Simple expression parser example using Boost::Spirit?

Question

Is anyone aware of an online resource where I can find out how to write a simple expression parser using Boost::Spirit?.

I do not necessarily need to evaluate the expression, but I need to parse it and be able to return a boolean to indicate whether the expression is parsable or not (e.g. brackets not matching etc).

I need the parser to be able recognise function names (e.g. foo and foobar), so this would also be a useful example to help me learn writing BNF notation.

The expressions will be normal arithmetic equations, i.e. comprising of the following symbols:

1. opening/closing brackets
2. arithmetic operators
3. recognized function names, and check for their required arguments

Lowe

Did you look into spirits documentation and examples? – Larina

Spirit's documentation is not nearly as simple as I'd wish for. I managed to learn by it, but better tutorial would certainly have made learning easier. – Jaunita

Thanks Tronic. That was very much my view when I perused through the docs on the Spirit home page. – Lowe

Have you recently looked at Spirit's web site (boost-spirit.com)? There is quite some material available, not really related to your expression parser question, but very useful as an addendum to the main documentation. – Donnenfeld

I am in the same situation with you. Any update of this question? – Optimistic

Answer 1

Here's some old Spirit prototype code I had laying around:

#include <iostream>
#include <algorithm>
#include <vector>
#include <string>
#include <exception>
#include <iterator>
#include <sstream>
#include <list>

#include <boost/spirit.hpp>
#include <boost/shared_ptr.hpp>

using namespace std;
using namespace boost::spirit;
using namespace boost;

void g(unsigned int i)
{   
    cout << "row: " << i << endl;
}

struct u
{
    u(const char* c): s(c) {}
    void operator()(const char* first, const char* last) const
    {
        cout << s << ": " << string(first, last) << endl;   
    }
private:
    string s;
};


struct Exp
{
};

struct Range: public Exp
{
};

struct Index: public Exp
{
};

struct String: public Exp
{
};

struct Op
{
    virtual ~Op() = 0;
    virtual string name() = 0;
};

Op::~Op() {}

struct CountIf: public Op
{
    string name() { return "CountIf"; }
};

struct Sum: public Op
{
    string name() { return "Sum"; }
};

struct Statement
{
    virtual ~Statement() = 0;
    virtual void print() = 0;
};

Statement::~Statement() {}

struct Formula: public Statement
{
    Formula(const char* first, const char* last): s(first, last), op(new CountIf)
    {
        typedef rule<phrase_scanner_t> r_t;

        r_t r_index     = (+alpha_p)[u("col")] >> uint_p[&g];
        r_t r_range     = r_index >> ':' >> r_index;
        r_t r_string    = ch_p('\"') >> *alnum_p >> '\"';
        r_t r_exp       = r_range | r_index | r_string; // will invoke actions for index twice due to range
        r_t r_list      = !(r_exp[u("arg")] % ',');
        r_t r_op        = as_lower_d["countif"] | as_lower_d["sum"];
        r_t r_formula   = r_op >> '(' >> r_list >> ')';

        cout << s << ": matched: " << boolalpha << parse(s.c_str(), r_formula, space_p).full << endl; 
    }
    void print() { cout << "Formula: " << s << " / " << op->name() << endl; }
private:
    string s;
    shared_ptr<Op> op;
    list<shared_ptr<Exp> > exp_list;
};

struct Comment: public Statement
{
    Comment(const char* first, const char* last): comment(first, last) {}
    void print() {cout << "Comment: " << comment << endl; }
private:
    string comment;
};


struct MakeFormula
{
    MakeFormula(list<shared_ptr<Statement> >& list_): list(list_) {}
    void operator()(const char* first, const char* last) const
    {
        cout << "MakeFormula: " << string(first, last) << endl;
        list.push_back(shared_ptr<Statement>(new Formula(first, last)));
    }
private:
    list<shared_ptr<Statement> >& list;
};

struct MakeComment
{
    MakeComment(list<shared_ptr<Statement> >& list_): list(list_) {}
    void operator()(const char* first, const char* last) const
    {
        cout << "MakeComment: " << string(first, last) << endl;
        list.push_back(shared_ptr<Statement>(new Comment(first, last)));
    }
private:
    list<shared_ptr<Statement> >& list;
};


int main(int argc, char* argv[])
try
{
    //typedef vector<string> v_t;
    //v_t v(argv + 1, argv + argc);
    // copy(v.begin(), v.end(), ostream_iterator<v_t::value_type>(cout, "\n"));

    string s;
    getline(cin, s);

    //        =COUNTIF(J2:J36, "Abc")

    typedef list<shared_ptr<Statement> > list_t;
    list_t list;

    typedef rule<phrase_scanner_t> r_t;

    r_t r_index     = (+alpha_p)[u("col")] >> uint_p[&g];
    r_t r_range     = r_index >> ':' >> r_index;
    r_t r_string    = ch_p('\"') >> *alnum_p >> '\"';
    r_t r_exp       = r_range | r_index | r_string; // will invoke actions for index twice due to range
    r_t r_list      = !(r_exp[u("arg")] % ',');
    r_t r_op        = as_lower_d["countif"] | as_lower_d["sum"];
    r_t r_formula   = r_op >> '(' >> r_list >> ')';
    r_t r_statement = (ch_p('=')  >> r_formula   [MakeFormula(list)])
                    | (ch_p('\'') >> (*anychar_p)[MakeComment(list)])
                    ;

    cout << s << ": matched: " << boolalpha << parse(s.c_str(), r_statement, space_p).full << endl; 

    for (list_t::const_iterator it = list.begin(); it != list.end(); ++it)
    {
        (*it)->print();
    }
}
catch(const exception& ex)
{
    cerr << "Error: " << ex.what() << endl;
}

Try running it and entering a line like:

=COUNTIF(J2:J36, "Abc")

Amphimixis

Nice example! Well, but is this really simple? – Cloy

Well, it's not completely trivial, but it's still only a page (maybe double-sided!) worth of printed code that I wrote on a domestic flight. :) – Amphimixis

Answer 2

The current version of Spirit (V2.x) contains a whole series of calculator examples from the very simple to a full fledged mini-c interpreter. You should have a look there as those are a perfect starting point for writing your own expression parser.

Donnenfeld

Answer 3

I'm not sure if this qualifies as simple either, but I've used this uri-grammar available at http://code.google.com/p/uri-grammar/source/browse/trunk/src/uri/grammar.hpp. It may not be trivial, but at least its parsing something that you probably understand already (URIs). When reading these grammars, its best to read from the bottom up, since that's where the most generic tokens tend to be defined.

Katinakatine