Parallelization in the python bindings

README.md

@@ -113,6 +113,26 @@ Once installed, you can import and use the Digest library in Python:
 ['ACGTA', 'CGTAG', 'AGCTA', 'TAGCT', 'GCTGA', 'TTACA', 'TACAT', 'GTATG', 'GCAAG', 'TGATC', 'CGTAG', 'TAGTG', 'ATGCT']
 ```
 
+We have also implemented parallel execution in the python library:
+```
+>>> import timeit
+>>> timeit.timeit('window_minimizer(seq, k=5, w=11)', 
+...     setup='from digest import window_minimizer; seq = open("tests/bench/chrY.fa", "rb").read()',
+...     number=10)
+19.85955114942044
+>>> timeit.timeit('window_minimizer(seq, k=5, w=11, num_threads=4)',
+...     setup='from digest import window_minimizer; seq = open("tests/bench/chrY.fa", "rb").read()',
+...     number=10)
+10.327348310500383
+```
+
+Functions can also return numpy arrays with the `*_np` function variant:
+```
+>>> from digest import window_minimizer_np
+>>> window_minimizer_np(b'ACGTACGTAGCTAGCTAGCTAGCTGATTACATACTGTATGCAAGCTAGCTGATCGATCGTAGCTAGTGATGCTAGCTAC', k=5, w=11)
+array([ 4,  5, 16, 19, 21, 26, 27, 35, 39, 49, 57, 63, 68], dtype=uint32)
+```
+
 <!---
 # Implementation
 Supports Mod Minimizers, Window Minimizers, and Syncmers  

include/digest/thread_out.hpp

@@ -173,7 +173,7 @@ void thread_mod(
 		ind += assigned_kmer_am;
 	}
 	for (auto &t : thread_vector) {
-		vec.emplace_back(t.get());
+		vec.emplace_back(std::move(t.get()));
 	}
 }
 
@@ -235,7 +235,7 @@ void thread_mod(
 		ind += assigned_kmer_am;
 	}
 	for (auto &t : thread_vector) {
-		vec.emplace_back(t.get());
+		vec.emplace_back(std::move(t.get()));
 	}
 }
 
@@ -305,14 +305,14 @@ void thread_wind(
 			extras--;
 		}
 
-		thread_vector.emplace_back(std::async(thread_wind_roll1<P, T>, seq, ind,
+		thread_vector.push_back(std::async(thread_wind_roll1<P, T>, seq, ind,
 											  k, large_wind_kmer_am,
 											  minimized_h, assigned_lwind_am));
 
 		ind += assigned_lwind_am;
 	}
 	for (auto &t : thread_vector) {
-		vec.emplace_back(t.get());
+		vec.emplace_back(std::move(t.get()));
 	}
 
 	// handle duplicates
@@ -365,7 +365,7 @@ void thread_wind(
 	unsigned lwinds_per_thread = num_lwinds / thread_count;
 	unsigned extras = num_lwinds % thread_count;
 	vec.reserve(thread_count);
-	std::vector<std::future<std::pair<uint32_t, uint32_t>>> thread_vector;
+	std::vector<std::future<std::vector<std::pair<uint32_t, uint32_t>>>> thread_vector;
 
 	size_t ind = start;
 	for (unsigned i = 0; i < thread_count; i++) {
@@ -377,14 +377,14 @@ void thread_wind(
 			extras--;
 		}
 
-		thread_vector.emplace_back(std::async(thread_wind_roll2<P, T>, seq, ind,
+		thread_vector.push_back(std::async(thread_wind_roll2<P, T>, seq, ind,
 											  k, large_wind_kmer_am,
 											  minimized_h, assigned_lwind_am));
 
 		ind += assigned_lwind_am;
 	}
 	for (auto &t : thread_vector) {
-		vec.emplace_back(t.get());
+		vec.emplace_back(std::move(t.get()));
 	}
 
 	// handle duplicates
@@ -473,7 +473,7 @@ void thread_sync(
 	}
 	for (auto &t : thread_vector) {
 
-		vec.emplace_back(t.get());
+		vec.emplace_back(std::move(t.get()));
 	}
 }
 
@@ -535,7 +535,7 @@ void thread_sync(
 		ind += assigned_lwind_am;
 	}
 	for (auto &t : thread_vector) {
-		vec.emplace_back(t.get());
+		vec.emplace_back(std::move(t.get()));
 	}
 }
 

pybind/bindings.cpp

@@ -6,15 +6,53 @@ namespace py = pybind11;
 
 PYBIND11_MODULE(digest, m) {
 	m.doc() = "bindings for digest";
-	m.def("window_minimizer", &window_minimizer,
-		  "A function that runs window minimizer digestion", py::arg("seq"),
-		  py::arg("k") = 31, py::arg("w") = 11,
+
+	m.def("window_minimizer", 
+		  [](std::string_view seq, unsigned k, unsigned w, unsigned num_threads, bool include_hash) {
+			  return window_minimizer_wrapper(seq.data(), seq.size(), k, w, num_threads, include_hash);
+		  },
+		  "A function that runs window minimizer digestion",
+		  py::call_guard<py::gil_scoped_release>(),
+		  py::arg("seq"), py::arg("k") = 31, py::arg("w") = 11, py::arg("num_threads") = 1,
 		  py::arg("include_hash") = false);
-	m.def("modimizer", &modimizer,
-		  "A function that runs mod-minimizer digestion", py::arg("seq"),
-		  py::arg("k") = 31, py::arg("mod") = 100,
+
+	m.def("modimizer", 
+		  [](std::string_view seq, unsigned k, uint32_t mod, unsigned num_threads, bool include_hash) {
+			  return modimizer_wrapper(seq.data(), seq.size(), k, mod, num_threads, include_hash);
+		  },
+		  "A function that runs mod-minimizer digestion",
+		  py::call_guard<py::gil_scoped_release>(),
+		  py::arg("seq"), py::arg("k") = 31, py::arg("mod") = 100, py::arg("num_threads") = 1,
+		  py::arg("include_hash") = false);
+
+	m.def("syncmer", 
+		  [](std::string_view seq, unsigned k, unsigned w, unsigned num_threads, bool include_hash) {
+			  return syncmer_wrapper(seq.data(), seq.size(), k, w, num_threads, include_hash);
+		  },
+		  "A function that runs syncmer digestion",
+		  py::call_guard<py::gil_scoped_release>(),
+		  py::arg("seq"), py::arg("k") = 31, py::arg("w") = 11, py::arg("num_threads") = 1,
+		  py::arg("include_hash") = false);
+
+	m.def("window_minimizer_np",
+		  [](std::string_view seq, unsigned k, unsigned w, unsigned num_threads, bool include_hash) {
+			  return window_minimizer_numpy(seq.data(), seq.size(), k, w, num_threads, include_hash);
+		  },
+		  "A function that runs window minimizer digestion and returns a numpy array",
+		  py::arg("seq"), py::arg("k") = 31, py::arg("w") = 11, py::arg("num_threads") = 1,
+		  py::arg("include_hash") = false);
+	m.def("modimizer_np",
+		  [](std::string_view seq, unsigned k, uint32_t mod, unsigned num_threads, bool include_hash) {
+			  return modimizer_numpy(seq.data(), seq.size(), k, mod, num_threads, include_hash);
+		  },
+		  "A function that runs mod-minimizer digestion and returns a numpy array",
+		  py::arg("seq"), py::arg("k") = 31, py::arg("mod") = 100, py::arg("num_threads") = 1,
 		  py::arg("include_hash") = false);
-	m.def("syncmer", &syncmer, "A function that runs syncmer digestion",
-		  py::arg("seq"), py::arg("k") = 31, py::arg("w") = 11,
+	m.def("syncmer_np",
+		  [](std::string_view seq, unsigned k, unsigned w, unsigned num_threads, bool include_hash) {
+			  return syncmer_numpy(seq.data(), seq.size(), k, w, num_threads, include_hash);
+		  },
+		  "A function that runs syncmer digestion and returns a numpy array",
+		  py::arg("seq"), py::arg("k") = 31, py::arg("w") = 11, py::arg("num_threads") = 1,
 		  py::arg("include_hash") = false);
 }