[RFC] Break down the static library into multiple files

CMakeLists.txt

@@ -613,7 +613,18 @@ target_include_directories(
 
 # Helper function used for `stringzilla_shared` and `stringzilla_bare` targets
 function (define_stringzilla_shared target)
-    add_library(${target} SHARED c/stringzilla.c)
+    add_library(${target} SHARED
+        c/compare.c
+        c/memory.c
+        c/find.c
+        c/hash.c
+        c/utf8.c
+        c/utf8_case.c
+        c/utf8_word.c
+        c/sort.c
+        c/intersect.c
+        c/stringzilla.c
+    )
     add_library(${PROJECT_NAME}::${target} ALIAS ${target})
 
     set_target_properties(

c/compare.c

@@ -0,0 +1,27 @@
+/**
+*   @file       compare.c
+ *  @brief      StringZilla compare functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+// Now include compare.h which will use our overridden macros
+#include <stringzilla/compare.h>

c/find.c

@@ -0,0 +1,27 @@
+/**
+ *  @file       find.c
+ *  @brief      StringZilla find functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Now override the macros AFTER types.h has defined them
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+#include <stringzilla/find.h>

c/hash.c

@@ -0,0 +1,27 @@
+/**
+ *  @file       hash.c
+ *  @brief      StringZilla hash functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+
+#include <stringzilla/hash.h>

c/intersect.c

@@ -0,0 +1,27 @@
+/**
+ *  @file       intersect.c
+ *  @brief      StringZilla intersect functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+
+#include <stringzilla/intersect.h>

c/memory.c

@@ -0,0 +1,27 @@
+/**
+ *  @file       memory.c
+ *  @brief      StringZilla memory functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+
+#include <stringzilla/memory.h>

c/sort.c

@@ -0,0 +1,27 @@
+/**
+ *  @file       sort.c
+ *  @brief      StringZilla sort functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+
+#include <stringzilla/sort.h>

c/utf8.c

@@ -0,0 +1,27 @@
+/**
+ *  @file       utf8.c
+ *  @brief      StringZilla UTF-8 functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+
+#include <stringzilla/utf8.h>

c/utf8_case.c

@@ -0,0 +1,27 @@
+/**
+ *  @file       utf8_case.c
+ *  @brief      StringZilla UTF-8 case functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Now override the macros AFTER types.h has defined them
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+#include <stringzilla/utf8_case.h>

c/utf8_word.c

@@ -0,0 +1,27 @@
+/**
+ *  @file       utf8_word.c
+ *  @brief      StringZilla UTF-8 word functions compiled into library
+ *  @author     Raul Marin
+ *  @date       February 5, 2026
+ */
+
+#ifdef SZ_DYNAMIC_DISPATCH
+#undef SZ_DYNAMIC_DISPATCH
+#endif
+#define SZ_DYNAMIC_DISPATCH 0
+
+// Include types.h first to let it define the macros
+#include <stringzilla/types.h>
+
+// Use weak symbols to allow multiple definitions (linker will merge them)
+#undef SZ_PUBLIC
+#undef SZ_C_INLINE
+#undef SZ_INTERNAL
+#define SZ_C_INLINE __attribute__((weak))
+#define SZ_PUBLIC __attribute__((weak))
+#define SZ_INTERNAL static inline
+#undef SZ_DYNAMIC
+#define SZ_DYNAMIC static inline
+
+
+#include <stringzilla/utf8_word.h>

include/stringzilla/compare.h

@@ -120,6 +120,8 @@ SZ_PUBLIC sz_ordering_t sz_order_neon(sz_cptr_t a, sz_size_t a_length, sz_cptr_t
 
 #pragma endregion // Core API
 
+#if !SZ_DYNAMIC_DISPATCH
+
 /* SSE implementation of the string search algorithms for Westmere processors and newer.
  *  Very minimalistic (compared to AVX-512), but still faster than the serial implementation.
  */
@@ -516,7 +518,6 @@ SZ_PUBLIC sz_ordering_t sz_order_sve(sz_cptr_t a, sz_size_t a_length, sz_cptr_t
  *  To override this behavior and precompile all backends - set `SZ_DYNAMIC_DISPATCH` to 1.
  */
 #pragma region Compile Time Dispatching
-#if !SZ_DYNAMIC_DISPATCH
 
 SZ_DYNAMIC sz_bool_t sz_equal(sz_cptr_t a, sz_cptr_t b, sz_size_t length) {
 #if SZ_USE_SKYLAKE
@@ -546,9 +547,10 @@ SZ_DYNAMIC sz_ordering_t sz_order(sz_cptr_t a, sz_size_t a_length, sz_cptr_t b,
 #endif
 }
 
-#endif            // !SZ_DYNAMIC_DISPATCH
 #pragma endregion // Compile Time Dispatching
 
+#endif            // !SZ_DYNAMIC_DISPATCH
+
 #ifdef __cplusplus
 }
 #endif // __cplusplus

include/stringzilla/find.h

@@ -240,6 +240,8 @@ SZ_PUBLIC sz_cptr_t sz_find_delimiters_utf8_ice(sz_cptr_t text, sz_size_t length
 SZ_PUBLIC sz_cptr_t sz_find_delimiters_utf8_neon(sz_cptr_t text, sz_size_t length, sz_size_t *matched_length);
 #endif
 
+#if !SZ_DYNAMIC_DISPATCH
+
 #pragma endregion // Core API
 
 #pragma region Helper Shortcuts
@@ -2005,7 +2007,6 @@ SZ_PUBLIC sz_cptr_t sz_find_sve(sz_cptr_t h, sz_size_t h_length, sz_cptr_t n, sz
  *  To override this behavior and precompile all backends - set `SZ_DYNAMIC_DISPATCH` to 1.
  */
 #pragma region Compile Time Dispatching
-#if !SZ_DYNAMIC_DISPATCH
 
 #pragma region Core Functionality
 
@@ -2096,9 +2097,10 @@ SZ_DYNAMIC sz_cptr_t sz_rfind_byteset(sz_cptr_t text, sz_size_t length, sz_bytes
 }
 
 #pragma endregion
-#endif            // !SZ_DYNAMIC_DISPATCH
 #pragma endregion // Compile Time Dispatching
 
+#endif            // !SZ_DYNAMIC_DISPATCH
+
 #ifdef __cplusplus
 }
 #endif // __cplusplus

include/stringzilla/hash.h

@@ -323,6 +323,15 @@ SZ_PUBLIC void sz_sha256_state_update_goldmont(sz_sha256_state_t *state, sz_cptr
 /** @copydoc sz_sha256_state_digest */
 SZ_PUBLIC void sz_sha256_state_digest_goldmont(sz_sha256_state_t const *state, sz_u8_t digest[sz_at_least_(32)]);
 
+/** @copydoc sz_sha256_state_init */
+SZ_PUBLIC void sz_sha256_state_init_serial(sz_sha256_state_t *state);
+
+/** @copydoc sz_sha256_state_update */
+SZ_PUBLIC void sz_sha256_state_update_serial(sz_sha256_state_t *state, sz_cptr_t text, sz_size_t length);
+
+/** @copydoc sz_sha256_state_digest */
+SZ_PUBLIC void sz_sha256_state_digest_serial(sz_sha256_state_t const *state, sz_u8_t digest[sz_at_least_(32)]);
+
 #endif
 
 #if SZ_USE_HASWELL
@@ -459,6 +468,8 @@ SZ_PUBLIC sz_u64_t sz_hash_state_digest_sve2(sz_hash_state_t const *state);
 
 #pragma endregion // Core API
 
+#if !SZ_DYNAMIC_DISPATCH
+
 #pragma region Helper Methods
 
 /**
@@ -4151,7 +4162,6 @@ SZ_PUBLIC void sz_fill_random_sve2(sz_ptr_t text, sz_size_t length, sz_u64_t non
  *  To override this behavior and precompile all backends - set `SZ_DYNAMIC_DISPATCH` to 1.
  */
 #pragma region Compile Time Dispatching
-#if !SZ_DYNAMIC_DISPATCH
 
 SZ_DYNAMIC sz_u64_t sz_bytesum(sz_cptr_t text, sz_size_t length) {
 #if SZ_USE_ICE
@@ -4287,9 +4297,10 @@ SZ_DYNAMIC void sz_sha256_state_digest(sz_sha256_state_t const *state, sz_u8_t d
 #endif
 }
 
-#endif            // !SZ_DYNAMIC_DISPATCH
 #pragma endregion // Compile Time Dispatching
 
+#endif            // !SZ_DYNAMIC_DISPATCH
+
 #ifdef __cplusplus
 }
 #endif // __cplusplus