Merge pull request #41 from yhoogstrate/mount_2bit

v1.5.0: Allows mounting UCSC 2bit files onto the filesystem using FUSE

CMakeLists.txt

@@ -8,7 +8,7 @@ project(fastafs)
 # Do this once in a while - find different bugs
 #set(CMAKE_CXX_COMPILER             "clang++")
 
-set(PROJECT_VERSION "1.4.0")
+set(PROJECT_VERSION "1.5.0")
 set(PACKAGE_URL "https://github.com/yhoogstrate/fastafs")
 set(PACKAGE_BUGREPORT "${PACKAGE_URL}/issues")
 
@@ -75,7 +75,7 @@ add_custom_target(tidy DEPENDS make_tidy )
 
 add_subdirectory(src)
 include_directories(include)
-add_definitions(-std=c++14)
+add_definitions(-std=c++17)
 
 # Boost
 find_package(Boost COMPONENTS unit_test_framework REQUIRED)
@@ -103,6 +103,7 @@ add_executable(fastafs
 					src/fasta_to_fastafs.cpp
 					src/ucsc2bit_to_fastafs.cpp
 					src/fastafs.cpp
+					src/ucsc2bit.cpp
 					src/twobit_byte.cpp
 					src/database.cpp
 					src/utils.cpp
@@ -125,8 +126,9 @@ add_test(test_cache               "${BUILD_TEST_DIR}/test_cache")
 add_test(test_view                "${BUILD_TEST_DIR}/test_view")
 #add_test(test_tree                "${BUILD_TEST_DIR}/test_tree")
 add_test(test_fastafs             "${BUILD_TEST_DIR}/test_fastafs")
-add_test(test_ucsc2bit            "${BUILD_TEST_DIR}/test_ucsc2bit")
+add_test(test_fastafs_as_ucsc2bit "${BUILD_TEST_DIR}/test_ucsc2bit")
 add_test(test_ucsc2bit_to_fastafs "${BUILD_TEST_DIR}/test_ucsc2bit_to_fastafs")
+add_test(test_ucsc2bit_as_fasta   "${BUILD_TEST_DIR}/test_ucsc2bit_as_fasta")
 add_test(test_utils               "${BUILD_TEST_DIR}/test_utils")
 
 

Changelog

@@ -1,3 +1,9 @@
+2019-08-02  Youri Hoogstrate
+
+    * v1.5.0
+    * Support mounting 2bit files (without FASTAFS file(s))
+       - Requires c++-17 for several std::filesystem calls
+
 2019-06-05  Youri Hoogstrate
 
     * Changed SHA1 to MD5 hashes for BAM compatibility

include/config.hpp.in

@@ -31,13 +31,34 @@
 #define GIT_SHA1_STRING "@GIT_SHA1_STRING@"
 
 
+/* https://stackoverflow.com/questions/26652904/boost-check-equal-and-dervatives-add-custom-message/26687584 */
+#define BOOST_CHECK_EQUAL_MESSAGE(L, R, M)      { BOOST_TEST_INFO(M); BOOST_CHECK_EQUAL(L, R); }
+
+
+
 #define NUCLEOTIDE_T 0 // 00
 #define NUCLEOTIDE_C 1 // 01
 #define NUCLEOTIDE_A 2 // 10
 #define NUCLEOTIDE_G 3 // 11
 
-//#define TWOBIT_MAGIC "\x43\x27\x41\x1a";
-//#define TWOBIT_VERSION "\x00\x00\x00\x00"
 
-/* https://stackoverflow.com/questions/26652904/boost-check-equal-and-dervatives-add-custom-message/26687584 */
-#define BOOST_CHECK_EQUAL_MESSAGE(L, R, M)      { BOOST_TEST_INFO(M); BOOST_CHECK_EQUAL(L, R); }
+
+
+#ifndef CONFIG_HPP
+#define CONFIG_HPP
+
+using namespace std::literals;
+
+
+static const int READ_BUFFER_SIZE = 4096;
+
+static const std::string UCSC2BIT_MAGIC = "\x43\x27\x41\x1a"s;
+static const std::string UCSC2BIT_VERSION = "\x00\x00\x00\x00"s;
+
+static const std::string FASTAFS_MAGIC = "\x0F\x0A\x46\x53"s;
+static const std::string FASTAFS_VERSION = "\x00\x00\x00\x00"s;
+
+static const std::string DICT_HEADER = "@HD\tVN:1.0\tSO:unsorted\n";
+
+
+#endif

include/fastafs.hpp

@@ -126,18 +126,5 @@ class fastafs
 };
 
 
-// Constants:
-using namespace std::literals;
-static const std::string UCSC2BIT_MAGIC = "\x43\x27\x41\x1a"s;
-static const std::string UCSC2BIT_VERSION = "\x00\x00\x00\x00"s;
-
-static const std::string FASTAFS_MAGIC = "\x0F\x0A\x46\x53"s;
-static const std::string FASTAFS_VERSION = "\x00\x00\x00\x00"s;
-
-static const int READ_BUFFER_SIZE = 4096;
-
-
-static const std::string DICT_HEADER = "@HD\tVN:1.0\tSO:unsorted\n";
-
 
 #endif

include/ucsc2bit.hpp

@@ -0,0 +1,64 @@
+
+#include <vector>
+
+
+#include "utils.hpp"
+
+#ifndef UCSC2BIT_HPP
+#define UCSC2BIT_HPP
+
+
+
+
+class ucsc2bit_seq
+{
+public:
+    explicit ucsc2bit_seq();
+
+    std::string name;//may not exceed 255 chars in current datatype
+    uint32_t data_position;// file offset to start reading sequence data
+    uint32_t sequence_data_position;// file offset to start reading sequence data
+
+    uint32_t n;// number nucleotides [ACTG + N]
+
+    std::vector<uint32_t> n_starts;// start positions (nucleotide positions; 0-based)
+    std::vector<uint32_t> n_ends;// end positions (nucleotide positions; 0-based)
+
+    std::vector<uint32_t> m_starts;// start positions (nucleotide positions; 0-based)
+    std::vector<uint32_t> m_ends;// end positions (nucleotide positions; 0-based)
+
+    uint32_t fasta_filesize(uint32_t);
+    uint32_t n_padding(uint32_t, uint32_t, uint32_t);
+
+    uint32_t view_fasta_chunk(uint32_t, char*, size_t, off_t, std::ifstream*);
+};
+
+
+
+class ucsc2bit
+{
+public:
+
+    explicit ucsc2bit(std::string);
+    ~ucsc2bit();
+
+    std::string name;// needed as basename for mounting
+    std::string filename;
+
+    void load(std::string);
+
+    std::vector<ucsc2bit_seq*> data;
+
+
+    uint32_t n();
+    size_t fasta_filesize(uint32_t);
+
+
+    uint32_t view_fasta_chunk(uint32_t, char*, size_t, off_t);
+    uint32_t view_faidx_chunk(uint32_t, char*, size_t, off_t);
+    std::string get_faidx(uint32_t);//@todo get rid of this, make it full chunked
+};
+
+
+
+#endif