我将避开你的大部分代码，原因很简单，经验告诉我 Lex 和 utree 通常不是你想要使用的。
你想要的是定义一个 AST 来表示你的命令语言，然后想出一个语法来构建它。
AST
namespace Ast {
struct NoValue {
bool operator==(NoValue const &) const { return true; }
};
template <typename Tag> struct GenericCommand {};
namespace tag {
struct boot;
struct help;
struct load;
struct exit;
struct set;
struct show;
};
template <> struct GenericCommand<tag::load> { std::string name; };
template <> struct GenericCommand<tag::set> {
std::string property;
boost::variant<NoValue, std::string, bool> value; // optional
};
using BootCmd = GenericCommand<tag::boot>;
using HelpCmd = GenericCommand<tag::help>;
using ExitCmd = GenericCommand<tag::exit>;
using ShowCmd = GenericCommand<tag::show>;
using LoadCmd = GenericCommand<tag::load>;
using SetCmd = GenericCommand<tag::set>;
using Command = boost::variant<BootCmd, HelpCmd, ExitCmd, ShowCmd, LoadCmd, SetCmd>;
using Commands = std::list<Command>;
}
完整代码仅添加调试输出帮助器。这是完整的 Fusion Adaption：
BOOST_FUSION_ADAPT_TPL_STRUCT((Tag), (Ast::GenericCommand) (Tag), )
BOOST_FUSION_ADAPT_STRUCT(Ast::LoadCmd, name)
BOOST_FUSION_ADAPT_STRUCT(Ast::SetCmd, property, value)
语法
这里我做了一些选择：
-
让我们让空格和大小写不敏感，允许使用行分隔的命令：（另请参阅Boost spirit skipper issues）
start = skip(blank) [lazy_command % eol];
-
让我们使用Nabialek Trick 将命令与前缀相关联。我使用了一个非常简单的 sn-p 代码来生成唯一前缀：
std::set<std::string> const verbs { "boot", "exit", "help", "-help", "/help", "load", "quit", "set", "show", };
for (auto const full : verbs)
for (auto partial=full; partial.length(); partial.resize(partial.size()-1)) {
auto n = std::distance(verbs.lower_bound(partial), verbs.upper_bound(full));
if (n < 2) std::cout << "(\"" << partial << "\", &" << full << "_command)\n";
}
您可以对属性执行相同的操作，但我认为当前设置更简单：
template <typename Iterator>
struct command_grammar : qi::grammar<Iterator, Ast::Commands()> {
command_grammar() : command_grammar::base_type(start) {
using namespace qi;
start = skip(blank) [lazy_command % eol];
// nabialek trick
lazy_command = no_case [ commands [ _a = _1 ] > lazy(*_a) [ _val = _1 ] ];
on_off.add("on", true)("off", false);
commands.add
("-help", &help_command) ("-hel", &help_command) ("-he", &help_command) ("-h", &help_command)
("/help", &help_command) ("/hel", &help_command) ("/he", &help_command) ("/h", &help_command)
("help", &help_command) ("hel", &help_command) ("he", &help_command) ("h", &help_command)
("boot", &boot_command) ("boo", &boot_command) ("bo", &boot_command) ("b", &boot_command)
("exit", &exit_command) ("exi", &exit_command) ("ex", &exit_command) ("e", &exit_command)
("quit", &exit_command) ("qui", &exit_command) ("qu", &exit_command) ("q", &exit_command)
("load", &load_command) ("loa", &load_command) ("lo", &load_command) ("l", &load_command)
("set", &set_command) ("se", &set_command)
("show", &show_command) ("sho", &show_command) ("sh", &show_command);
quoted_string = '"' >> +~char_('"') >> '"';
// nullary commands
boot_command_ = eps;
exit_command_ = eps;
help_command_ = eps;
show_command_ = eps;
// non-nullary commands
load_command_ = quoted_string;
drive_ = char_("A-Z") >> ':';
set_command_ = no_case[lit("drive")|"driv"|"dri"|"dr"] >> attr("DRIVE") >> drive_
| no_case[ (lit("debug")|"debu"|"deb"|"de") >> attr("DEBUG") >> on_off ]
| no_case[ (lit("trace")|"trac"|"tra"|"tr"|"t") >> attr("TRACE") >> on_off ]
;
BOOST_SPIRIT_DEBUG_NODES(
(start)(lazy_command)
(boot_command) (exit_command) (help_command) (show_command) (set_command) (load_command)
(boot_command_)(exit_command_)(help_command_)(show_command_)(set_command_)(load_command_)
(quoted_string)(drive_)
)
on_error<fail>(start, error_handler_(_4, _3, _2));
on_error<fail>(lazy_command, error_handler_(_4, _3, _2));
boot_command = boot_command_;
exit_command = exit_command_;
help_command = help_command_;
load_command = load_command_;
exit_command = exit_command_;
set_command = set_command_;
show_command = show_command_;
}
private:
struct error_handler_t {
template <typename...> struct result { typedef void type; };
void operator()(qi::info const &What, Iterator Err_pos, Iterator Last) const {
std::cout << "Error! Expecting " << What << " here: \"" << std::string(Err_pos, Last) << "\"" << std::endl;
}
};
boost::phoenix::function<error_handler_t> const error_handler_ = error_handler_t {};
qi::rule<Iterator, Ast::Commands()> start;
using Skipper = qi::blank_type;
using CommandRule = qi::rule<Iterator, Ast::Command(), Skipper>;
qi::symbols<char, bool> on_off;
qi::symbols<char, CommandRule const*> commands;
qi::rule<Iterator, std::string()> drive_property, quoted_string, drive_;
qi::rule<Iterator, Ast::Command(), Skipper, qi::locals<CommandRule const*> > lazy_command;
CommandRule boot_command, exit_command, help_command, load_command, set_command, show_command;
qi::rule<Iterator, Ast::BootCmd(), Skipper> boot_command_;
qi::rule<Iterator, Ast::ExitCmd(), Skipper> exit_command_;
qi::rule<Iterator, Ast::HelpCmd(), Skipper> help_command_;
qi::rule<Iterator, Ast::LoadCmd(), Skipper> load_command_;
qi::rule<Iterator, Ast::SetCmd(), Skipper> set_command_;
qi::rule<Iterator, Ast::ShowCmd(), Skipper> show_command_;
};
测试用例
Live On Coliru
int main() {
typedef std::string::const_iterator It;
command_grammar<It> const commands;
for (std::string const input : {
"help",
"set drive C:",
"SET DRIVE C:",
"loAD \"XYZ\"",
"load \"anything \nat all\"",
// multiline
"load \"ABC\"\nhelp\n-he\n/H\nsh\nse t off\nse debug ON\nb\nq"
})
{
std::cout << "----- '" << input << "' -----\n";
It f = input.begin(), l = input.end();
Ast::Commands parsed;
bool result = parse(f, l, commands, parsed);
if (result) {
for (auto& cmd : parsed) {
std::cout << "Parsed " << cmd << "\n";
}
} else {
std::cout << "Parse failed\n";
}
if (f != l) {
std::cout << "Remaining unparsed '" << std::string(f, l) << "'\n";
}
}
}
打印：
----- 'help' -----
Parsed HELP ()
----- 'set drive C:' -----
Parsed SET (DRIVE C)
----- 'SET DRIVE C:' -----
Parsed SET (DRIVE C)
----- 'loAD "XYZ"' -----
Parsed LOAD (XYZ)
----- 'load "anything
at all"' -----
Parsed LOAD (anything
at all)
----- 'load "ABC"
help
-he
/H
sh
se t off
se debug ON
b
q' -----
Parsed LOAD (ABC)
Parsed HELP ()
Parsed HELP ()
Parsed HELP ()
Parsed SHOW ()
Parsed SET (TRACE 0)
Parsed SET (DEBUG 1)
Parsed BOOT ()
Parsed EXIT ()
完整列表
Live On Coliru
//#define BOOST_SPIRIT_DEBUG
#include <boost/fusion/include/io.hpp>
#include <boost/fusion/adapted/struct.hpp>
#include <boost/spirit/include/phoenix.hpp>
#include <boost/spirit/include/qi.hpp>
namespace qi = boost::spirit::qi;
namespace Ast {
struct NoValue {
bool operator==(NoValue const &) const { return true; }
friend std::ostream& operator<<(std::ostream& os, NoValue) { return os; }
};
template <typename Tag> struct GenericCommand {};
namespace tag {
struct boot {};
struct help {};
struct load {};
struct exit {};
struct set {};
struct show {};
static std::ostream& operator<<(std::ostream& os, boot) { return os << "BOOT"; }
static std::ostream& operator<<(std::ostream& os, help) { return os << "HELP"; }
static std::ostream& operator<<(std::ostream& os, load) { return os << "LOAD"; }
static std::ostream& operator<<(std::ostream& os, exit) { return os << "EXIT"; }
static std::ostream& operator<<(std::ostream& os, set ) { return os << "SET"; }
static std::ostream& operator<<(std::ostream& os, show) { return os << "SHOW"; }
};
template <> struct GenericCommand<tag::load> { std::string name; };
template <> struct GenericCommand<tag::set> {
std::string property;
boost::variant<NoValue, std::string, bool> value; // optional
};
using BootCmd = GenericCommand<tag::boot>;
using HelpCmd = GenericCommand<tag::help>;
using ExitCmd = GenericCommand<tag::exit>;
using ShowCmd = GenericCommand<tag::show>;
using LoadCmd = GenericCommand<tag::load>;
using SetCmd = GenericCommand<tag::set>;
using Command = boost::variant<BootCmd, HelpCmd, ExitCmd, ShowCmd, LoadCmd, SetCmd>;
using Commands = std::list<Command>;
template <typename Tag>
static inline std::ostream& operator<<(std::ostream& os, Ast::GenericCommand<Tag> const& command) {
return os << Tag{} << " " << boost::fusion::as_vector(command);
}
}
BOOST_FUSION_ADAPT_TPL_STRUCT((Tag), (Ast::GenericCommand) (Tag), )
BOOST_FUSION_ADAPT_STRUCT(Ast::LoadCmd, name)
BOOST_FUSION_ADAPT_STRUCT(Ast::SetCmd, property, value)
template <typename Iterator>
struct command_grammar : qi::grammar<Iterator, Ast::Commands()> {
command_grammar() : command_grammar::base_type(start) {
using namespace qi;
start = skip(blank) [lazy_command % eol];
// nabialek trick
lazy_command = no_case [ commands [ _a = _1 ] > lazy(*_a) [ _val = _1 ] ];
on_off.add("on", true)("off", false);
commands.add
("-help", &help_command) ("-hel", &help_command) ("-he", &help_command) ("-h", &help_command)
("/help", &help_command) ("/hel", &help_command) ("/he", &help_command) ("/h", &help_command)
("help", &help_command) ("hel", &help_command) ("he", &help_command) ("h", &help_command)
("boot", &boot_command) ("boo", &boot_command) ("bo", &boot_command) ("b", &boot_command)
("exit", &exit_command) ("exi", &exit_command) ("ex", &exit_command) ("e", &exit_command)
("quit", &exit_command) ("qui", &exit_command) ("qu", &exit_command) ("q", &exit_command)
("load", &load_command) ("loa", &load_command) ("lo", &load_command) ("l", &load_command)
("set", &set_command) ("se", &set_command)
("show", &show_command) ("sho", &show_command) ("sh", &show_command);
quoted_string = '"' >> +~char_('"') >> '"';
// nullary commands
boot_command_ = eps;
exit_command_ = eps;
help_command_ = eps;
show_command_ = eps;
// non-nullary commands
load_command_ = quoted_string;
drive_ = char_("A-Z") >> ':';
set_command_ = no_case[lit("drive")|"driv"|"dri"|"dr"] >> attr("DRIVE") >> drive_
| no_case[ (lit("debug")|"debu"|"deb"|"de") >> attr("DEBUG") >> on_off ]
| no_case[ (lit("trace")|"trac"|"tra"|"tr"|"t") >> attr("TRACE") >> on_off ]
;
BOOST_SPIRIT_DEBUG_NODES(
(start)(lazy_command)
(boot_command) (exit_command) (help_command) (show_command) (set_command) (load_command)
(boot_command_)(exit_command_)(help_command_)(show_command_)(set_command_)(load_command_)
(quoted_string)(drive_)
)
on_error<fail>(start, error_handler_(_4, _3, _2));
on_error<fail>(lazy_command, error_handler_(_4, _3, _2));
boot_command = boot_command_;
exit_command = exit_command_;
help_command = help_command_;
load_command = load_command_;
exit_command = exit_command_;
set_command = set_command_;
show_command = show_command_;
}
private:
struct error_handler_t {
template <typename...> struct result { typedef void type; };
void operator()(qi::info const &What, Iterator Err_pos, Iterator Last) const {
std::cout << "Error! Expecting " << What << " here: \"" << std::string(Err_pos, Last) << "\"" << std::endl;
}
};
boost::phoenix::function<error_handler_t> const error_handler_ = error_handler_t {};
qi::rule<Iterator, Ast::Commands()> start;
using Skipper = qi::blank_type;
using CommandRule = qi::rule<Iterator, Ast::Command(), Skipper>;
qi::symbols<char, bool> on_off;
qi::symbols<char, CommandRule const*> commands;
qi::rule<Iterator, std::string()> drive_property, quoted_string, drive_;
qi::rule<Iterator, Ast::Command(), Skipper, qi::locals<CommandRule const*> > lazy_command;
CommandRule boot_command, exit_command, help_command, load_command, set_command, show_command;
qi::rule<Iterator, Ast::BootCmd(), Skipper> boot_command_;
qi::rule<Iterator, Ast::ExitCmd(), Skipper> exit_command_;
qi::rule<Iterator, Ast::HelpCmd(), Skipper> help_command_;
qi::rule<Iterator, Ast::LoadCmd(), Skipper> load_command_;
qi::rule<Iterator, Ast::SetCmd(), Skipper> set_command_;
qi::rule<Iterator, Ast::ShowCmd(), Skipper> show_command_;
};
int main() {
typedef std::string::const_iterator It;
command_grammar<It> const commands;
for (std::string const input : {
"help",
"set drive C:",
"SET DRIVE C:",
"loAD \"XYZ\"",
"load \"anything \nat all\"",
// multiline
"load \"ABC\"\nhelp\n-he\n/H\nsh\nse t off\nse debug ON\nb\nq"
})
{
std::cout << "----- '" << input << "' -----\n";
It f = input.begin(), l = input.end();
Ast::Commands parsed;
bool result = parse(f, l, commands, parsed);
if (result) {
for (auto& cmd : parsed) {
std::cout << "Parsed " << cmd << "\n";
}
} else {
std::cout << "Parse failed\n";
}
if (f != l) {
std::cout << "Remaining unparsed '" << std::string(f, l) << "'\n";
}
}
}
脚本后
问。如何注释解析器以使用 utree 创建 AST？
问。如何在 utree 构建后遍历它，以发现被解析的内容？
问。我想添加一个评论字符，“！”。那么，在那之后我怎么能忽略所有的东西——除非它出现在带引号的字符串中？
问。为什么当我输入无效输入时我的错误处理程序没有被调用？
- 因为您没有标记期望点（
operator> 而不是operator>>）。如果您不这样做，则无法匹配子表达式只会回溯。
问。如何使命令标记不区分大小写，但不更改带引号的字符串的内容？