[MISC] Separate kmer counting from sketches. (#217)

* [MISC] Separate kmer counting from sketches. * [MISC] automatic linting * fix test * [MISC] automatic linting * fix doc * fix benchmark * review updates * fix test --------- Co-authored-by: seqan-actions[bot] <[email protected]>
seqan · Aug 12, 2024 · ed7fc71 · ed7fc71
1 parent 5c570cc
commit ed7fc71
Show file tree

Hide file tree

Showing 6 changed files with 15 additions and 47 deletions.
diff --git a/include/hibf/sketch/compute_sketches.hpp b/include/hibf/sketch/compute_sketches.hpp
@@ -17,21 +17,12 @@ namespace seqan::hibf::sketch
 /*!\brief Computes the kmer_counts and sketches and stores them in the respective vectors for further use.
  * \ingroup hibf_layout
  * \param[in] config The configuration to compute the layout with.
- * \param[in,out] kmer_counts The vector that will store the kmer counts (estimations).
  * \param[in,out] hll_sketches The vector that will store the sketches.
  */
-void compute_sketches(config const & config,
-                      std::vector<size_t> & kmer_counts,
-                      std::vector<sketch::hyperloglog> & hll_sketches);
-
-//!\overload
-void compute_sketches(config const & config,
-                      std::vector<sketch::hyperloglog> & hll_sketches,
-                      std::vector<sketch::minhashes> & minhash_sketches);
+void compute_sketches(config const & config, std::vector<sketch::hyperloglog> & hll_sketches);
 
 //!\overload
 void compute_sketches(config const & config,
-                      std::vector<size_t> & kmer_counts,
                       std::vector<sketch::hyperloglog> & hll_sketches,
                       std::vector<sketch::minhashes> & minhash_sketches);
 

diff --git a/src/hierarchical_interleaved_bloom_filter.cpp b/src/hierarchical_interleaved_bloom_filter.cpp
@@ -32,7 +32,8 @@
 #include <hibf/misc/iota_vector.hpp>
 #include <hibf/misc/timer.hpp>              // for concurrent_timer
 #include <hibf/sketch/compute_sketches.hpp> // for compute_sketches
-#include <hibf/sketch/hyperloglog.hpp>      // for hyperloglog
+#include <hibf/sketch/estimate_kmer_counts.hpp>
+#include <hibf/sketch/hyperloglog.hpp> // for hyperloglog
 
 namespace seqan::hibf
 {
@@ -204,10 +205,12 @@ hierarchical_interleaved_bloom_filter::hierarchical_interleaved_bloom_filter(con
 {
     configuration.validate_and_set_defaults();
 
-    std::vector<size_t> kmer_counts{};
     std::vector<sketch::hyperloglog> sketches{};
+    std::vector<size_t> kmer_counts{};
+
     layout_compute_sketches_timer.start();
-    sketch::compute_sketches(configuration, kmer_counts, sketches);
+    sketch::compute_sketches(configuration, sketches);
+    hibf::sketch::estimate_kmer_counts(sketches, kmer_counts);
     layout_compute_sketches_timer.stop();
 
     // If rearrangement is enabled, i.e. seqan::hibf::config::disable_rearrangement is false:

diff --git a/src/sketch/compute_sketches.cpp b/src/sketch/compute_sketches.cpp
@@ -25,13 +25,10 @@
 namespace seqan::hibf::sketch
 {
 
-void compute_sketches(config const & config,
-                      std::vector<size_t> & kmer_counts,
-                      std::vector<sketch::hyperloglog> & hll_sketches)
+void compute_sketches(config const & config, std::vector<sketch::hyperloglog> & hll_sketches)
 {
     // compute hll_sketches
     hll_sketches.resize(config.number_of_user_bins);
-    kmer_counts.resize(config.number_of_user_bins);
 
     robin_hood::unordered_flat_set<uint64_t> kmers;
 #pragma omp parallel for schedule(dynamic) num_threads(config.threads) private(kmers)
@@ -47,8 +44,6 @@ void compute_sketches(config const & config,
 
         hll_sketches[i] = std::move(hll_sketch);
     }
-
-    sketch::estimate_kmer_counts(hll_sketches, kmer_counts);
 }
 
 /*!\brief Encapsulates handling of too few kmers to compute minHash sketches.
@@ -164,14 +159,4 @@ void compute_sketches(config const & config,
     too_few_kmers.check_and_throw();
 }
 
-void compute_sketches(config const & config,
-                      std::vector<size_t> & kmer_counts,
-                      std::vector<sketch::hyperloglog> & hll_sketches,
-                      std::vector<sketch::minhashes> & minhash_sketches)
-{
-    compute_sketches(config, hll_sketches, minhash_sketches);
-    kmer_counts.resize(config.number_of_user_bins);
-    sketch::estimate_kmer_counts(hll_sketches, kmer_counts);
-}
-
 } // namespace seqan::hibf::sketch
diff --git a/test/performance/sketch/compute_sketches_benchmark.cpp b/test/performance/sketch/compute_sketches_benchmark.cpp
@@ -31,7 +31,6 @@ void compute_sketches(benchmark::State & state)
             it = i;
     };
 
-    std::vector<size_t> kmer_counts;
     [[maybe_unused]] std::vector<seqan::hibf::sketch::minhashes> minhash_sketches;
     std::vector<seqan::hibf::sketch::hyperloglog> hyperloglog_sketches;
 
@@ -43,9 +42,9 @@ void compute_sketches(benchmark::State & state)
     for (auto _ : state)
     {
         if constexpr (sketch_t == sketch::MinHashes)
-            seqan::hibf::sketch::compute_sketches(config, kmer_counts, hyperloglog_sketches, minhash_sketches);
+            seqan::hibf::sketch::compute_sketches(config, hyperloglog_sketches, minhash_sketches);
         else
-            seqan::hibf::sketch::compute_sketches(config, kmer_counts, hyperloglog_sketches);
+            seqan::hibf::sketch::compute_sketches(config, hyperloglog_sketches);
     }
 }
 

diff --git a/test/unit/hibf/layout/compute_layout_test.cpp b/test/unit/hibf/layout/compute_layout_test.cpp
@@ -14,7 +14,8 @@
 #include <hibf/misc/iota_vector.hpp>
 #include <hibf/misc/timer.hpp>              // for concurrent_timer
 #include <hibf/sketch/compute_sketches.hpp> // for compute_sketches
-#include <hibf/sketch/hyperloglog.hpp>      // for hyperloglog
+#include <hibf/sketch/estimate_kmer_counts.hpp>
+#include <hibf/sketch/hyperloglog.hpp> // for hyperloglog
 
 TEST(compute_layout, dispatch)
 {
@@ -33,7 +34,8 @@ TEST(compute_layout, dispatch)
     std::vector<seqan::hibf::sketch::hyperloglog> sketches;
     std::vector<size_t> kmer_counts;
 
-    seqan::hibf::sketch::compute_sketches(config, kmer_counts, sketches);
+    seqan::hibf::sketch::compute_sketches(config, sketches);
+    seqan::hibf::sketch::estimate_kmer_counts(sketches, kmer_counts);
 
     auto layout1 = seqan::hibf::layout::compute_layout(config, kmer_counts, sketches);
 

diff --git a/test/unit/hibf/sketch/compute_sketches_test.cpp b/test/unit/hibf/sketch/compute_sketches_test.cpp
@@ -80,9 +80,8 @@ class compute_sketches_test : public ::testing::Test
 
 TEST_F(compute_sketches_test, hyperloglog_and_kmer_counts)
 {
-    seqan::hibf::sketch::compute_sketches(this->config, this->kmer_counts, this->hyperloglog_sketches);
+    seqan::hibf::sketch::compute_sketches(this->config, this->hyperloglog_sketches);
 
-    this->check_kmer_counts();
     this->check_hyperloglog_sketches();
 }
 
@@ -94,17 +93,6 @@ TEST_F(compute_sketches_test, with_minHash)
     this->check_minhash_sketches();
 }
 
-TEST_F(compute_sketches_test, with_minHash_and_kmer_counts)
-{
-    seqan::hibf::sketch::compute_sketches(this->config,
-                                          this->kmer_counts,
-                                          this->hyperloglog_sketches,
-                                          this->minhash_sketches);
-
-    this->check_hyperloglog_sketches();
-    this->check_minhash_sketches();
-}
-
 TEST_F(compute_sketches_test, too_few_hashes)
 {
     this->config.number_of_user_bins = 1;