dalla-data-processing 0.0.1__py3-none-any.whl

dalla/deduplication/onion/src_sc/onion.c

@@ -0,0 +1,854 @@
+/*********************************************************************
+ * Copyright (c) 2011-2020 Jan Pomikalek, Milos Jakubicek            *
+ * All rights reserved.                                              *
+ *                                                                   *
+ * This software is licensed as described in the file COPYING, which *
+ * you should have received as part of this distribution.            *
+ *********************************************************************
+ *                                                                   *
+ * Refactored in 2023 by Sally Choker and Fadi Zaraket.              *
+ * Modifications:                                                    *
+ * - Adapted to take as input a list of file paths instead of        *
+ *   a single file.                                                  *
+ * - Enhanced to detect the source of duplicate files                *
+ * - Calculate the score of duplication from each duplicate file     *
+ *********************************************************************/
+
+#include "buzhash.h"
+#include "version.h"
+#include <algorithm>
+#include <errno.h>
+#include <fstream>
+#include <iostream>
+#include <sstream>
+#include <stdio.h>
+#include <string.h>
+#include <sys/resource.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <time.h>
+#include <unistd.h>
+#include <unordered_map>
+#include <vector>
+
+#define GOOGLE_SPARSE
+
+#if defined GOOGLE_SPARSE
+#include <sparsehash/sparse_hash_map>
+// #include <sparsehash/sparse_hash_set>
+// using google::sparse_hash_set;
+#elif defined __GNUC__ || defined __APPLE__
+#include <ext/hash_map>
+namespace std { using namespace __gnu_cxx; }
+#else
+#include <hash_map>
+#endif
+using namespace std;
+
+#ifdef GOOGLE_SPARSE
+// typedef sparse_hash_set<uint64_t> ngrhash;
+typedef google::sparse_hash_map<uint64_t, int> ngrhash;
+#else
+// typedef hash_map<uint64_t,bool> ngrhash;
+typedef hash_map<uint64_t, int> ngrhash;
+#endif
+
+// initialize vector to store pair filenames and duplicates of all files
+std::vector<std::pair<string, int>> foundFilenames;
+
+// initailize vector to store only pairs of duplicated filenames
+std::vector<std::pair<string, int>> DuplicateFilenames;
+// initialize a vector to store documents not checked for deduplication (don't
+// fit in the buffer)
+std::vector<int> UncheckedFilenames;
+
+// initialize a vector to stroe duplicate filenames and their score:
+// bad_tokens/total_tokens
+std::vector<std::pair<string, float>> DuplicateFilenamesScores;
+
+#define BITMASK_HIGH63 0xfffffffffffffffeul
+
+#define NGRAM_SIZE 10
+#define DUPL_THRES 0.9
+#define DOC_TAG "doc"
+#define PAR_TAG "p"
+#define TRIM_HASHES 64
+#define MAX_STUB_LENGTH 20
+#define BUFFER_SIZE 335544320
+
+// options
+int Ngram_size = NGRAM_SIZE;
+float Dupl_thres = DUPL_THRES;
+const char *Doc_tag = DOC_TAG;
+const char *Par_tag = PAR_TAG;
+int Strip_dupl = 0;
+int No_smoothing = 0;
+int Trim_hashes = TRIM_HASHES;
+int Max_stub_length = MAX_STUB_LENGTH;
+long Buffer_size = BUFFER_SIZE;
+const char *output_dir = NULL;
+int Quiet = 0;
+char *Dupl_hashes_path = NULL;
+FILE *Input;
+long int Input_size;
+char current_file_name[2048];
+int fileNameIndex = 0;
+
+void print_usage(FILE *stream) {
+  fprintf(stream, "\
+Usage: onion [OPTIONS] [FILE]\n\
+Mark duplicate text parts in the input vertical file.\n\
+\n\
+ -f FILE   hashes of duplicate n-grams\n\
+ -n NUM    n-gram length (default: %i)\n\
+ -t NUM    duplicate content threshold (default: %.1f)\n\
+ -d STR    document tag (default: %s)\n\
+ -p STR    paragraph tag (default: %s)\n\
+ -s        strip duplicate parts (rather than mark)\n\
+ -m        no smoothing\n\
+ -T NUM    trim n-gram hashes to NUM bits (default: %i)\n\
+ -l NUM    max stub length (default: %i)\n\
+ -b NUM    buffer size, in bytes (default: %i)\n\
+ -q        quiet; suppress all output except for errors\n\
+\n\
+ -V        print version information and exit\n\
+ -h        display this help and exit\n\
+\n\
+With no FILE, or when FILE is -, read standard input.\n\
+Output is written to standard output.\n\
+\n\
+Project home page: <http://code.google.com/p/onion/>\n",
+          NGRAM_SIZE, DUPL_THRES, DOC_TAG, PAR_TAG, TRIM_HASHES,
+          MAX_STUB_LENGTH, BUFFER_SIZE);
+}
+
+void print_progress(const char *task_descr, unsigned long int processed_bytes,
+                    float percent_done) {
+  time_t now;
+  struct rusage usage;
+  time(&now);
+  getrusage(RUSAGE_SELF, &usage);
+  fprintf(stderr, "[%.24s] onion: %s: %6li MB processed", ctime(&now),
+          task_descr, processed_bytes / (1024 * 1024));
+  if (percent_done >= 0)
+    fprintf(stderr, " (%6.2f%%)", percent_done);
+  fprintf(stderr, "\t%6li MB RAM used", usage.ru_maxrss / 1024);
+  fprintf(stderr, "\n");
+}
+
+void saveGlobalHashmap(const ngrhash &global, const std::string &filename) {
+  std::ofstream outFile(filename.c_str(), std::ios::binary | std::ios::app);
+  if (!outFile.is_open()) {
+    throw std::runtime_error("Failed to open file for writing");
+  }
+
+  for (const auto &pair : global) {
+    outFile.write(reinterpret_cast<const char *>(&pair.first),
+                  sizeof(pair.first));
+    outFile.write(reinterpret_cast<const char *>(&pair.second),
+                  sizeof(pair.second));
+  }
+
+  outFile.close();
+}
+
+bool fileExists(const std::string &name) {
+  struct stat map;
+  return (stat(name.c_str(), &map) == 0);
+}
+
+void writeFilenameDuplicateToCSV(
+    const std::vector<std::pair<std::string, int>> &vec,
+    const std::string &fullPath) {
+  std::ifstream infile(fullPath);
+  bool fileExists = infile.is_open();
+  infile.close();
+
+  std::ofstream file(fullPath, fileExists ? std::ios::app : std::ios::out);
+  if (!file.is_open()) {
+    std::cerr << "Failed to open the file: " << fullPath << std::endl;
+    return;
+  }
+
+  for (const auto &pair : vec) {
+    file << "\"" << pair.first << "\"," << pair.second << "\n";
+  }
+
+  file.close();
+}
+
+void writeFilenameDuplicateScoreToCSV(
+    const std::vector<std::pair<std::string, float>> &vec,
+    const std::string &fullPath) {
+
+  std::ifstream infile(fullPath);
+  bool fileExists = infile.is_open();
+  infile.close();
+
+  std::ofstream file(fullPath, fileExists ? std::ios::app : std::ios::out);
+
+  if (!file.is_open()) {
+    std::cerr << "Failed to open the file: " << fullPath << std::endl;
+    return;
+  }
+
+  for (const auto &pair : vec) {
+    file << pair.first << "," << pair.second << "\n";
+  }
+}
+
+void writeUncheckedFilenamesToCSV(const std::vector<int> &vec,
+                                  const std::string &fullPath) {
+  std::ifstream infile(fullPath);
+  bool fileExists = infile.is_open();
+  infile.close();
+
+  std::ofstream file(fullPath, fileExists ? std::ios::app : std::ios::out);
+  if (!file.is_open()) {
+    std::cerr << "Failed to open the file: " << fullPath << std::endl;
+    return;
+  }
+
+  for (const auto &num : vec) {
+    file << num << "\n";
+  }
+
+  file.close();
+}
+
+int process_one_par(int *&pars, char **&tokens, buzhash_buffer_t &bh_buffer,
+                    hash_t &hash_bitmask, ngrhash &local, ngrhash &global,
+                    int &bad_tokens, int &tok_i, int &par_i, int &total_tokens,
+                    int &prev_bad_tokens, int &have_dupl_ngrams) {
+  ngrhash::const_iterator it;
+  for (tok_i = pars[par_i]; tok_i < pars[par_i + 1]; tok_i++) {
+    char *token = tokens[tok_i];
+    if (token[0] == '<')
+      continue;
+    total_tokens++;
+    // prev_bad_tokens--;
+    // if (prev_bad_tokens < 0)
+    // prev_bad_tokens = 0;
+    hash_t hash = buzhash(token, &bh_buffer);
+    hash_t masked_hash = hash & hash_bitmask;
+    if (!buzhash_is_full_buffer(&bh_buffer))
+      continue;
+    it = local.find(hash);
+    if (it == local.end()) {
+      if (have_dupl_ngrams) {
+        // test with the last bit set to 1
+        // // (check against already seen duplicate ngrams)
+        it = global.find(masked_hash | 1);
+      } else {
+        it = global.find(masked_hash);
+      }
+    }
+    if (it != global.end()) {
+      bad_tokens++; // bad_tokens += Ngram_size - prev_bad_tokens;
+      // prev_bad_tokens = Ngram_size;
+      // if (fileNameIndex != it->second) {
+      foundFilenames.push_back(
+          make_pair(current_file_name, it->second)); // Save the filename.
+                                                     //}
+    }
+#ifdef GOOGLE_SPARSE
+    local.insert(std::make_pair(hash, fileNameIndex)); // local.insert(hash);
+#else
+    local[hash] = fileNameIndex; // local[hash] = true;
+#endif
+  }
+}
+
+int process_one_file(int &buffer_content, char **&tokens, int *&pars,
+                     int *&par_len, char *&bad_par,
+
+                     int *docs,
+
+                     char *doc_tag, int doc_tag_len, char *doc_end_tag,
+                     int doc_end_tag_len, char *par_tag, int par_tag_len,
+                     char *par_end_tag, int par_end_tag_len,
+
+                     int &have_dupl_ngrams,
+                     unsigned long int &total_processed_bytes,
+
+                     // make sure these become mutable
+                     ngrhash &global, ngrhash &local,
+                     buzhash_buffer_t &bh_buffer, char *buffer) {
+  int bytes_read = fread(buffer + buffer_content, sizeof(char),
+                         Buffer_size - buffer_content, Input);
+  // print("Buffer %x, BufferContent %d, BufferSize %ld, \n", buffer,
+  // buffer_content, Buffer_size);
+  hash_t hash_bitmask = 0xfffffffffffffffful;
+
+  int buffer_size = buffer_content + bytes_read;
+  buffer[buffer_size] = '\0'; // make it a string
+  char *buffer_pos = buffer;
+
+  // find tokens
+  int token_count = 0;
+  tokens[token_count++] = buffer_pos++;
+  while ((buffer_pos = strchr(buffer_pos, '\n')) != NULL) {
+    buffer_pos[0] = '\0';
+    tokens[token_count++] = ++buffer_pos;
+  }
+
+  // find docs and paragraphs
+  int doc_count = 0;
+  int par_count = 0;
+  docs[doc_count++] = 0;
+  pars[par_count++] = 0;
+  int start_doc_next = 0;
+  int start_par_next = 0;
+  int i;
+
+  for (i = 1; i < token_count; i++) {
+    // "<doc>" or "<doc "
+    if (start_doc_next ||
+        (strncmp(tokens[i], doc_tag, doc_tag_len) == 0 &&
+         (tokens[i][doc_tag_len] == ' ' || tokens[i][doc_tag_len] == '>'))) {
+      docs[doc_count++] = par_count;
+      pars[par_count++] = i;
+      start_doc_next = 0;
+    }
+    // "</doc>"
+    else if (strncmp(tokens[i], doc_end_tag, doc_end_tag_len) == 0) {
+      start_doc_next = 1;
+    }
+    // "<p>" or "<p "
+    else if (start_par_next || (strncmp(tokens[i], par_tag, par_tag_len) == 0 &&
+                                (tokens[i][par_tag_len] == ' ' ||
+                                 tokens[i][par_tag_len] == '>'))) {
+      pars[par_count++] = i;
+      start_par_next = 0;
+    }
+    // "</p>"
+    else if (strncmp(tokens[i], par_end_tag, par_end_tag_len) == 0) {
+      start_par_next = 1;
+    }
+  }
+
+  if (doc_count == 1 && !feof(Input)) {
+    // full buffer contains only one document
+    // by default, documents with higher than 16MB are not checked for dedup
+    fprintf(stderr, "Too long document at byte %li.\n", total_processed_bytes);
+    UncheckedFilenames.push_back(fileNameIndex);
+    return 1;
+  }
+
+  if (feof(Input)) {
+    // create sentinels
+    docs[doc_count++] = par_count;
+    if (strlen(tokens[token_count - 1]) == 0) {
+      // files ending with a newline have a natural sentinel
+      // (the last zero-length token)
+      pars[par_count++] = token_count - 1;
+    } else {
+      // for other files, we need to make up the last token
+      pars[par_count++] = token_count;
+      tokens[token_count++] = buffer + buffer_size;
+    }
+  }
+
+  // for all documents
+  int doc_i;
+  for (doc_i = 0; doc_i < doc_count - 1; doc_i++) {
+    buzhash_clear_buffer(&bh_buffer);
+    local.clear();
+    // for all paragraphs in the document
+    int par_i;
+    for (par_i = docs[doc_i]; par_i < docs[doc_i + 1]; par_i++) {
+      int total_tokens = 0;
+      int bad_tokens = 0;
+      /* prev_bad_tokens is the number of tokens in the current
+       * n-gram which are contained in one of the previous bad
+       * n-grams.
+       *
+       * At the beginning of a new paragraph we need to pretend that
+       * there are Ngram_size prev_bad_tokens so that the leading
+       * bad n-grams do not generate too many bod tokens. */
+      int prev_bad_tokens = Ngram_size;
+      // for all tokens in the paragraph
+      // for every token in the current paragraph, it computes its hash and
+      // checks if it is present in the local and global hash maps
+      int tok_i;
+      float score;
+      for (tok_i = pars[par_i]; tok_i < pars[par_i + 1]; tok_i++) {
+        process_one_par(pars, tokens, bh_buffer, hash_bitmask, local, global,
+                        bad_tokens, tok_i, par_i, total_tokens, prev_bad_tokens,
+                        have_dupl_ngrams);
+      }
+
+      // remember the length of the paragraph
+      par_len[par_i] = total_tokens;
+
+      // mark bad paragraphs
+      bad_par[par_i] =
+          (total_tokens > 0 && (1.0 * bad_tokens / total_tokens) > Dupl_thres);
+
+      // get score for each file
+      score = (1.0 * bad_tokens / total_tokens);
+      if (score > Dupl_thres) {
+        DuplicateFilenamesScores.push_back(make_pair(current_file_name, score));
+      }
+      // DuplicateFilenamesScores.push_back(make_pair(fileNameIndex, score));
+    }
+
+    // smoothing
+    if (!No_smoothing) {
+      int last_bad_par = docs[doc_i] - 1;
+      int stub_length = 0;
+      for (par_i = docs[doc_i]; par_i <= docs[doc_i + 1]; par_i++) {
+        if (par_i == docs[doc_i + 1] || bad_par[par_i]) {
+          if (stub_length <= Max_stub_length) {
+            // remove stub
+            int par_j;
+            for (par_j = last_bad_par + 1; par_j < par_i; par_j++)
+              bad_par[par_j] = 1;
+          }
+          last_bad_par = par_i;
+          stub_length = 0;
+        } else {
+          stub_length += par_len[par_i];
+        }
+      }
+    }
+
+    int count_bad = 0;
+    for (par_i = docs[doc_i]; par_i < docs[doc_i + 1]; par_i++) {
+      if (!bad_par[par_i] && par_len[par_i] > 0) {
+      } else {
+        count_bad += 1;
+      }
+    }
+    if (count_bad > 0) {
+      std::sort(foundFilenames.begin(), foundFilenames.end());
+      // auto uniqueEnd = std::unique(foundFilenames.begin(),
+      // foundFilenames.end()); foundFilenames.erase(uniqueEnd,
+      // foundFilenames.end());
+      printf("%s is %d bad\n", current_file_name, count_bad);
+      DuplicateFilenames.insert(DuplicateFilenames.end(),
+                                foundFilenames.begin(), foundFilenames.end());
+      // DuplicateFilenames.push_back(current_file_name);
+    }
+
+    foundFilenames.clear();
+
+    // is there at least one good paragraph?
+    int all_bad = 1;
+    for (par_i = docs[doc_i]; par_i < docs[doc_i + 1]; par_i++) {
+      if (!bad_par[par_i] && par_len[par_i] > 0) {
+        all_bad = 0;
+        break;
+      }
+    }
+
+    buzhash_clear_buffer(&bh_buffer);
+    // for all paragraphs in the document (again)
+    for (par_i = docs[doc_i]; par_i < docs[doc_i + 1]; par_i++) {
+      int first_token = pars[docs[doc_i]];
+      int last_token = pars[docs[doc_i + 1]] - 1;
+      // for all tokens in the paragraph
+      int tok_i;
+      for (tok_i = pars[par_i]; tok_i < pars[par_i + 1]; tok_i++) {
+        char *token = tokens[tok_i];
+        int bad_token = bad_par[par_i];
+        // tags at document boundaries are preserved if there is
+        // at least one good paragraph, removed otherwise
+        if (tok_i == first_token && strncmp(token, doc_tag, doc_tag_len) == 0)
+          bad_token = all_bad;
+        if (tok_i == last_token && strcmp(token, doc_end_tag) == 0)
+          bad_token = all_bad;
+        // print output
+        // if (Strip_dupl) {
+        // if (!bad_token)
+        // printf("%s\n", token);
+        //}
+        // else {
+        // printf("%i\t%s\n", bad_token, token);
+        //}
+        if (token[0] == '<')
+          continue;
+        // store hashes of n-grams
+        hash_t hash = buzhash(token, &bh_buffer);
+        hash_t masked_hash = hash & hash_bitmask;
+        if (!buzhash_is_full_buffer(&bh_buffer))
+          continue;
+        if (!bad_par[par_i]) {
+          if (have_dupl_ngrams) {
+            // If we have the list of hashes of all duplicate
+            // n-grams, we set the least significant bit of the
+            // stored hash to 1 if we have seen the matching
+            // duplicate n-gram to indicate it has been seen.
+            // Unique n-grams are ignored.
+            if (global.erase(masked_hash & BITMASK_HIGH63))
+#ifdef GOOGLE_SPARSE
+              global.insert(std::make_pair(masked_hash | 1, fileNameIndex));
+            // global.insert (masked_hash | 1);
+#else
+              global[masked_hash | 1] =
+                  fileNameIndex; // global[masked_hash | 1] = true;
+#endif
+          } else {
+            // otherwise we have to store hashes of all n-grams
+#ifdef GOOGLE_SPARSE
+            global.insert(std::make_pair(
+                masked_hash, fileNameIndex)); // global.insert(masked_hash);
+#else
+            global[masked_hash] = fileNameIndex; // global[masked_hash] = true;
+#endif
+          }
+        }
+      }
+    }
+  }
+
+  // copy the unprocessed data to the beginning of the buffer
+  if (!feof(Input)) {
+    char *last_doc_fst_tok = tokens[pars[docs[doc_count - 1]]];
+    int processed_bytes = last_doc_fst_tok - buffer;
+    total_processed_bytes += processed_bytes;
+    int remaining_bytes = buffer_size - processed_bytes;
+    char *buffer_end = last_doc_fst_tok + remaining_bytes;
+    char *pos;
+    // replace \0s with EOLs (revert the buffer contents to original)
+    for (pos = last_doc_fst_tok; pos < buffer_end; pos++)
+      if (pos[0] == '\0')
+        pos[0] = '\n';
+    // print progress information
+    if (!Quiet) {
+      float percent_done = -1;
+      if (Input_size > 0)
+        percent_done = 100.0 * total_processed_bytes / Input_size;
+      print_progress("removing duplicates", total_processed_bytes,
+                     percent_done);
+    }
+    memmove(buffer, last_doc_fst_tok, remaining_bytes);
+    buffer_content = remaining_bytes;
+  }
+  printf("end of process_one_file\n");
+}
+
+int main(int argc, char **argv) {
+  printf("Dupl Threshold %f \n", Dupl_thres);
+  printf("N-gram size %d \n", Ngram_size);
+  // get options
+  int c;
+  char *endptr;
+  char *datasetname = NULL;
+  char *listOfFilesPath = NULL;
+
+  while ((c = getopt(argc, argv, "f:n:t:d:p:smT:l:b:qVhD:L:O:")) != -1) {
+    errno = 0;
+    switch (c) {
+    case 'f':
+      Dupl_hashes_path = optarg;
+      break;
+    case 'n':
+      Ngram_size = strtol(optarg, &endptr, 10);
+      if (errno != 0 || *endptr != '\0') {
+        fprintf(stderr, "Integer value expected for -n, got: %s\n", optarg);
+        print_usage(stderr);
+        return 1;
+      }
+      break;
+    case 't':
+      Dupl_thres = strtod(optarg, &endptr);
+      if (errno != 0 || *endptr != '\0') {
+        fprintf(stderr, "Float value expected for -t, got: %s\n", optarg);
+        print_usage(stderr);
+        return 1;
+      }
+      break;
+    case 'd':
+      Doc_tag = optarg;
+      break;
+    case 'p':
+      Par_tag = optarg;
+      break;
+    case 's':
+      Strip_dupl = 1;
+      break;
+    case 'm':
+      No_smoothing = 1;
+      break;
+    case 'T':
+      Trim_hashes = strtol(optarg, &endptr, 10);
+      if (errno != 0 || *endptr != '\0') {
+        fprintf(stderr, "Integer value expected for -T, got: %s\n", optarg);
+        print_usage(stderr);
+        return 1;
+      }
+      break;
+    case 'l':
+      Max_stub_length = strtol(optarg, &endptr, 10);
+      if (errno != 0 || *endptr != '\0') {
+        fprintf(stderr, "Integer value expected for -l, got: %s\n", optarg);
+        print_usage(stderr);
+        return 1;
+      }
+      break;
+    case 'b':
+      Buffer_size = strtol(optarg, &endptr, 10);
+      if (errno != 0 || *endptr != '\0') {
+        fprintf(stderr, "Integer value expected for -b, got: %s\n", optarg);
+        print_usage(stderr);
+        return 1;
+      }
+      break;
+    case 'q':
+      Quiet = 1;
+      break;
+    case 'V':
+      print_version("onion");
+      return 0;
+    case 'h':
+      print_usage(stdout);
+      return 0;
+    case '?':
+      print_usage(stderr);
+      return 1;
+    case 'D': // Dataset name
+      datasetname = optarg;
+      break;
+    case 'L': // List of files path
+      listOfFilesPath = optarg;
+      break;
+    case 'O': // Output directory
+      output_dir = optarg;
+      break;
+    }
+  }
+
+  Input = stdin;
+  Input_size = -1;
+  if (optind < argc) {
+    char *filename = argv[optind];
+    if (strcmp(filename, "-") != 0) {
+      errno = 0;
+      Input = fopen(filename, "r");
+      if (errno != 0) {
+        fprintf(stderr, "Unable to open %s for reading.\n", filename);
+        return 1;
+      }
+      fseek(Input, 0L, SEEK_END);
+      Input_size = ftell(Input);
+      fseek(Input, 0L, SEEK_SET);
+    }
+  }
+
+  if (output_dir != NULL) {
+    struct stat st = {0};
+    if (stat(output_dir, &st) == -1) {
+      if (mkdir(output_dir, 0700) != 0) {
+        perror("Unable to create output directory");
+        return 1;
+      }
+    }
+  } else {
+    fprintf(stderr, "You must specify an output directory.\n");
+    print_usage(stderr);
+    return 1;
+  }
+
+  if (datasetname == NULL || listOfFilesPath == NULL) {
+    fprintf(stderr, "You must specify dataset name and list of files path.\n");
+    print_usage(stderr);
+    return 1;
+  }
+
+  FILE *List_of_Files = fopen(listOfFilesPath, "r");
+  if (List_of_Files == NULL) {
+    perror("Unable to open list of files");
+    return 1;
+  }
+
+  unsigned long int total_processed_bytes = 0;
+
+  // patterns
+  char *doc_tag = (char *)malloc((strlen(Doc_tag) + 1 + 1) * sizeof(char));
+  strcat(strcpy(doc_tag, "<"), Doc_tag);
+  char *doc_end_tag = (char *)malloc((strlen(Doc_tag) + 3 + 1) * sizeof(char));
+  strcat(strcat(strcpy(doc_end_tag, "</"), Doc_tag), ">");
+  char *par_tag = (char *)malloc((strlen(Par_tag) + 1 + 1) * sizeof(char));
+  strcat(strcpy(par_tag, "<"), Par_tag);
+  char *par_end_tag = (char *)malloc((strlen(Par_tag) + 3 + 1) * sizeof(char));
+  strcat(strcat(strcpy(par_end_tag, "</"), Par_tag), ">");
+
+  int doc_tag_len = strlen(doc_tag);
+  int doc_end_tag_len = strlen(doc_end_tag);
+  int par_tag_len = strlen(par_tag);
+  int par_end_tag_len = strlen(par_end_tag);
+
+  // bitmask for trimming ngram hashes
+  hash_t hash_bitmask = 0xfffffffffffffffful;
+  int bitshift = 64 - Trim_hashes;
+  if (bitshift > 0)
+    hash_bitmask >>= bitshift;
+
+  // data structures
+  int buffer_size = 0;
+  int buffer_content = 0;
+  char *buffer = (char *)malloc((Buffer_size + 1) * sizeof(char));
+  char **tokens = (char **)malloc((Buffer_size + 1) * sizeof(char *));
+  int *pars = (int *)malloc((Buffer_size + 1) *
+                            sizeof(int)); // array of starting tokens
+  int *par_len = (int *)malloc((Buffer_size + 1) * sizeof(int));
+  char *bad_par = (char *)malloc((Buffer_size + 1) * sizeof(char));
+  int *docs =
+      (int *)malloc((Buffer_size + 1) * sizeof(int)); // array of starting pars
+  int token_count, par_count, doc_count;
+
+  // buzhash
+  buzhash_buffer_t bh_buffer;
+  buzhash_init_buffer(&bh_buffer, Ngram_size);
+
+  // global hash table stores the hashes of all files read so far (in our case
+  // the full document since we did not mark paragraph separators) local hash
+  // table stores the hahses of n-grams found within the currently processed
+  // file (doc as a whole in our case)
+  ngrhash global, local;
+#ifdef GOOGLE_SPARSE
+  global.set_deleted_key(0);
+  local.set_deleted_key(0);
+#endif
+
+  // read hashes of duplicate n-grams if available
+  int have_dupl_ngrams = 0;
+  if (Dupl_hashes_path != NULL) {
+    have_dupl_ngrams = 1;
+    errno = 0;
+    FILE *ngrams_fp = fopen(Dupl_hashes_path, "r");
+    if (errno != 0) {
+      fprintf(stderr, "Unable to open %s for reading.\n", Dupl_hashes_path);
+      return 1;
+    }
+    fseek(ngrams_fp, 0L, SEEK_END);
+    unsigned long int ngrams_size = ftell(ngrams_fp);
+    fseek(ngrams_fp, 0L, SEEK_SET);
+
+    unsigned long int bytes_read = 0;
+    hash_t hash;
+    while (fread(&hash, sizeof(hash), 1, ngrams_fp)) {
+      printf("reading");
+      bytes_read += sizeof(hash);
+      hash_t masked_hash = hash & hash_bitmask;
+      // store only the 63 most significant bits of the hash;
+      // reserve the last bit as a flag (seen / unseen)
+      // #ifdef GOOGLE_SPARSE
+      // global.insert(std::make_pair(masked_hash | 1,
+      // fileNameIndex));//global.insert (masked_hash & BITMASK_HIGH63); #else
+      // global[masked_hash & BITMASK_HIGH63] = true;
+      // #endif
+
+      // print progress information
+      if (!Quiet && bytes_read % (10000000 * sizeof(hash)) == 0) {
+        float percent_done = -1;
+        if (ngrams_size > 0)
+          percent_done = 100.0 * bytes_read / ngrams_size;
+        print_progress("reading hashes", bytes_read, percent_done);
+      }
+    }
+
+    if (!Quiet)
+      print_progress("reading hashes", bytes_read, 100);
+  }
+
+  int fileIndex =
+      0; // File index t store every 300 duplicates in a new csv file
+
+  while (!feof(List_of_Files)) {
+    fileNameIndex++;
+    /* read the name of each file in the list of files*/
+    if (fgets(current_file_name, sizeof(current_file_name), List_of_Files) ==
+        NULL) {
+      break;
+    }
+    int last_ch = current_file_name[strlen(current_file_name) - 1];
+    if (last_ch == '\n') {
+      current_file_name[strlen(current_file_name) - 1] = '\0';
+    }
+    /* use fopen to assign Input to it*/
+    Input = fopen(current_file_name, "r");
+    if (Input == NULL) {
+      char err_msg[1024];
+      snprintf(err_msg, 1023, "Error to open data file %s--",
+               current_file_name);
+      perror(err_msg);
+      fprintf(stderr, "Skipping file: %s\n", current_file_name);
+      continue;
+    }
+
+    printf("File Number: %d \n", fileNameIndex);
+
+    /* Process it as below*/
+    // it modifies the data structures passed to it as arguments, processes the
+    // input file, and updates global state
+    process_one_file(buffer_content, /*   int */
+                     tokens,         /*   char** */
+                     pars,           /*   int* */
+                     par_len,        /*   int* */
+                     bad_par,        /*   char* */
+
+                     docs, /*   int* */
+
+                     doc_tag,         /*   char* */
+                     doc_tag_len,     /*   int */
+                     doc_end_tag,     /*   char* */
+                     doc_end_tag_len, /*   int */
+                     par_tag,         /*   char* */
+                     par_tag_len,     /*   int */
+                     par_end_tag,     /*   char* */
+                     par_end_tag_len, /*   int */
+
+                     have_dupl_ngrams,      /*   long int int */
+                     total_processed_bytes, /*   unsigned */
+                     // make sure these become mutable
+                     global,    /*   ngrhash */
+                     local,     /*   ngrhash */
+                     bh_buffer, /*   buzhash_buffer_t */
+                     buffer     /*   char* */
+    );
+
+    if (DuplicateFilenames.size() > 3000) {
+      fileIndex++;
+      std::string filename =
+          std::string(output_dir) + "/" + "Duplicate_pair_files_" +
+          std::string(datasetname) + "_" + to_string(fileIndex) + ".csv";
+
+      writeFilenameDuplicateToCSV(DuplicateFilenames, filename);
+
+      std::string filename_scores =
+          std::string(output_dir) + "/" + "Duplicate_scores_" +
+          std::string(datasetname) + "_" + to_string(fileIndex) + ".csv";
+
+      writeFilenameDuplicateScoreToCSV(DuplicateFilenamesScores,
+                                       filename_scores);
+      DuplicateFilenames.clear();
+      DuplicateFilenamesScores.clear();
+      printf(" 3000 Duplicated files are saved\n");
+    }
+
+    fclose(Input);
+  }
+
+  // write the remaining duplicate files
+  fileIndex++;
+  std::string filename = std::string(output_dir) + "/" +
+                         "Duplicate_pair_files_" + std::string(datasetname) +
+                         "_" + to_string(fileIndex) + ".csv";
+  writeFilenameDuplicateToCSV(DuplicateFilenames, filename);
+  std::string name = std::string(output_dir) + "/" + "Unchecked_files_" +
+                     std::string(datasetname) + "_" + ".csv";
+  writeUncheckedFilenamesToCSV(UncheckedFilenames, name);
+
+  // print progress information
+  total_processed_bytes += buffer_size;
+  if (!Quiet)
+    print_progress("removing duplicates", total_processed_bytes, 100);
+
+  //    save the global hash map
+  // saveGlobalHashmap(global);  // This will append to the existing file or
+  // create a new one if not present
+
+  return 0;
+}