Generalized suffix array (#396)

* gsa

* gsa

* save

* save

sais/gsa/gsa_test.go

@@ -0,0 +1,40 @@
+package gsa
+
+import (
+	"fmt"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func ExampleGSA() {
+	R := [][]byte{[]byte("hihihi")}
+	str, n := ConcatAll(R)
+	sa2 := make([]uint, SaSize(n))
+	lcp := make([]int, LcpSize(n))
+	_ = GSA(str, sa2, lcp, nil)
+	for i := 0; i < n; i++ {
+		j := sa2[i]
+		for ; int(j) < n; j++ {
+			if str[j] == 1 {
+				fmt.Printf("$")
+				break
+			} else if str[j] == 0 {
+				fmt.Printf("#")
+			} else {
+				fmt.Printf("%c", str[j]-1)
+			}
+		}
+		fmt.Printf("\n")
+	}
+	fmt.Printf("%d\n", sa2)
+}
+
+func TestGSA(t *testing.T) {
+	R := [][]byte{{4, 5, 6, 4, 5, 6, 4, 5, 6}}
+	str, n := ConcatAll(R)
+	sa := make([]uint, SaSize(n))
+	lcp := make([]int, n)
+	_ = GSA(str, sa, lcp, nil)
+	assert.Equal(t, []uint{10, 9, 6, 3, 0, 7, 4, 1, 8, 5, 2}, sa[:n])
+}

sais/gsa/gsaca.go

@@ -0,0 +1,76 @@
+package gsa
+
+/*
+#include "gsacak.h"
+*/
+import "C"
+import (
+	"unsafe"
+)
+
+// Implementation from https://github.com/felipelouza/gsufsort
+
+func SaSize(l int) int {
+	var a uint
+	return l * int(unsafe.Sizeof(a))
+}
+func LcpSize(l int) int {
+	var a uint
+	return l * int(unsafe.Sizeof(a))
+}
+func GSA(data []byte, sa []uint, lcp []int, da []int32) error {
+	tPtr := unsafe.Pointer(&data[0]) // source "text"
+	var lcpPtr, saPtr, daPtr unsafe.Pointer
+	if sa != nil {
+		saPtr = unsafe.Pointer(&sa[0])
+	}
+	if lcp != nil {
+		lcpPtr = unsafe.Pointer(&lcp[0])
+	}
+	if da != nil {
+		daPtr = unsafe.Pointer(&da[0])
+	}
+	depth := C.gsacak(
+		(*C.uchar)(tPtr),
+		(*C.uint_t)(saPtr),
+		(*C.int_t)(lcpPtr),
+		(*C.int_da)(daPtr),
+		C.uint_t(len(data)),
+	)
+	_ = depth
+	return nil
+}
+
+func ConcatAll(R [][]byte) (str []byte, n int) {
+	for i := 0; i < len(R); i++ {
+		n += len(R[i]) + 1
+	}
+
+	n++ //add 0 at the end
+	str = make([]byte, n)
+	var l, max int
+	k := len(R)
+
+	for i := 0; i < k; i++ {
+		m := len(R[i])
+		if m > max {
+			max = m
+		}
+		for j := 0; j < m; j++ {
+			if R[i][j] < 255 && R[i][j] > 1 {
+				str[l] = R[i][j] + 1
+				l++
+			}
+		}
+		if m > 0 {
+			if str[l-1] > 1 {
+				str[l] = 1
+				l++
+			} //add 1 as separator (ignores empty entries)
+		}
+	}
+	str[l] = 0
+	l++
+	n = l
+	return str, n
+}

sais/gsa/gsacak.h

@@ -0,0 +1,147 @@
+// vim: noai:ts=2:sw=2
+
+/* 
+ * Authors: Felipe A. Louza, Simon Gog, Guilherme P. Telles
+ * contact: louza@ic.unicamp.br
+ * 03/04/2017
+ */
+
+/* 
+ * This code is a modification of SACA-K algorithm by G. Nong, which can be
+ * retrieved at: http://code.google.com/p/ge-nong/ 
+ *
+ * Our version of SACA-K, called gSACA-K, maintain the theoretical bounds of the
+ * original algorithm to construct the generalized suffix array.
+ *
+ * Our algorithm gSACA-K can also computes the LCP-array and the Document-array
+ * with no additional costs.
+ * 
+ * gsacak(s, SA, NULL, NULL, n) //computes only SA
+ * gsacak(s, SA, LCP,  NULL, n) //computes SA and LCP
+ * gsacak(s, SA, NULL, DA, n)   //computes SA and DA
+ * gsacak(s, SA, LCP,  DA, n)   //computes SA, LCP and DA
+ * 
+ */
+
+/******************************************************************************/
+
+#ifndef GSACAK_H
+#define GSACAK_H
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <limits.h>
+#include <inttypes.h>
+#include <string.h>
+#include <time.h>
+
+#define max(a,b) ((a) > (b) ? (a) : (b))
+
+#ifndef DEBUG
+  #define DEBUG 0
+#endif
+
+#ifndef M64
+	#define M64 1
+#endif
+
+#if M64
+	typedef int64_t	int_t;
+	typedef uint64_t uint_t;
+	#define PRIdN	PRId64
+	#define U_MAX	UINT64_MAX
+	#define I_MAX	INT64_MAX
+	#define I_MIN	INT64_MIN
+#else
+	typedef int32_t int_t;
+	typedef uint32_t uint_t;
+	#define PRIdN	PRId32
+	#define U_MAX	UINT32_MAX
+	#define I_MAX	INT32_MAX
+	#define I_MIN	INT32_MIN
+#endif
+
+/*! @option type of s[0,n-1] for integer alphabets 
+ *
+ *  @constraint sizeof(int_t) >= sizeof(int_text) 
+ */
+typedef uint32_t int_text;	//4N bytes for s[0..n-1]
+#define PRIdT	PRIu32
+
+/*! @option type for array DA
+ */
+typedef int32_t int_da;
+#define PRIdA	PRId32
+
+/******************************************************************************/
+
+/** @brief computes the suffix array of string s[0..n-1] 
+ *
+ *  @param s	input string with s[n-1]=0
+ *  @param SA		suffix array 
+ *  @param n	string length
+ *  @return -1 if an error occured, otherwise the depth of the recursive calls.
+ */
+int sacak(unsigned char *s, uint_t *SA, uint_t n);
+
+/** @brief computes the suffix array of string s[0..n-1]
+ *
+ *  @param k	alphabet size+1 (0 is reserved)
+ */
+int sacak_int(int_text *s, uint_t *SA, uint_t n, uint_t k);
+
+/******************************************************************************/
+
+/** @brief Computes the suffix array SA (LCP, DA) of T^cat in s[0..n-1]
+ *
+ *  @param s		input concatenated string, using separators s[i]=1 and with s[n-1]=0
+ *  @param SA		Suffix array 
+ *  @param LCP	LCP array 
+ *  @param DA		Document array
+ *  @param n		String length
+ *  
+ *  @return depth of the recursive calls.
+ */
+int gsacak(unsigned char *s, uint_t *SA, int_t *LCP, int_da *DA, uint_t n);
+
+/** @brief Computes the suffix array SA (LCP, DA) of T^cat in s[0..n-1]
+ *
+ *  @param s		input concatenated string, using separators s[i]=1 and with s[n-1]=0
+ *  @param SA		Suffix array 
+ *  @param LCP	LCP array 
+ *  @param DA		Document array
+ *  @param n		String length
+ *  @param k    alphabet size+2 (0 and 1 are reserved)
+ *
+ *  @return depth of the recursive calls.
+ */
+int gsacak_int(int_text *s, uint_t *SA, int_t *LCP, int_da *DA, uint_t n, uint_t k);
+
+/******************************************************************************/
+
+
+#define m64 1
+
+#if m64
+	typedef int64_t  int_t;
+	typedef uint64_t uint_t;
+	#define PRIdN	 PRId64
+#else
+	typedef int32_t  int_t;
+	typedef uint32_t uint_t;
+	#define PRIdN	 PRId32
+#endif
+
+typedef uint32_t int_text;
+
+
+
+int_t SACA_K(int_t	*s, uint_t *SA,
+  uint_t n, unsigned int K,
+  uint_t m, int cs, int level);
+
+int_t gSACA_K(uint_t *s, uint_t *SA,
+  uint_t n, unsigned int K,
+  int cs, uint_t separator, int level);
+
+#endif