[MISC] Remove merge sketches again.

seqan · Oct 18, 2023 · c02fece · c02fece
1 parent 71e23c0
commit c02fece
Show file tree

Hide file tree

Showing 3 changed files with 6 additions and 33 deletions.
diff --git a/src/display_layout/general.cpp b/src/display_layout/general.cpp
@@ -48,17 +48,6 @@ void keep_duplicates(robin_hood::unordered_set<uint64_t> & shared, std::vector<u
     shared = std::move(result);
 }
 
-size_t merge_sketches_and_estimate(std::vector<seqan::hibf::sketch::hyperloglog> const & sketches)
-{
-    assert(sketches.size() > 0);
-    seqan::hibf::sketch::hyperloglog temp_hll = sketches[0];
-
-    for (size_t j = 1; j < sketches.size(); ++j)
-        temp_hll.merge(sketches[j]);
-
-    return temp_hll.estimate();
-}
-
 int execute(config const & cfg)
 {
     std::ifstream layout_file{cfg.input};
@@ -103,7 +92,6 @@ int execute(config const & cfg)
         });
 
     seqan::hibf::sketch::hyperloglog sketch{hibf_config.sketch_bits}; // store one sketch computeted by iterative add
-    std::vector<seqan::hibf::sketch::hyperloglog> sketches; // store a sketch for each user bin to merge afterwards
 
     robin_hood::unordered_set<uint64_t> current_kmer_set{};
     robin_hood::unordered_set<uint64_t> shared_kmers{};
@@ -137,30 +125,27 @@ int execute(config const & cfg)
     output_stream << "# Layout: " << cfg.input.c_str() << '\n' //
                   << "tb_index\t"
                   << "size\t"
+                  << "estimated_size\t"
                   << "shared_size\t"
                   << "ub_count\t"
                   << "kind\t"
-                  << "splits\t"
-                  << "estimated_size_single_sketch\t"
-                  << "estimated_size_merged_sketch\n";
+                  << "splits\n";
 
     auto print_result_line = [&]()
     {
         bool const is_merged{bin_kinds[current_idx] == chopper::layout::hibf_statistics::bin_kind::merged};
         size_t const avg_kmer_count = (current_kmer_set.size() + split_count - 1u) / split_count;
-        size_t const one_sketch_estimate = (sketch.estimate() + split_count - 1u) / split_count;
-        size_t const merge_estimate = merge_sketches_and_estimate(sketches);
+        size_t const sketch_estimate = (sketch.estimate() + split_count - 1u) / split_count;
 
         for (size_t i{}, total{split_count}; i < total; ++i)
         {
             output_stream << current_idx + i << '\t'                  //
                           << avg_kmer_count << '\t'                   //
+                          << sketch_estimate << '\t'                  //
                           << shared_kmers.size() << '\t'              //
                           << ub_count << '\t'                         //
                           << (is_merged ? "merged" : "split") << '\t' //
-                          << split_count << '\t'                      //
-                          << one_sketch_estimate << '\t'              //
-                          << (is_merged ? merge_estimate : one_sketch_estimate) << '\n';
+                          << split_count << '\n';
             split_count = 0u; // Subsequent split bins display 0, the first split bin displays the actual split count.
         }
     };
@@ -191,7 +176,6 @@ int execute(config const & cfg)
 
             // reset all current data
             sketch.reset();
-            sketches.clear();
             current_kmers.clear();
             shared_kmers.clear();
             shared_kmers_initialised = false;
@@ -213,7 +197,7 @@ int execute(config const & cfg)
         {
             ++ub_count; // This assumes that each user bin has exactly one associated file. Currently the case.
 
-            process_file(filename, current_kmer_set, current_kmers, sketch, sketches, fill_current_kmers, chopper_config.k);
+            process_file(filename, current_kmer_set, current_kmers, sketch, fill_current_kmers, chopper_config.k);
         }
 
         // Compute set intersection: shared_kmers = shared_kmers ∩ current_kmers

diff --git a/src/display_layout/process_file.cpp b/src/display_layout/process_file.cpp
@@ -25,13 +25,9 @@ void process_file(std::string const & filename,
                   robin_hood::unordered_set<uint64_t> & current_kmer_set,
                   std::vector<uint64_t> & current_kmers,
                   seqan::hibf::sketch::hyperloglog & sketch,
-                  std::vector<seqan::hibf::sketch::hyperloglog> & sketches,
                   bool const fill_current_kmers,
                   uint8_t const kmer_size)
 {
-    seqan::hibf::sketch::hyperloglog local_ub_sketch = sketch; // copy sketch configuration
-    local_ub_sketch.reset();
-
     if (filename.ends_with(".minimiser"))
     {
         uint64_t hash{};
@@ -47,7 +43,6 @@ void process_file(std::string const & filename,
                 current_kmers.push_back(hash);
                 current_kmer_set.insert(hash);
                 sketch.add(hash);
-                local_ub_sketch.add(hash);
             }
         }
         else
@@ -56,7 +51,6 @@ void process_file(std::string const & filename,
             {
                 sketch.add(hash);
                 current_kmer_set.insert(hash);
-                local_ub_sketch.add(hash);
             }
         }
     }
@@ -77,7 +71,6 @@ void process_file(std::string const & filename,
                     current_kmers.push_back(hash_value);
                     current_kmer_set.insert(hash_value);
                     sketch.add(hash_value);
-                    local_ub_sketch.add(hash_value);
                 }
             }
         }
@@ -89,13 +82,10 @@ void process_file(std::string const & filename,
                 {
                     current_kmer_set.insert(hash_value);
                     sketch.add(hash_value);
-                    local_ub_sketch.add(hash_value);
                 }
             }
         }
     }
-
-    sketches.push_back(local_ub_sketch);
 }
 
 void process_file(std::string const & filename,

diff --git a/src/display_layout/shared.hpp b/src/display_layout/shared.hpp
@@ -34,6 +34,5 @@ void process_file(std::string const & filename,
                   robin_hood::unordered_set<uint64_t> & current_kmer_set,
                   std::vector<uint64_t> & current_kmers,
                   seqan::hibf::sketch::hyperloglog & sketch,
-                  std::vector<seqan::hibf::sketch::hyperloglog> & sketches,
                   bool const fill_current_kmers,
                   uint8_t const kmer_size);