// tdriver.cc

#include <cstdio>
#include <cstddef>
#include <cstdint>
#include <cstdlib>
#include <cstring>
#include <ctime>

#include <string>
#include <array>
#include <vector>

#include <utility>
#include <iterator>
#include <execution>
#include <algorithm>

// #include <boost/container_hash/hash.hpp>

#include <iostream>
#include <iomanip>
#include <fstream>
#include <sstream>

// ----------------------------------------------------------------------------

typedef int_fast64_t llil_int_type;

// All words in big1.txt, big2.txt, big3.txt are <= 6 chars in length.
// big.txt  max word length is 6
// long.txt max word length is 208
//
// Based on rough benchmarking, the short fixed string hack below is only
// worth trying for MAX_STR_LEN_L up to about 22.
// See also https://backlinko.com/google-keyword-study
//
// To use (limited length) fixed length strings uncomment the next line.
#define MAX_STR_LEN_L 6

#ifdef MAX_STR_LEN_L
struct str_type : std::array<char, MAX_STR_LEN_L> {
   bool operator==( const str_type& o ) const {
      return ::memcmp(this->data(), o.data(), MAX_STR_LEN_L) == 0;
   }
   bool operator<( const str_type& o ) const {
      return ::memcmp(this->data(), o.data(), MAX_STR_LEN_L) < 0;
   }
};
#else
struct str_type : std::basic_string<char> {
   bool operator==( const str_type& o ) const {
      return ::strcmp(this->data(), o.data()) == 0;
   }
   bool operator<( const str_type& o ) const {
      return ::strcmp(this->data(), o.data()) < 0;
   }
};
#endif

using str_int_type     = std::pair<str_type, llil_int_type>;
using vec_str_int_type = std::vector<str_int_type>;

// inject specialization of std::hash for str_type into namespace std
namespace std {
   template<> struct hash<str_type> {
      std::size_t operator()( str_type const& v ) const noexcept {
#if 0
         return boost::hash_range( v.cbegin(), v.cend() );
#else
         std::basic_string_view<char> bv {
            reinterpret_cast<const char*>(v.data()), v.size() * sizeof(char) };
         return std::hash<std::basic_string_view<char>>()(bv);
#endif
      }
   };
}

// Test with std::map, std::unordered_map or phmap::parallel_flat_hash_map
// or boost::unorderedmap or absl::flat_hash_map
#define MT_STD_MAP_L                 0
#define MT_STD_UNORDERED_MAP_L       1
#define MT_PARALLEL_FLAT_HASH_MAP_L  2
#define MT_BOOST_UNORDERED_MAP_L     3
#define MT_GOOGLE_HASH_MAP_L         4
#define MT_GOOGLE_NODE_MAP_L         5

// Uncomment to use absl::Hash function object
// #define MT_USE_ABSL_HASH_L    1

// Uncomment one of the map types below
// #define MAP_TYPE_L  MT_STD_MAP_L
// #define MAP_TYPE_L  MT_STD_UNORDERED_MAP_L
// #define MAP_TYPE_L  MT_PARALLEL_FLAT_HASH_MAP_L
// #define MAP_TYPE_L  MT_BOOST_UNORDERED_MAP_L
#define MAP_TYPE_L  MT_GOOGLE_HASH_MAP_L
// #define MAP_TYPE_L  MT_GOOGLE_NODE_MAP_L

#if MAP_TYPE_L == MT_STD_MAP_L
#include <map>
using map_str_int_type = std::map<str_type, llil_int_type>;
#elif MAP_TYPE_L == MT_STD_UNORDERED_MAP_L
#include <unordered_map>
using map_str_int_type = std::unordered_map<str_type, llil_int_type>;
#elif MAP_TYPE_L == MT_PARALLEL_FLAT_HASH_MAP_L
#include <parallel_hashmap/phmap.h>
// create the parallel_flat_hash_map without internal mutexes
using map_str_int_type = phmap::parallel_flat_hash_map<
   str_type, llil_int_type,
   phmap::priv::hash_default_hash<str_type>,
   phmap::priv::hash_default_eq<str_type>,
   phmap::priv::Allocator<phmap::priv::Pair<const str_type, llil_int_type>>,
   8, phmap::NullMutex
>;
#elif MAP_TYPE_L == MT_BOOST_UNORDERED_MAP_L
#include <boost/unordered_map.hpp>
using map_str_int_type = boost::unordered_map<str_type, llil_int_type>;
#elif MAP_TYPE_L == MT_GOOGLE_HASH_MAP_L
#include <absl/container/flat_hash_map.h>
#include <absl/container/node_hash_map.h>
#ifdef MT_USE_ABSL_HASH_L
using map_str_int_type = absl::flat_hash_map<str_type, llil_int_type, absl::Hash<str_type>>;
#else
using map_str_int_type = absl::flat_hash_map<str_type, llil_int_type>;
#endif
#elif MAP_TYPE_L == MT_GOOGLE_NODE_MAP_L
#include <absl/hash/hash.h>
using map_str_int_type = absl::node_hash_map<str_type, llil_int_type>;
#else
#error "Unsupported map_str_int_type"
#endif

// Simple RAII timer -----------------------------------------------------------
// Create a MyTimer object in a scope:
//    {
//       MyTimer tt;
//       ...
//    }
// to automatically print the time taken in the block to stderr

#include <chrono>

inline double elaspe_time(
   std::chrono::high_resolution_clock::time_point cend,
   std::chrono::high_resolution_clock::time_point cstart)
{
   return double( std::chrono::duration_cast<std::chrono::milliseconds>(cend - cstart).count() ) * 1e-3;
}

class MyTimer {
public:
   MyTimer()  { stnow_m = std::chrono::high_resolution_clock::now(); }
   ~MyTimer() {
      auto endnow = std::chrono::high_resolution_clock::now();
      std::cerr << " (" << elaspe_time(endnow, stnow_m) << " seconds)\n";
   }
private:
   std::chrono::time_point<std::chrono::high_resolution_clock> stnow_m;
};

// ---------------------------------------------------------------------

// #include "get_properties.inl"

// get_properties.inl
// Note: str_type, llil_int_type and map_str_int_type are not defined here.
// llil_int_type is normally defined as: typedef int_fast64_t llil_int_type;
// Note: int_fast64_t is defined in <cstdint>
// map_str_int_type is keyed by str_type with value of llil_int_type.
// These three types must be defined by the code that includes this .inl file.
// map_str_int_type can be many different types, std::map, std::unordered_map, ...
// so long as all operations on map_upd below are supported.

#include <cstdint>
#include <cstdio>
#include <array>
#include <algorithm>

inline llil_int_type fast_atoll64( const char* str )
{
   llil_int_type val  = 0;
   // int     sign = 0;
   // if ( *str == '-' ) {
   //    sign = 1, ++str;
   // }
   uint8_t digit;
   while ((digit = uint8_t(*str++ - '0')) <= 9) val = val * 10 + digit;
   // return sign ? -val : val;
   return val;
}

// Limit line length and use ANSI C functions to try to boost performance
#define MAX_LINE_LEN_L 255

// Update map_upd with the properties found in file fname
// Return the number of properties in file fname (or -1 if fname could not be opened)
static llil_int_type get_properties(
   const char*        fname,     // in   : the input file name
   map_str_int_type&  map_upd    // inout: the map to be updated
)
{
   std::array<char, MAX_LINE_LEN_L + 1> line;
   char* found;
   llil_int_type count;
   llil_int_type nprop = 0;
   FILE* fh = ::fopen(fname, "r");
   if ( fh == NULL ) return -1;
   while ( ::fgets( line.data(), static_cast<int>(MAX_LINE_LEN_L), fh ) != NULL ) {
      ++nprop;
      found = std::find( line.begin(), line.end(), '\t' );
      count = fast_atoll64( found + 1 );
      // Note: using emplace() is faster than map_upd[fixword] += count;
#ifdef MAX_STR_LEN_L
      str_type fixword {};  // {} initializes all elements of fixword to '\0'
      std::copy( line.begin(), found, fixword.begin() );
      const auto [it, success] = map_upd.emplace( fixword, count );
#else
      *found = '\0';
      const auto [it, success] = map_upd.emplace(str_type{ line.data() }, count);
#endif
      if (!success) it->second += count;
   }
   ::fclose(fh);
   return nprop;
}

// ---------------------------------------------------------------------

int main(int argc, char* argv[])
{
   if (argc < 2) {
      std::cerr << "usage: tdriver file1 file2 ... >out.txt\n";
      return 1;
   }

#ifdef MAX_STR_LEN_L
   std::cerr << "tdriver (fixed string length=" << MAX_STR_LEN_L << ") start\n";
#else
   std::cerr << "tdriver start\n";
#endif

#if MAP_TYPE_L == MT_STD_MAP_L
   std::cerr << "use std::map\n";
#elif MAP_TYPE_L == MT_STD_UNORDERED_MAP_L
   std::cerr << "use std::unordered_map\n";
#elif MAP_TYPE_L == MT_PARALLEL_FLAT_HASH_MAP_L
   std::cerr << "use phmap::parallel_flat_hash_map\n";
#elif MAP_TYPE_L == MT_BOOST_UNORDERED_MAP_L
   std::cerr << "use boost::unordered_map\n";
#elif MAP_TYPE_L == MT_GOOGLE_HASH_MAP_L
   std::cerr << "use absl::flat_hash_map\n";
#elif MAP_TYPE_L == MT_GOOGLE_NODE_MAP_L
   std::cerr << "use absl::node_hash_map\n";
#else
#error "Unsupported map_str_int_type"
#endif
#ifdef MT_USE_ABSL_HASH_L
   std::cerr << "use absl::Hash function object in ctor\n";
#endif

   // Get the list of input files from the command line
   int    nfiles = argc - 1;
   char** fname  = &argv[1];

   map_str_int_type mymap;
   for ( int i = 0; i < nfiles; ++i ) {
      MyTimer tt;
      llil_int_type nlines = get_properties(fname[i], mymap);
      std::cerr << fname[i] << ": nlines=" << nlines;
   }

   // Store the properties into a vector
   vec_str_int_type propvec( mymap.begin(), mymap.end() );
   mymap.clear();

   // Sort the vector by (count) in reverse order, (name) in lexical order
   std::sort(
      // Standard sort
      propvec.begin(), propvec.end(),
      [](const str_int_type& left, const str_int_type& right) {
         return left.second != right.second
            ? left.second > right.second
            : left.first  < right.first;
      }
   );

   // Output the sorted vector
#ifdef MAX_STR_LEN_L
   for ( auto const& n : propvec )
      ::printf("%.*s\t%ld\n", MAX_STR_LEN_L, n.first.data(), n.second);
#else
   for ( auto const& n : propvec )
      std::cout << n.first << "\t" << n.second << "\n";
#endif

   return 0;
}