fix hang on small files

lean/TODO.org

@@ -1,3 +1,8 @@
+
+* Improvements
+  - *~70-80x* speedup from shuffle3 1.0
+  - Can properly handle large files without core dumping
+
 * Todo
 - [X] impl rng
 - [X] impl shuffling

lean/include/reinterpret.hpp

@@ -28,12 +28,12 @@ struct span {
 
 	inline const T& operator[](std::size_t i) const
 	{
-		if (i >= len) panic("Out of bounds access");
+		if (i >= len) panic("Out of bounds access: %lu >= %lu", i, len);
 		return ptr[i];
 	}
 	inline T& operator[](std::size_t i)
 	{
-		if(i >= len) panic("Out of bounds access");
+		if (i >= len) panic("Out of bounds access: %lu >= %lu", i, len);
 		return ptr[i];
 	}
 

lean/include/shuffle.hpp

@@ -9,6 +9,7 @@ namespace rng {
 	template<typename T, typename R>
 	inline void shuffle(R& rng, span<T> span)
 	{
+		if(!span.size()) return;
 		fmt::print(" -> shuffling {} objects...", span.size());
 		for(std::size_t i=span.size()-1;i>0;i--)
 		{
@@ -21,6 +22,7 @@ namespace rng {
 	template<typename T, typename R>
 	inline void unshuffle(R& rng, span<T> span)
 	{
+		if(!span.size()) return;
 		std::vector<std::size_t> rng_values;
 
 		fmt::print(" -> unshuffling {} objects...", span.size());

lean/src/main.c

@@ -47,9 +47,9 @@ int main(int argc, char** argv)
 	
 	//rng_test();
 	
-	rng_t r = rng_new(RNG_INIT(RNG_KIND_FRNG, frng = { { 1.0, 2.0 } }));
-	rng_test_spec(r);
-	rng_free(r);
+	//rng_t r = rng_new(RNG_INIT(RNG_KIND_FRNG, frng = { { 1.0, 2.0 } }));
+	//rng_test_spec(r);
+	//rng_free(r);
 
 	if( argv[1] ) {
 		//map_and_then(argv[1], &map_callback, &args);

lean/src/work.cpp

@@ -15,18 +15,16 @@
 template<typename T, typename Fn>
 std::tuple<T, T> minmax_t(const span<T>& array, Fn keep)
 {
-	T lowest;
 	T highest;
+	T lowest;
 	for(std::size_t i=0;i<array.size();i++)
 	{
 		if(!keep(array[i])) continue;
 
-		if(!i) lowest = highest = array[i];
-		else {
 		auto item = array[i];
-			if(item < lowest) lowest = item;
-			if(item > highest) highest = item;
-		}
+		if(!i) lowest = highest = item;
+		else if(item < lowest) lowest = item;
+		else if(item > highest) highest = item;
 	}
 	//fmt::print("MMX {}, {}\n", lowest, highest);
 	return {lowest, highest};