[MISC] Minor changes that will ease the transition once we move to th…

…e HIBF library (#291)

* [INFRA] Update chopper

* [MISC] Resolve circular include of hierarchical_build.hpp and initialise_max_bin_kmers.hpp

* [MISC] Remove doubling of prefixes

* [MISC] automatic linting

---------

Co-authored-by: seqan-actions[bot] <[email protected]>

src/argument_parsing/build_parsing.cpp

@@ -13,14 +13,14 @@
 #include <cereal/archives/json.hpp>
 
 #include <chopper/configuration.hpp>
+#include <chopper/prefixes.hpp>
 
 #include <raptor/argument_parsing/build_parsing.hpp>
 #include <raptor/argument_parsing/compute_bin_size.hpp>
 #include <raptor/argument_parsing/init_shared_meta.hpp>
 #include <raptor/argument_parsing/parse_bin_path.hpp>
 #include <raptor/argument_parsing/shared.hpp>
 #include <raptor/argument_parsing/validators.hpp>
-#include <raptor/build/hibf/bin_prefixes.hpp>
 #include <raptor/build/raptor_build.hpp>
 
 namespace raptor
@@ -234,7 +234,7 @@ bool input_is_pack_file(std::filesystem::path const & path)
     std::string line{};
     while (std::getline(file, line) && line.starts_with("##")) // Skip parameter information
     {}
-    return line.starts_with(raptor::hibf::pack_file_first_line_prefix);
+    return line.starts_with(chopper::prefix::first_header_line);
 }
 
 void build_parsing(sharg::parser & parser)

src/build/hibf/CMakeLists.txt

@@ -8,7 +8,6 @@ if (NOT TARGET raptor_build_hibf)
                  construct_ibf.cpp
                  create_ibfs_from_chopper_pack.cpp
                  hierarchical_build.cpp
-                 initialise_max_bin_kmers.cpp
                  insert_into_ibf.cpp
                  loop_over_children.cpp
                  parse_chopper_pack_header.cpp

src/build/hibf/hierarchical_build.cpp

@@ -15,7 +15,6 @@
 #include <raptor/build/hibf/compute_kmers.hpp>
 #include <raptor/build/hibf/construct_ibf.hpp>
 #include <raptor/build/hibf/hierarchical_build.hpp>
-#include <raptor/build/hibf/initialise_max_bin_kmers.hpp>
 #include <raptor/build/hibf/insert_into_ibf.hpp>
 #include <raptor/build/hibf/loop_over_children.hpp>
 #include <raptor/build/hibf/update_user_bins.hpp>
@@ -38,6 +37,33 @@ size_t hierarchical_build(robin_hood::unordered_flat_set<size_t> & parent_kmers,
     std::vector<int64_t> filename_indices(current_node_data.number_of_technical_bins, -1);
     robin_hood::unordered_flat_set<size_t> kmers{};
 
+    auto initialise_max_bin_kmers = [](robin_hood::unordered_flat_set<size_t> & kmers,
+                                       std::vector<int64_t> & ibf_positions,
+                                       std::vector<int64_t> & filename_indices,
+                                       lemon::ListDigraph::Node const & node,
+                                       build_data<data_layout_mode> & data,
+                                       build_arguments const & arguments) -> size_t
+    {
+        auto & node_data = data.node_map[node];
+
+        if (node_data.favourite_child != lemon::INVALID) // max bin is a merged bin
+        {
+            // recursively initialize favourite child first
+            ibf_positions[node_data.max_bin_index] =
+                hierarchical_build(kmers, node_data.favourite_child, data, arguments, false);
+            return 1;
+        }
+        else // max bin is not a merged bin
+        {
+            // we assume that the max record is at the beginning of the list of remaining records.
+            auto const & record = node_data.remaining_records[0];
+            compute_kmers(kmers, arguments, record);
+            update_user_bins(data, filename_indices, record);
+
+            return record.number_of_bins.back();
+        }
+    };
+
     // initialize lower level IBF
     size_t const max_bin_tbs =
         initialise_max_bin_kmers(kmers, ibf_positions, filename_indices, current_node, data, arguments);

src/build/hibf/parse_chopper_pack_header.cpp

@@ -18,7 +18,6 @@
 
 #include <chopper/prefixes.hpp>
 
-#include <raptor/build/hibf/bin_prefixes.hpp>
 #include <raptor/build/hibf/parse_chopper_pack_header.hpp>
 
 namespace raptor::hibf
@@ -61,11 +60,11 @@ size_t parse_chopper_pack_header(lemon::ListDigraph & ibf_graph,
            && std::string_view{line}.substr(1, 1) == chopper::prefix::header_config)
         ; // skip config in header
 
-    assert(line[0] == '#');                                    // we are reading header lines
-    assert(line.substr(1, hibf_prefix.size()) == hibf_prefix); // first line should always be High level IBF
+    assert(line[0] == '#'); // we are reading header lines
+    assert(line.substr(1, chopper::prefix::high_level.size()) == chopper::prefix::high_level);
 
     // parse High Level max bin index
-    assert(line.substr(hibf_prefix.size() + 2, 11) == "max_bin_id:");
+    assert(line.substr(chopper::prefix::high_level.size() + 2, 11) == "max_bin_id:");
     std::string_view const hibf_max_bin_str{line.begin() + 27, line.end()};
 
     auto high_level_node = ibf_graph.addNode(); // high-level node = root node
@@ -76,14 +75,15 @@ size_t parse_chopper_pack_header(lemon::ListDigraph & ibf_graph,
     // first read and parse header records, in order to sort them before adding them to the graph
     while (std::getline(chopper_pack_file, line) && line.substr(0, 6) != "#FILES")
     {
-        assert(line.substr(1, merged_bin_prefix.size()) == merged_bin_prefix);
+        assert(line.substr(1, chopper::prefix::merged_bin.size()) == chopper::prefix::merged_bin);
 
         // parse header line
-        std::string_view const indices_str{line.begin() + 1 /*#*/ + merged_bin_prefix.size() + 1 /*_*/,
-                                           std::find(line.begin() + merged_bin_prefix.size() + 2, line.end(), ' ')};
+        std::string_view const indices_str{
+            line.begin() + 1 /*#*/ + chopper::prefix::merged_bin.size() + 1 /*_*/,
+            std::find(line.begin() + chopper::prefix::merged_bin.size() + 2, line.end(), ' ')};
 
-        assert(line.substr(merged_bin_prefix.size() + indices_str.size() + 3, 11) == "max_bin_id:");
-        std::string_view const max_id_str{line.begin() + merged_bin_prefix.size() + indices_str.size() + 14,
+        assert(line.substr(chopper::prefix::merged_bin.size() + indices_str.size() + 3, 11) == "max_bin_id:");
+        std::string_view const max_id_str{line.begin() + chopper::prefix::merged_bin.size() + indices_str.size() + 14,
                                           line.end()};
 
         header_records.emplace_back(parse_bin_indices(indices_str), parse_first_bin(max_id_str));

test/unit/cli/argument_parsing/options_test.cpp

@@ -5,7 +5,8 @@
 // shipped with this file and also available at: https://github.com/seqan/raptor/blob/main/LICENSE.md
 // --------------------------------------------------------------------------------------------------
 
-#include <raptor/build/hibf/bin_prefixes.hpp>
+#include <chopper/prefixes.hpp>
+
 #include <raptor/test/cli_test.hpp>
 #include <raptor/test/tmp_test_file.hpp>
 
@@ -27,7 +28,7 @@ std::filesystem::path const
     dummy_sequence_file    = test_files.create("dummy.fasta",   ">ID\nACGTCACGATCGTACGATCGATCGATCG"),
     tmp_bin_list_file      = test_files.create("all_bins.txt",   dummy_sequence_file.c_str()),
     tmp_bin_list_empty     = test_files.create("empty.txt"),
-    tmp_bin_list_corrupted = test_files.create("corrupted.txt",  raptor::hibf::pack_file_first_line_prefix),
+    tmp_bin_list_corrupted = test_files.create("corrupted.txt",  chopper::prefix::first_header_line),
     empty_sequence_file    = test_files.create("empty.fasta"),
     tmp_empty_bin_file     = test_files.create("empty_bin.txt",  empty_sequence_file.c_str()),
     header_file            = test_files.create("bin1.header",    "1111111111111111111\t19\t0\t1\n"),