autogram/src/engine.cpp

// robinsonizer engine by Bob Polis
// copyright (c) 1994-2019

// TODO Use correct plurals: adjust numeral file format to include this.

// TODO Don't assume 26-letter alphabet, but allow for arbitrary character list.

// TODO Switch to get_long_options in main.

// C++
#include <iostream>
#include <string>
#include <fstream>
#include <cstring>
using namespace std;

// libcommon
#include <libcommon.hpp>

// Project
#include "engine.hpp"

const int s_index = 's' - 'a';

engine::engine(std::string start,
                 int maxiter,
                 int maxseed,
                 std::string numerals_file,
                 std::string characters_file,
                 unsigned int vl,
                 robinsonizer_mode mode,
                 bool easy_parsing,
                 int engine_id) :
_start {start},
_maxiter {maxiter},
_verbosity_level {vl},
_mode {mode},
_easy_parsing {easy_parsing},
_engine_id {engine_id}
{
    // read numerals for desired language from text file, init letter frequency table
    {
        ifstream file {numerals_file};
        string line;
        while (getline(file, line)) {
            _numerals.push_back(line);
        }
    }
    
    // setup random distribution
    _dist.param(uniform_int_distribution<>::param_type {0, min<int>(abs(maxseed), _numerals.size())});
    
    // now we know how many numerals we have, so we can allocate our efficient buffers
    int n;
    for (n = 0; n < _numerals.size(); ++n) {
        vector<int> vec;
        vec.resize(26, 0);
        _frequencies.push_back(vec);
        _freq.push_back(_frequencies.back().data());
    }
    _start_freq.resize(26, 0);
    _old.resize(26, 0);
    _new.resize(26, 0);
    _used.resize(26, 0);
    
    // init numeral letter frequency table
    for (n = 0; n < _numerals.size(); n++) {
        for (char c : _numerals[n]) {
            if (c >= 'a' && c <= 'z') {
                ++_freq[n][c - 'a'];
            }
        }
    }
    if (_verbosity_level > 1) {
        numeral_frequencies(cerr);
    }
    
    // get letter frequencies from sentence start
    for (char c : start) {
        // optionally translate upper- to lowercase
        const char up_lo_dif = 'a' - 'A';
        if (c >= 'A' && c <= 'Z') {
            c += up_lo_dif;
        }
        // skip non-alphabetical chars
        if (c >= 'a' && c <= 'z') {
            ++_start_freq[c - 'a'];
        }
    }
    // add letters from 'and', reset 'and' vector
    for (n = 0; n < 26; n++) {
        _start_freq[n] += _freq[0][n];
        _freq[0][n] = 0;
    }
    // build 'used' table for correct autogram seeding
    for (n = 0; n < 26; ++n) {
        _used[n] = _start_freq[n];
    }
    for (n = 0; n < _numerals.size(); ++n) {
        for (int i = 0; i < 26; ++i) {
            _used[i] += _freq[n][i];
        }
    }
}


void engine::run()
{
    auto prev = _old.data();
    auto next = _new.data();
    auto freq = _freq.data();
    auto start = _start_freq.data();

    do {
        // setup
        int num_iter = 0;
        unsigned int k, n;

        // create random seed vector
        switch (_mode) {
        case robinsonizer_mode::pangram:
        case robinsonizer_mode::strict_autogram:
            for (n = 0; n < 26; n++) {
                prev[n] = _dist(_random_engine);
            }
            break;
        case robinsonizer_mode::lax_autogram:
            for (n = 0; n < 26; ++n) {
                if (_used[n]) { // only if letter occurs in numerals or sentence start
                    prev[n] = _dist(_random_engine);
                } else {
                    prev[n] = 0;
                }
            }
            break;
        default:
            break;
        }

#if DEBUG
        // logging, if desired
        if (_verbosity_level > 1) {
            frequencies(cerr, const_cast<const int*>(prev));
        }
#endif

        do {
            // setup
            num_iter++;
            _total_iterations++;
            memcpy(next, start, 26 * sizeof(int));

            // count letters in resulting sentence by incrementing result freqmap elements
            for (n = 0; n < 26; n++) {
                if (static_cast<unsigned int>(prev[n]) < _numerals.size()) {
                    auto p = freq[prev[n]];
                    for (k = 0; k < 26; k++) {
                        next[k] += p[k];
                    }
                } else {
                    char c = 'a' + n;
                    if (_easy_parsing) {
                        cout << "OVFL[" << _engine_id << "] " << c << " (" << prev[n] << ")" << endl;
                    } else {
                        cerr << endl << "overflow: " << c << " (" << prev[n] << ")";
                    }
                    break;
                }
            }

            // increment frequency for 's' for every letter which occurs more than once,
            // and increment the count for every letter which is (or should be) mentioned
            for (n = 0; n < 26; n++) {
                switch (_mode) {
                case robinsonizer_mode::pangram:
                    ++next[n];
                    break;
                case robinsonizer_mode::strict_autogram:
                    if (next[n]) {
                        ++next[n];
                    }
                    break;
                case robinsonizer_mode::lax_autogram:
                    if (prev[n]) {
                        ++next[n];
                    }
                    break;
                default:
                    break;
                }
                if (next[n] > 1) {
                    ++next[s_index];
                }
            }

#if DEBUG
            // debug output, only if verbosity level is high enough
            if (_verbosity_level > 1) {
                write_result(cerr);
                cerr << endl;
            }
#endif

            // test if our result equals the previous one (if so, we found a valid sentence)
            _found = memcmp(next, prev, 26 * sizeof(int)) == 0;

            if (_found) {
                break;
            }
            if (num_iter == _maxiter) {
                break;
            } else {
                memcpy(prev, next, 26 * sizeof(int));
            }

        } while (true);

    } while (!_found);
}

void engine::frequencies(ostream& os, const int fm[]) const
{
    bool output = false;
    for (unsigned int n = 0; n < 26; n++) {
        if (fm[n]) {
            if (output) {
                os << ", ";
            }
            char c = n + 'a';
            os << c << " (" << fm[n] << ")";
            output = true;
        }
    }
    os << endl;
}

void engine::numeral_frequencies(ostream& os) const
{
    for (unsigned int i = 0; i < _numerals.size(); i++) {
        os << _numerals[i] << ": ";
        frequencies(os, _freq[i]);
    }
}

void engine::write_result(ostream& os) const
{
    if (_easy_parsing) {
        os << "RSLT[" << _engine_id << "] ";
    }
    os << _start << " ";
    unsigned int n;
    unsigned int first = 0;
    unsigned int last = 25;
    bool first_found = false;
    for (n = 0; n < 26; n++) { // pre-scan
        if (_new[n]) {
            if (!first_found) {
                first = n;
                first_found = true;
            }
            last = n;
        }
    }
    for (n = 0; n < 26; n++) {
        if (_new[n]) {
            if (n == last) {
                os << " " << _numerals[0] << " ";
            } else if (n > first) {
                os << ", ";
            }
            char c = n + 'a';
            os << _numerals[_new[n]] << " " << c;
            if (_new[n] > 1) {
                os << "'s";
            }
        }
    }
    os << ".";
}

ostream& operator<<(ostream& os, const engine& engine) {
    engine.write_result(os);
    return os;
}