#998 Add MZ_ICU option for string encoding conversion

Add ICU as an alternative backend to libiconv using ucnv_convert.
ICU is detected via CMake's FindICU module on non-Windows platforms.
Add CI build configuration for ICU in GitHub Actions.

Author: tbeu

.github/workflows/build.yml

@@ -176,6 +176,15 @@ jobs:
             packages: clang-14 llvm-14
             gcov-exec: llvm-cov-14 gcov
 
+          - name: Ubuntu Clang ICU
+            os: ubuntu-latest
+            compiler: clang-14
+            cxx-compiler: clang++-14
+            cmake-args: -D MZ_CODE_COVERAGE=ON -D MZ_ICU=ON -D MZ_ICONV=OFF
+            codecov: ubuntu_clang_icu
+            packages: clang-14 llvm-14 libicu-dev
+            gcov-exec: llvm-cov-14 gcov
+
           # No code coverage supported
           - name: Windows MSVC
             os: windows-latest

CMakeLists.txt

@@ -60,6 +60,7 @@ option(MZ_OPENSSL "Enables OpenSSL for encryption" ${UNIX})
 cmake_dependent_option(MZ_LIBBSD "Builds with libbsd crypto random" ON "UNIX" OFF)
 # Character conversion options
 option(MZ_ICONV "Enables iconv for string encoding conversion" ON)
+option(MZ_ICU "Enables ICU for string encoding conversion" OFF)
 # Code generation options
 option(MZ_COMPRESS_ONLY "Only support compression" OFF)
 option(MZ_DECOMPRESS_ONLY "Only support decompression" OFF)
@@ -586,6 +587,12 @@ else()
         list(APPEND MINIZIP_DEF -DMZ_ZIP_NO_CRYPTO)
     endif()
 
+    # Iconv and ICU are mutually exclusive; ICU takes precedence
+    if (MZ_ICU AND MZ_ICONV)
+        message(STATUS "ICU enabled, disabling Iconv")
+        set(MZ_ICONV OFF)
+    endif()
+
     # Iconv is only necessary when it is not already built-in
     # FindIconv requires cmake 3.11 or higher
     if (MZ_ICONV)
@@ -606,6 +613,24 @@ else()
 
         set(MZ_ICONV OFF)
     endif()
+
+    if (MZ_ICU)
+        find_package(ICU COMPONENTS uc QUIET)
+    endif()
+
+    if(ICU_FOUND)
+        message(STATUS "Using ICU ${ICU_VERSION}")
+
+        list(APPEND MINIZIP_DEF -DHAVE_ICU)
+        list(APPEND MINIZIP_DEP_PKG ICU)
+        list(APPEND MINIZIP_LIB ICU::uc)
+        set(PC_PRIVATE_LIBS "${PC_PRIVATE_LIBS} -licuuc")
+    else()
+        if(MZ_ICU)
+            message(STATUS "ICU library not found")
+            set(MZ_ICU OFF)
+        endif()
+    endif()
 endif()
 
 # Setup predefined macros
@@ -1056,6 +1081,7 @@ add_feature_info(MZ_OPENSSL MZ_OPENSSL "Enables OpenSSL for encryption")
 add_feature_info(MZ_LIBBSD MZ_LIBBSD "Builds with libbsd crypto random")
 # Character conversion options
 add_feature_info(MZ_ICONV MZ_ICONV "Enables iconv string encoding conversion library")
+add_feature_info(MZ_ICU MZ_ICU "Enables ICU string encoding conversion library")
 # Code generation options
 add_feature_info(MZ_COMPRESS_ONLY MZ_COMPRESS_ONLY "Only support compression")
 add_feature_info(MZ_DECOMPRESS_ONLY MZ_DECOMPRESS_ONLY "Only support decompression")

README.md

@@ -82,6 +82,7 @@ cmake --build build
 | MZ_OPENSSL          | Enables OpenSSL encryption                                     |     UNIX      |
 | MZ_LIBBSD           | Builds with libbsd crypto random                               |     UNIX      |
 | MZ_ICONV            | Enables iconv encoding conversion                              |      ON       |
+| MZ_ICU              | Enables ICU encoding conversion                                |      OFF      |
 | MZ_COMPRESS_ONLY    | Only support compression                                       |      OFF      |
 | MZ_DECOMPRESS_ONLY  | Only support decompression                                     |      OFF      |
 | MZ_FILE32_API       | Builds using posix 32-bit file api                             |      OFF      |

mz_os_posix.c

@@ -18,6 +18,9 @@
 #if defined(HAVE_ICONV)
 #  include <iconv.h>
 #endif
+#if defined(HAVE_ICU)
+#  include <unicode/ucnv.h>
+#endif
 #include <string.h>
 #include <sys/types.h>
 #include <sys/stat.h>
@@ -88,6 +91,50 @@ char *mz_os_utf8_string_create(const char *string, int32_t encoding) {
 
     return string_utf8;
 }
+#elif defined(HAVE_ICU)
+char *mz_os_utf8_string_create(const char *string, int32_t encoding) {
+    char string_encoding[16];
+    const char *from_encoding = NULL;
+    int32_t string_length = 0;
+    int32_t string_utf8_size = 0;
+    char *string_utf8 = NULL;
+    int32_t result = 0;
+    UErrorCode status = U_ZERO_ERROR;
+
+    if (!string || encoding <= 0)
+        return NULL;
+
+    if (encoding == MZ_ENCODING_UTF8)
+        from_encoding = "UTF-8";
+    else if (encoding == MZ_ENCODING_CODEPAGE_437)
+        from_encoding = "ibm-437";
+    else if (encoding == MZ_ENCODING_CODEPAGE_932)
+        from_encoding = "windows-932-2000";
+    else if (encoding == MZ_ENCODING_CODEPAGE_936)
+        from_encoding = "windows-936-2000";
+    else if (encoding == MZ_ENCODING_CODEPAGE_950)
+        from_encoding = "windows-950-2000";
+    else {
+        snprintf(string_encoding, sizeof(string_encoding), "windows-%" PRId32 "-2000", encoding);
+        from_encoding = string_encoding;
+    }
+
+    string_length = (int32_t)strlen(string);
+    string_utf8_size = string_length * 4 + 1;
+    string_utf8 = (char *)calloc(string_utf8_size, sizeof(char));
+
+    if (!string_utf8)
+        return NULL;
+
+    result = ucnv_convert("UTF-8", from_encoding, string_utf8, string_utf8_size, string, string_length, &status);
+
+    if (U_FAILURE(status) || result < 0) {
+        free(string_utf8);
+        string_utf8 = NULL;
+    }
+
+    return string_utf8;
+}
 #else
 char *mz_os_utf8_string_create(const char *string, int32_t encoding) {
     return strdup(string);