2 files changed, 70 insertions, 9 deletions

diff --git a/native/jni/src/suggest/core/session/prev_words_info.h b/native/jni/src/suggest/core/session/prev_words_info.h
index 640f6a2fc..e350c6996 100644
--- a/native/jni/src/suggest/core/session/prev_words_info.h
+++ b/native/jni/src/suggest/core/session/prev_words_info.h
@@ -25,7 +25,6 @@
 namespace latinime {
 
 // TODO: Support n-gram.
-// This class does not take ownership of any code point buffers.
 class PrevWordsInfo {
  public:
     // No prev word information.
@@ -33,21 +32,52 @@ class PrevWordsInfo {
         clear();
     }
 
+    PrevWordsInfo(PrevWordsInfo &&prevWordsInfo) {
+        for (size_t i = 0; i < NELEMS(mPrevWordCodePoints); ++i) {
+            mPrevWordCodePointCount[i] = prevWordsInfo.mPrevWordCodePointCount[i];
+            memmove(mPrevWordCodePoints[i], prevWordsInfo.mPrevWordCodePoints[i],
+                    sizeof(mPrevWordCodePoints[i][0]) * mPrevWordCodePointCount[i]);
+            mIsBeginningOfSentence[i] = prevWordsInfo.mIsBeginningOfSentence[i];
+        }
+    }
+
+    // Construct from previous words.
+    PrevWordsInfo(const int prevWordCodePoints[][MAX_WORD_LENGTH],
+            const int *const prevWordCodePointCount, const bool *const isBeginningOfSentence,
+            const size_t prevWordCount) {
+        clear();
+        for (size_t i = 0; i < std::min(NELEMS(mPrevWordCodePoints), prevWordCount); ++i) {
+            if (prevWordCodePointCount[i] < 0 || prevWordCodePointCount[i] > MAX_WORD_LENGTH) {
+                continue;
+            }
+            memmove(mPrevWordCodePoints[i], prevWordCodePoints[i],
+                    sizeof(mPrevWordCodePoints[i][0]) * prevWordCodePointCount[i]);
+            mPrevWordCodePointCount[i] = prevWordCodePointCount[i];
+            mIsBeginningOfSentence[i] = isBeginningOfSentence[i];
+        }
+    }
+
+    // Construct from a previous word.
     PrevWordsInfo(const int *const prevWordCodePoints, const int prevWordCodePointCount,
             const bool isBeginningOfSentence) {
         clear();
-        mPrevWordCodePoints[0] = prevWordCodePoints;
+        if (prevWordCodePointCount > MAX_WORD_LENGTH || !prevWordCodePoints) {
+            return;
+        }
+        memmove(mPrevWordCodePoints[0], prevWordCodePoints,
+                sizeof(mPrevWordCodePoints[0][0]) * prevWordCodePointCount);
         mPrevWordCodePointCount[0] = prevWordCodePointCount;
         mIsBeginningOfSentence[0] = isBeginningOfSentence;
     }
 
     bool isValid() const {
-        for (size_t i = 0; i < NELEMS(mPrevWordCodePoints); ++i) {
-            if (mPrevWordCodePointCount[i] > MAX_WORD_LENGTH) {
-                return false;
-            }
+        if (mPrevWordCodePointCount[0] > 0) {
+            return true;
+        }
+        if (mIsBeginningOfSentence[0]) {
+            return true;
         }
-        return true;
+        return false;
     }
 
     void getPrevWordsTerminalPtNodePos(
@@ -168,13 +198,12 @@ class PrevWordsInfo {
 
     void clear() {
         for (size_t i = 0; i < NELEMS(mPrevWordCodePoints); ++i) {
-            mPrevWordCodePoints[i] = nullptr;
             mPrevWordCodePointCount[i] = 0;
             mIsBeginningOfSentence[i] = false;
         }
     }
 
-    const int *mPrevWordCodePoints[MAX_PREV_WORD_COUNT_FOR_N_GRAM];
+    int mPrevWordCodePoints[MAX_PREV_WORD_COUNT_FOR_N_GRAM][MAX_WORD_LENGTH];
     int mPrevWordCodePointCount[MAX_PREV_WORD_COUNT_FOR_N_GRAM];
     bool mIsBeginningOfSentence[MAX_PREV_WORD_COUNT_FOR_N_GRAM];
 };
diff --git a/native/jni/src/utils/jni_data_utils.h b/native/jni/src/utils/jni_data_utils.h
index 3514aeeb0..cb82d3c3b 100644
--- a/native/jni/src/utils/jni_data_utils.h
+++ b/native/jni/src/utils/jni_data_utils.h
@@ -21,6 +21,7 @@
 
 #include "defines.h"
 #include "jni.h"
+#include "suggest/core/session/prev_words_info.h"
 #include "suggest/core/policy/dictionary_header_structure_policy.h"
 #include "suggest/policyimpl/dictionary/header/header_read_write_utils.h"
 #include "utils/char_utils.h"
@@ -95,6 +96,37 @@ class JniDataUtils {
         }
     }
 
+    static PrevWordsInfo constructPrevWordsInfo(JNIEnv *env, jobjectArray prevWordCodePointArrays,
+            jbooleanArray isBeginningOfSentenceArray) {
+        int prevWordCodePoints[MAX_PREV_WORD_COUNT_FOR_N_GRAM][MAX_WORD_LENGTH];
+        int prevWordCodePointCount[MAX_PREV_WORD_COUNT_FOR_N_GRAM];
+        bool isBeginningOfSentence[MAX_PREV_WORD_COUNT_FOR_N_GRAM];
+        jsize prevWordsCount = env->GetArrayLength(prevWordCodePointArrays);
+        for (size_t i = 0; i < NELEMS(prevWordCodePoints); ++i) {
+            prevWordCodePointCount[i] = 0;
+            isBeginningOfSentence[i] = false;
+            if (prevWordsCount <= static_cast<int>(i)) {
+                continue;
+            }
+            jintArray prevWord = (jintArray)env->GetObjectArrayElement(prevWordCodePointArrays, i);
+            if (!prevWord) {
+                continue;
+            }
+            jsize prevWordLength = env->GetArrayLength(prevWord);
+            if (prevWordLength > MAX_WORD_LENGTH) {
+                continue;
+            }
+            env->GetIntArrayRegion(prevWord, 0, prevWordLength, prevWordCodePoints[i]);
+            prevWordCodePointCount[i] = prevWordLength;
+            jboolean isBeginningOfSentenceBoolean = JNI_FALSE;
+            env->GetBooleanArrayRegion(isBeginningOfSentenceArray, i, 1 /* len */,
+                    &isBeginningOfSentenceBoolean);
+            isBeginningOfSentence[i] = isBeginningOfSentenceBoolean == JNI_TRUE;
+        }
+        return PrevWordsInfo(prevWordCodePoints, prevWordCodePointCount, isBeginningOfSentence,
+                MAX_PREV_WORD_COUNT_FOR_N_GRAM);
+    }
+
     static void putBooleanToArray(JNIEnv *env, jbooleanArray array, const int index,
             const jboolean value) {
         env->SetBooleanArrayRegion(array, index, 1 /* len */, &value);