to_group.back() = true;
        }

        return *this;
    }

    options_description_easy_init
    options_description::add_options()
    {
        return options_description_easy_init(this);
    }

    const option_description&
    options_description::find(const std::string& name,
                              bool approx,
                              bool long_ignore_case,
                              bool short_ignore_case) const
    {
        const option_description* d = find_nothrow(name, approx,
                                       long_ignore_case, short_ignore_case);
        if (!d)
            boost::throw_exception(unknown_option(name));
        return *d;
    }

    const std::vector< shared_ptr<option_description> >&
    options_description::options() const
    {
        return m_options;
    }

    const option_description*
    options_description::find_nothrow(const std::string& name,
                                      bool approx,
                                      bool long_ignore_case,
                                      bool short_ignore_case) const
    {
        int found = -1;
        vector<string> approximate_matches;
        vector<string> full_matches;

        // We use linear search because matching specified option
        // name with the declaraed option name need to take care about
        // case sensitivity and trailing '*' and so we can't use simple map.
        for(unsigned i = 0; i < m_options.size(); ++i)
        {
            option_description::match_result r =
                m_options[i]->match(name, approx, long_ignore_case, short_ignore_case);

            if (r == option_description::no_match)
                continue;

            if (r == option_description::full_match)
            {
                full_matches.push_back(m_options[i]->key(name));

                // Use this full match regardless of other matches found.
                found = i;
            }
            else
            {
                // FIXME: the use of 'key' here might not
                // be the best approach.
                approximate_matches.push_back(m_options[i]->key(name));

                // If this is the first match found the use it.  If         
                // a later full match is found that will be used instead.
                if (found == -1)
                  found = i;
            }
        }
        if (full_matches.size() > 1)
            boost::throw_exception(
                ambiguous_option(name, full_matches));

        // If we have a full match, and an approximate match,
        // ignore approximate match instead of reporting error.
        // Say, if we have options "all" and "all-chroots", then
        // "--all" on the command line should select the first one,
        // without ambiguity.
        if (full_matches.empty() && approximate_matches.size() > 1)
            boost::throw_exception(
                ambiguous_option(name, approximate_matches));

        return found>=0
          ? m_options[found].get()
          : 0;
    }

    BOOST_PROGRAM_OPTIONS_DECL
    std::ostream& operator<<(std::ostream& os, const options_description& desc)
    {
        desc.print(os);
        return os;
    }

    namespace {

        /* Given a string 'par', that contains no newline characters
           outputs it to 'os' with wordwrapping, that is, as several
           line.

           Each output line starts with 'indent' space characters, 
           following by characters from 'par'. The total length of
           line is no longer than 'line_length'.
                                      
        */
        void format_paragraph(std::ostream& os,
                              std::string par,
                              unsigned indent,
                              unsigned line_length)
        {
            // Through reminder of this function, 'line_length' will
            // be the length available for characters, not including
            // indent.
            assert(indent < line_length);
            line_length -= indent;

            // index of tab (if present) is used as additional indent relative
            // to first_column_width if paragrapth is spanned over multiple
            // lines if tab is not on first line it is ignored
            string::size_type par_indent = par.find('\t');

            if (par_indent == string::npos)
            {
                par_indent = 0;
            }
            else
            {
                // only one tab per paragraph allowed
                if (count(par.begin(), par.end(), '\t') > 1)
                {
                    boost::throw_exception(program_options::error(
                        "Only one tab per paragraph is allowed"));
                }

                // erase tab from string
                par.erase(par_indent, 1);

                // this assert may fail due to user error or 
                // environment conditions!
                assert(par_indent < line_length);

                // ignore tab if not on first line
                if (par_indent >= line_length)
                {
                    par_indent = 0;
                }
            }

            if (par.size() < line_length)
            {
                os << par;
            }
            else
            {
                string::const_iterator       line_begin = par.begin();
                const string::const_iterator par_end = par.end();

                bool first_line = true; // of current paragraph!        

                while (line_begin < par_end)  // paragraph lines
                {
                    if (!first_line)
                    {
                        // If line starts with space, but second character
                        // is not space, remove the leading space.
                        // We don't remove double spaces because those
                        // might be intentianal.
                        if ((*line_begin == ' ') &&
                            ((line_begin + 1 < par_end) &&
                             (*(line_begin + 1) != ' ')))
                        {
                            line_begin += 1;  // line_begin != line_end
                        }
                    }

                    // Take care to never increment the iterator past
                    // the end, since MSVC 8.0 (brokenly), assumes that
                    // doing that, even if no access happens, is a bug.
                    unsigned remaining = distance(line_begin, par_end);
                    string::const_iterator line_end = line_begin +
                        ((remaining < line_length) ? remaining : line_length);

                    // prevent chopped words
                    // Is line_end between two non-space characters?
                    if ((*(line_end - 1) != ' ') &&
                        ((line_end < par_end) && (*line_end != ' ')))
                    {
                        // find last ' ' in the second half of the current paragraph line
                        string::const_iterator last_space =
                            find(reverse_iterator<string::const_iterator>(line_end),
                                 reverse_iterator<string::const_iterator>(line_begin),
                                 ' ')
                            .base();

                        if (last_space != line_begin)
                        {
                            // is last_space within the second half ot the 
                            // current line
                            if (static_cast<unsigned>(distance(last_space, line_end)) <
                                (line_length / 2))
                            {
                                line_end = last_space;
                            }
                        }
                    } // prevent chopped words

                    // write line to stream
                    copy(line_begin, line_end, ostream_iterator<char>(os));

                    if (first_line)
                    {
                        indent += par_indent;
                        line_length -= par_indent; // there's less to work with now
                        first_line = false;
                    }

                    // more lines to follow?
                    if (line_end != par_end)
                    {
                        os << '\n';

                        for(unsigned pad = indent; pad > 0; --pad)
                        {
                            os.put(' ');
                        }
                    }

                    // next line starts after of this line
                    line_begin = line_end;
                } // paragraph lines
            }
        }

        void format_description(std::ostream& os,
                                const std::string& desc,
                                unsigned first_column_width,
                                unsigned line_length)
        {
            // we need to use one char less per line to work correctly if actual
            // console has longer lines
            assert(line_length > 1);
            if (line_length > 1)
            {
                --line_length;
            }

            // line_length must be larger than first_column_width
            // this assert may fail due to user error or environment conditions!
            assert(line_length > first_column_width);

            // Note: can't use 'tokenizer' as name of typedef -- borland
            // will consider uses of 'tokenizer' below as uses of
            // boost::tokenizer, not typedef.
            typedef boost::tokenizer<boost::char_separator<char> > tok;

            tok paragraphs(
                desc,
                char_separator<char>("\n", "", boost::keep_empty_tokens));

            tok::const_iterator       par_iter = paragraphs.begin();
            const tok::const_iterator par_end = paragraphs.end();

            while (par_iter != par_end)  // paragraphs
            {
                format_paragraph(os, *par_iter, first_column_width,
                                 line_length);

                ++par_iter;

                // prepair next line if any
                if (par_iter != par_end)
                {
                    os << '\n';

                    for(unsigned pad = first_column_width; pad > 0; --pad)
                    {
                        os.put(' ');
                    }
                }
            }  // paragraphs
        }

        void format_one(std::ostream& os, const option_description& opt,
                        unsigned first_column_width, unsigned line_length)
        {
            stringstream ss;
            ss << "  " << opt.format_name() << ' ' << opt.format_parameter();

            // Don't use ss.rdbuf() since g++ 2.96 is buggy on it.
            os << ss.str();

            if (!opt.description().empty())
            {
                if (ss.str().size() >= first_column_width)
                {
                   os.put('\n'); // first column is too long, lets put description in new line
                   for (unsigned pad = first_column_width; pad > 0; --pad)
                   {
                      os.put(' ');
                   }
                } else {
                   for(unsigned pad = first_column_width - ss.str().size(); pad > 0; --pad)
                   {
                      os.put(' ');
                   }
                }

                format_description(os, opt.description(),
                                   first_column_width, line_length);
            }
        }
    }

    void
    options_description::print(std::ostream& os) const
    {
        if (!m_caption.empty())
            os << m_caption << ":\n";

        /* Find the maximum width of the option column */
        unsigned width(23);
        unsigned i; // vc6 has broken for loop scoping
        for (i = 0; i < m_options.size(); ++i)
        {
            const option_description& opt = *m_options[i];
            stringstream ss;
            ss << "  " << opt.format_name() << ' ' << opt.format_parameter();
            width = (max)(width, static_cast<unsigned>(ss.str().size()));
        }
        /* this is the column were description should start, if first
           column is longer, we go to a new line */
        const unsigned start_of_description_column = m_line_length - m_min_description_length;

        width = (min)(width, start_of_description_column-1);

        /* add an additional space to improve readability */
        ++width;

        /* The options formatting style is stolen from Subversion. */
        for (i = 0; i < m_options.size(); ++i)
        {
            if (belong_to_group[i])
                continue;

            const option_description& opt = *m_options[i];

            format_one(os, opt, width, m_line_length);

            os << "\n";
        }

        for (unsigned j = 0; j < groups.size(); ++j) {
            os << "\n" << *groups[j];
        }
    }

}}