Fixes boundary problem in indices conversion. am: 45c9b56574oreo-dr1-dev

am: 5e5fa7f6ed

Change-Id: I2519824d5c7992762f541ba5e10a2a2deb5ad38a

2 files changed, 30 insertions, 4 deletions

diff --git a/tests/textclassifier_jni_test.cc b/tests/textclassifier_jni_test.cc
index c441cf5..ffc193b 100644
--- a/tests/textclassifier_jni_test.cc
+++ b/tests/textclassifier_jni_test.cc
@@ -23,8 +23,27 @@ namespace libtextclassifier {
 namespace {
 
 TEST(TextClassifier, ConvertIndicesBMPUTF8) {
-  EXPECT_EQ(ConvertIndicesBMPToUTF8("hello", {0, 5}),
-            ConvertIndicesUTF8ToBMP("hello", {0, 5}));
+  // Test boundary cases.
+  EXPECT_EQ(ConvertIndicesBMPToUTF8("hello", {0, 5}), std::make_pair(0, 5));
+  EXPECT_EQ(ConvertIndicesUTF8ToBMP("hello", {0, 5}), std::make_pair(0, 5));
+
+  EXPECT_EQ(ConvertIndicesBMPToUTF8("hello world", {0, 5}),
+            std::make_pair(0, 5));
+  EXPECT_EQ(ConvertIndicesUTF8ToBMP("hello world", {0, 5}),
+            std::make_pair(0, 5));
+  EXPECT_EQ(ConvertIndicesBMPToUTF8("😁ello world", {0, 6}),
+            std::make_pair(0, 5));
+  EXPECT_EQ(ConvertIndicesUTF8ToBMP("😁ello world", {0, 5}),
+            std::make_pair(0, 6));
+
+  EXPECT_EQ(ConvertIndicesBMPToUTF8("hello world", {6, 11}),
+            std::make_pair(6, 11));
+  EXPECT_EQ(ConvertIndicesUTF8ToBMP("hello world", {6, 11}),
+            std::make_pair(6, 11));
+  EXPECT_EQ(ConvertIndicesBMPToUTF8("hello worl😁", {6, 12}),
+            std::make_pair(6, 11));
+  EXPECT_EQ(ConvertIndicesUTF8ToBMP("hello worl😁", {6, 11}),
+            std::make_pair(6, 12));
 
   // Simple example where the longer character is before the selection.
   //  character 😁 is 0x1f601
diff --git a/textclassifier_jni.cc b/textclassifier_jni.cc
index 84a1e32..8d64d87 100644
--- a/textclassifier_jni.cc
+++ b/textclassifier_jni.cc
@@ -116,8 +116,8 @@ CodepointSpan ConvertIndicesBMPUTF8(const std::string& utf8_str,
   }
 
   CodepointSpan result{-1, -1};
-  for (auto it = unicode_str.begin(); it != unicode_str.end();
-       ++it, ++unicode_index, ++bmp_index) {
+  std::function<void()> assign_indices_fn = [&result, &orig_indices,
+                                             &source_index, &target_index]() {
     if (orig_indices.first == *source_index) {
       result.first = *target_index;
     }
@@ -125,12 +125,19 @@ CodepointSpan ConvertIndicesBMPUTF8(const std::string& utf8_str,
     if (orig_indices.second == *source_index) {
       result.second = *target_index;
     }
+  };
+
+  for (auto it = unicode_str.begin(); it != unicode_str.end();
+       ++it, ++unicode_index, ++bmp_index) {
+    assign_indices_fn();
 
     // There is 1 extra character in the input for each UTF8 character > 0xFFFF.
     if (*it > 0xFFFF) {
       ++bmp_index;
     }
   }
+  assign_indices_fn();
+
   return result;
 }