Merge pull request #172 from bab2min/dev_18

Prepare v0.18.0
bab2min · Jul 2, 2024 · 4c87dcd · 4c87dcd
2 parents 30c9f7e + 0ab6c9d
commit 4c87dcd
Show file tree

Hide file tree

Showing 22 changed files with 265 additions and 48 deletions.
diff --git a/.github/workflows/macos.yml b/.github/workflows/macos.yml
@@ -9,22 +9,16 @@ jobs:
     strategy:
       matrix:
         include:
-          - name: "macOS 11 + Xcode 11.7"
-            os: macos-11
+          - name: "macOS 13 + Xcode 15.0"
+            os: macos-13
+            arch: x86_64
             compiler: xcode
-            version: "11.7"
-          - name: "macOS 11 + Xcode 12.2"
-            os: macos-11
+            version: "15.0"
+          - name: "macOS 14 Arm64 + Xcode 15.0"
+            os: macos-14
+            arch: arm64
             compiler: xcode
-            version: "12.4"
-          - name: "macOS 11 + gcc-10"
-            os: macos-11
-            compiler: gcc
-            version: "10"
-          - name: "macOS 11 + gcc-11"
-            os: macos-11
-            compiler: gcc
-            version: "11"
+            version: "15.0"
 
     runs-on: ${{ matrix.os }}
     name: ${{ matrix.name }}
@@ -42,8 +36,8 @@ jobs:
         else
           ls -ls /Applications/
           sudo xcode-select -switch /Applications/Xcode_${{ matrix.version }}.app
-          echo "CC=clang" >> $GITHUB_ENV
-          echo "CXX=clang++" >> $GITHUB_ENV
+          echo "CC=$(brew --prefix llvm@15)/bin/clang" >> $GITHUB_ENV
+          echo "CXX=$(brew --prefix llvm@15)/bin/clang++" >> $GITHUB_ENV
         fi
     - name: Configure Build
       run: mkdir build && cd build && cmake -DCMAKE_BUILD_TYPE=Release -DKIWI_JAVA_BINDING=1 ..
@@ -74,11 +68,17 @@ jobs:
         ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
         KIWI_ARCH_TYPE=none ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
         KIWI_ARCH_TYPE=balanced ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
-        KIWI_ARCH_TYPE=sse2 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
-        KIWI_ARCH_TYPE=sse4_1 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
-        KIWI_ARCH_TYPE=avx2 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
-        KIWI_ARCH_TYPE=avx2 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out --sbg kowiki1000.txt
-        KIWI_ARCH_TYPE=avx2 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out --typos 6 kowiki1000.txt
+        if [ "${{ matrix.arch }}" = "x86_64" ]; then
+          KIWI_ARCH_TYPE=sse2 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
+          KIWI_ARCH_TYPE=sse4_1 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
+          KIWI_ARCH_TYPE=avx2 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
+          KIWI_ARCH_TYPE=avx2 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out --sbg kowiki1000.txt
+          KIWI_ARCH_TYPE=avx2 ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out --typos 6 kowiki1000.txt
+        else
+          KIWI_ARCH_TYPE=neon ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out kowiki1000.txt
+          KIWI_ARCH_TYPE=neon ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out --sbg kowiki1000.txt
+          KIWI_ARCH_TYPE=neon ./build/kiwi-cli-* -m ./ModelGenerator -e -o test.out --typos 6 kowiki1000.txt
+        fi
     - name: Archive binaries
       uses: actions/upload-artifact@v2
       with:

diff --git a/.github/workflows/release.yml b/.github/workflows/release.yml
@@ -144,7 +144,7 @@ jobs:
   build-macos:
     strategy:
       matrix:
-          os: [macos-11]
+          os: [macos-13]
           arch: [x86_64, arm64]
 
     runs-on: ${{ matrix.os }}

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -1,6 +1,6 @@
 cmake_minimum_required(VERSION 3.12)
 
-project(kiwi VERSION 0.17.1 DESCRIPTION "Kiwi, Korean Intelligent Word Identifier")
+project(kiwi VERSION 0.18.0 DESCRIPTION "Kiwi, Korean Intelligent Word Identifier")
 
 set ( CMAKE_CXX_STANDARD 14 )
 set ( CMAKE_VERBOSE_MAKEFILE true )
@@ -48,14 +48,15 @@ set ( CORE_SRCS
   src/Form.cpp
   src/FeatureTestor.cpp
   src/FileUtils.cpp
-  src/HSDataset.cpp
+  src/Dataset.cpp
   src/Joiner.cpp
   src/Kiwi.cpp
   src/KiwiBuilder.cpp
   src/KTrie.cpp
   src/PatternMatcher.cpp
   src/search.cpp
   src/ScriptType.cpp
+  src/SubstringExtractor.cpp
   src/SwTokenizer.cpp
   src/TagUtils.cpp
   src/TypoTransformer.cpp

diff --git a/bindings/java/JniUtils.hpp b/bindings/java/JniUtils.hpp
@@ -924,7 +924,8 @@ namespace jni
 			{
 				return CppType{ env, v };
 			}
-			if (!env->IsInstanceOf(v, JIteratorBase::jClass)) throw std::runtime_error{ StringConcat_v<svNotInstanceOf, typeStr, svNullTerm>.data()};
+			// The following line crashes clang compiler. I don't know why, but it's not necessary. So I commented it out.
+			if (!env->IsInstanceOf(v, JIteratorBase::jClass)) throw std::runtime_error{ ""/*StringConcat_v<svNotInstanceOf, typeStr, svNullTerm>.data()*/};
 			return CppType{ env, v };
 		}
 	};

diff --git a/bindings/java/kr/pe/bab2min/Kiwi.java b/bindings/java/kr/pe/bab2min/Kiwi.java
@@ -12,7 +12,7 @@
 
 public class Kiwi implements AutoCloseable  {
 	private long _inst;
-	final private static String _version = "0.17.1";
+	final private static String _version = "0.18.0";
 
 	public static class Match {
 		final static public int none = 0,

diff --git a/include/kiwi/HSDataset.h → include/kiwi/Dataset.h b/include/kiwi/HSDataset.h → include/kiwi/Dataset.h
diff --git a/include/kiwi/Form.h b/include/kiwi/Form.h
@@ -2,8 +2,8 @@
  * @file Form.h
  * @author bab2min (bab2min@gmail.com)
  * @brief 형태 및 형태소에 관한 정보를 담는 구조체들이 선언된 헤더
- * @version 0.17.0
- * @date 2022-09-01
+ * @version 0.18.0
+ * @date 2024-07-01
  * 
  * 
  */

diff --git a/include/kiwi/Kiwi.h b/include/kiwi/Kiwi.h
@@ -2,8 +2,8 @@
  * @file Kiwi.h
  * @author bab2min (bab2min@gmail.com)
  * @brief Kiwi C++ API를 담고 있는 헤더 파일
- * @version 0.17.0
- * @date 2022-09-01
+ * @version 0.18.0
+ * @date 2024-07-01
  * 
  * 
  */
@@ -650,7 +650,7 @@ namespace kiwi
 		 * @param numThreads 모델 및 형태소 분석에 사용할 스레드 개수
 		 * @param options 생성 옵션. `kiwi::BuildOption`을 참조
 		 */
-		KiwiBuilder(const std::string& modelPath, size_t numThreads = 0, BuildOption options = BuildOption::integrateAllomorph | BuildOption::loadDefaultDict, bool useSBG = false);
+		KiwiBuilder(const std::string& modelPath, size_t numThreads = 0, BuildOption options = BuildOption::default_, bool useSBG = false);
 
 		/**
 		 * @brief 현재 KiwiBuilder 객체가 유효한 분석 모델을 로딩한 상태인지 알려준다.

diff --git a/include/kiwi/Macro.h b/include/kiwi/Macro.h
@@ -4,7 +4,7 @@
 #define KIWI_STR(x) KIWI_STR_HELPER(x)
 
 #define KIWI_VERSION_MAJOR 0
-#define KIWI_VERSION_MINOR 17
-#define KIWI_VERSION_PATCH 1
+#define KIWI_VERSION_MINOR 18
+#define KIWI_VERSION_PATCH 0
 
 #define KIWI_VERSION_STRING KIWI_STR(KIWI_VERSION_MAJOR) "." KIWI_STR(KIWI_VERSION_MINOR) "." KIWI_STR(KIWI_VERSION_PATCH)
diff --git a/include/kiwi/SubstringExtractor.h b/include/kiwi/SubstringExtractor.h
@@ -0,0 +1,16 @@
+#pragma once
+
+#include <vector>
+#include <string>
+
+namespace kiwi
+{
+	std::vector<std::pair<std::u16string, size_t>> extractSubstrings(
+		const char16_t* first, 
+		const char16_t* last, 
+		size_t minCnt, 
+		size_t minLength = 2, 
+		size_t maxLength = 32,
+		bool longestOnly = true,
+		char16_t stopChr = 0);
+}
diff --git a/include/kiwi/SwTokenizer.h b/include/kiwi/SwTokenizer.h
@@ -2,8 +2,8 @@
  * @file SwTokenizer.h
  * @author bab2min (bab2min@gmail.com)
  * @brief Subword Tokenizer
- * @version 0.16.1
- * @date 2022-07-28
+ * @version 0.18.0
+ * @date 2024-07-01
  *
  *
  */

diff --git a/include/kiwi/Types.h b/include/kiwi/Types.h
@@ -2,8 +2,8 @@
  * @file Types.h
  * @author bab2min (bab2min@gmail.com)
  * @brief Kiwi C++ API에 쓰이는 주요 타입들을 모아놓은 헤더 파일
- * @version 0.17.0
- * @date 2022-09-01
+ * @version 0.18.0
+ * @date 2024-07-01
  * 
  * 
  */

diff --git a/include/kiwi/capi.h b/include/kiwi/capi.h
@@ -2,8 +2,8 @@
  * @file capi.h
  * @author bab2min (bab2min@gmail.com)
  * @brief Kiwi C API를 담고 있는 헤더 파일
- * @version 0.17.0
- * @date 2022-09-01
+ * @version 0.18.0
+ * @date 2024-07-01
  * 
  * 
  */
@@ -45,7 +45,11 @@ typedef struct {
 	uint32_t line_number; /**< 줄 번호*/
 	uint16_t length; /**< 길이(UTF16 문자 기준) */
 	uint8_t tag; /**< 품사 태그 */
-	uint8_t sense_id; /**< 의미 번호 */
+	union
+	{
+		uint8_t sense_id; /**< 의미 번호 */
+		uint8_t script; /**< 유니코드 영역에 기반한 문자 타입 */
+	};
 	float score; /**< 해당 형태소의 언어모델 점수 */
 	float typo_cost; /**< 오타가 교정된 경우 오타 비용. 그렇지 않은 경우 0 */
 	uint32_t typo_form_id; /**< 교정 전 오타의 형태에 대한 정보 (typoCost가 0인 경우 의미 없음) */
@@ -1008,6 +1012,16 @@ DECL_DLL int kiwi_pt_add_token_to_span_w(kiwi_pretokenized_h handle, int span_id
  */
 DECL_DLL int kiwi_pt_close(kiwi_pretokenized_h handle);
 
+/**
+ * @brief `kiwi_token_info_t`의 `script`가 가리키는 문자 영역의 유니코드 상 이름을 반환합니다.
+ *
+ * @param script `kiwi_token_info_t`의 `script` 필드 값
+ * @return 유니코드 영역의 이름을 반환합니다. 알 수 없을 경우 "Unknown"을 반환합니다.
+ *
+ * @note 이 함수가 반환하는 값은 string literal이므로 별도로 해제할 필요가 없습니다.
+ */
+DECL_DLL const char* kiwi_get_script_name(uint8_t script);
+
 #ifdef __cplusplus  
 }
 #endif 
diff --git a/src/HSDataset.cpp → src/Dataset.cpp b/src/HSDataset.cpp → src/Dataset.cpp
@@ -1,4 +1,4 @@
-#include <kiwi/HSDataset.h>
+#include <kiwi/Dataset.h>
 #include "RaggedVector.hpp"
 
 using namespace kiwi;

diff --git a/src/KiwiBuilder.cpp b/src/KiwiBuilder.cpp
@@ -3,7 +3,7 @@
 
 #include <kiwi/Kiwi.h>
 #include <kiwi/Utils.h>
-#include <kiwi/HSDataset.h>
+#include <kiwi/Dataset.h>
 #include "ArchAvailable.h"
 #include "KTrie.h"
 #include "StrUtils.h"

diff --git a/src/PatternMatcher.cpp b/src/PatternMatcher.cpp
@@ -272,7 +272,7 @@ size_t PatternMatcherImpl::testAbbr(const char16_t* first, const char16_t* last)
 	if (b != last && *b == ' ')
 	{
 		if (l > (isUpperAlpha(*first) ? 5 : 3)) return 0; // reject too long patterns for abbreviation
-		++b;
+		return b - first;
 	}
 	else
 	{
@@ -287,7 +287,6 @@ size_t PatternMatcherImpl::testAbbr(const char16_t* first, const char16_t* last)
 
 		if (b != last && *b == '.') ++b;
 		else return b - first;
-		if (b != last && *b == ' ') ++b;
 	}
 	if (b[-1] == ' ') --b;
 	return b - first;

diff --git a/src/ScriptType.cpp b/src/ScriptType.cpp
@@ -563,7 +563,7 @@ namespace kiwi
         if (type == ScriptType::chess_symbols) return "Chess Symbols";
         if (type == ScriptType::symbols_for_legacy_computing) return "Symbols for Legacy Computing";
         if (type == ScriptType::tags) return "Tags";
-        return "unknown";
+        return "Unknown";
     }
 
     int isEmoji(char32_t c0, char32_t c1)