Merge pull request #165 from ashvardanian/main-dev

Improve: Signed args for `front`/`back`

.github/workflows/build_tools.sh

@@ -10,7 +10,7 @@ COMMON_FLAGS="-DSTRINGZILLA_BUILD_TEST=1 -DSTRINGZILLA_BUILD_BENCHMARK=1 -DSTRIN
 # Compiler specific settings
 case "$COMPILER" in
     "GCC")
-        COMPILER_FLAGS="-DCMAKE_CXX_COMPILER=gcc-12 -DCMAKE_CXX_COMPILER=g++-12"
+        COMPILER_FLAGS="-DCMAKE_CXX_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++"
         ;;
     "LLVM")
         COMPILER_FLAGS="-DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++"

.vscode/settings.json

@@ -124,6 +124,7 @@
     "tparam",
     "TPFLAGS",
     "unigram",
+    "unpoison",
     "usecases",
     "Vardanian",
     "vectorcallfunc",

README.md

@@ -816,7 +816,9 @@ When it's enabled, the _~~subjectively~~_ risky overloads from the Standard will
 using str = sz::string;
 
 str("a:b").front(1) == "a"; // no checks, unlike `substr`
+str("a:b").front(2) == "2"; // take first 2 characters
 str("a:b").back(-1) == "b"; // accepting negative indices
+str("a:b").back(-2) == ":b"; // similar to Python's `"a:b"[-2:]`
 str("a:b").sub(1, -1) == ":"; // similar to Python's `"a:b"[1:-1]`
 str("a:b").sub(-2, -1) == ":"; // similar to Python's `"a:b"[-2:-1]`
 str("a:b").sub(-2, 1) == ""; // similar to Python's `"a:b"[-2:1]`

include/stringzilla/stringzilla.hpp

@@ -1242,27 +1242,31 @@ class basic_string_slice {
      *  @warning The behavior is @b undefined if the position is beyond bounds.
      */
     reference sat(difference_type signed_offset) const noexcept {
-        size_type pos = (signed_offset < 0) ? size() + signed_offset : signed_offset;
+        size_type pos = static_cast<size_type>(signed_offset < 0 ? size() + signed_offset : signed_offset);
         assert(pos < size() && "string_slice::sat(i) out of bounds");
         return start_[pos];
     }
 
     /**
-     *  @brief  The opposite operation to `remove_prefix`, that does no bounds checking.
-     *  @warning The behavior is @b undefined if `n > size()`.
+     *  @brief  The slice that would be dropped by `remove_prefix`, that accepts signed arguments
+     *          and does no bounds checking. Equivalent to Python's `"abc"[:2]` and `"abc"[:-1]`.
+     *  @warning The behavior is @b undefined if `n > size() || n < -size() || n == -0`.
      */
-    string_slice front(size_type n) const noexcept {
-        assert(n <= size() && "string_slice::front(n) out of bounds");
-        return {start_, n};
+    string_slice front(difference_type signed_offset) const noexcept {
+        size_type pos = static_cast<size_type>(signed_offset < 0 ? size() + signed_offset : signed_offset);
+        assert(pos <= size() && "string_slice::front(signed_offset) out of bounds");
+        return {start_, pos};
     }
 
     /**
-     *  @brief  The opposite operation to `remove_prefix`, that does no bounds checking.
-     *  @warning The behavior is @b undefined if `n > size()`.
+     *  @brief  The slice that would be dropped by `remove_suffix`, that accepts signed arguments
+     *          and does no bounds checking. Equivalent to Python's `"abc"[2:]` and `"abc"[-1:]`.
+     *  @warning The behavior is @b undefined if `n > size() || n < -size() || n == -0`.
      */
-    string_slice back(size_type n) const noexcept {
-        assert(n <= size() && "string_slice::back(n) out of bounds");
-        return {start_ + length_ - n, n};
+    string_slice back(difference_type signed_offset) const noexcept {
+        size_type pos = static_cast<size_type>(signed_offset < 0 ? size() + signed_offset : signed_offset);
+        assert(pos <= size() && "string_slice::back(signed_offset) out of bounds");
+        return {start_ + pos, length_ - pos};
     }
 
     /**
@@ -2186,15 +2190,15 @@ class basic_string {
      *  @brief  The opposite operation to `remove_prefix`, that does no bounds checking.
      *  @warning The behavior is @b undefined if `n > size()`.
      */
-    string_view front(size_type n) const noexcept { return view().front(n); }
-    string_span front(size_type n) noexcept { return span().front(n); }
+    string_view front(difference_type n) const noexcept { return view().front(n); }
+    string_span front(difference_type n) noexcept { return span().front(n); }
 
     /**
      *  @brief  The opposite operation to `remove_prefix`, that does no bounds checking.
      *  @warning The behavior is @b undefined if `n > size()`.
      */
-    string_view back(size_type n) const noexcept { return view().back(n); }
-    string_span back(size_type n) noexcept { return span().back(n); }
+    string_view back(difference_type n) const noexcept { return view().back(n); }
+    string_span back(difference_type n) noexcept { return span().back(n); }
 
     /**
      *  @brief  Equivalent to Python's `"abc"[-3:-1]`. Exception-safe, unlike STL's `substr`.

scripts/test.cpp

@@ -190,11 +190,11 @@ static void test_api_readonly() {
     assert(str("hello", 4) == "hell"); // Construct from substring
 
     // Element access.
-    assert(str("test")[0] == 't');
-    assert(str("test").at(1) == 'e');
+    assert(str("rest")[0] == 'r');
+    assert(str("rest").at(1) == 'e');
+    assert(*str("rest").data() == 'r');
     assert(str("front").front() == 'f');
     assert(str("back").back() == 'k');
-    assert(*str("data").data() == 'd');
 
     // Iterators.
     assert(*str("begin").begin() == 'b' && *str("cbegin").cbegin() == 'c');
@@ -567,9 +567,30 @@ template <typename string_type>
 static void test_api_readonly_extensions() {
     using str = string_type;
 
-    // Signed offset lokups and slices.
+    // Signed offset lookups and slices.
     assert(str("hello").sat(0) == 'h');
     assert(str("hello").sat(-1) == 'o');
+    assert(str("rest").sat(1) == 'e');
+    assert(str("rest").sat(-1) == 't');
+    assert(str("rest").sat(-4) == 'r');
+
+    assert(str("front").front() == 'f');
+    assert(str("front").front(1) == "f");
+    assert(str("front").front(2) == "fr");
+    assert(str("front").front(2) == "fr");
+    assert(str("front").front(-2) == "fro");
+    assert(str("front").front(0) == "");
+    assert(str("front").front(5) == "front");
+    assert(str("front").front(-5) == "");
+
+    assert(str("back").back() == 'k');
+    assert(str("back").back(1) == "ack");
+    assert(str("back").back(2) == "ck");
+    assert(str("back").back(-1) == "k");
+    assert(str("back").back(-2) == "ck");
+    assert(str("back").back(-4) == "back");
+    assert(str("back").back(4) == "");
+
     assert(str("hello").sub(1) == "ello");
     assert(str("hello").sub(-1) == "o");
     assert(str("hello").sub(1, 2) == "e");
@@ -1022,7 +1043,7 @@ void test_search_with_misaligned_repetitions(std::string_view haystack_pattern,
                                              std::size_t misalignment) {
     constexpr std::size_t max_repeats = 128;
 
-    // Allocate a buffer to store the haystack with enough padding to misalign it.
+    // Allocate a buffer to store the haystack with enough padding to mis-align it.
     std::size_t haystack_buffer_length = max_repeats * haystack_pattern.size() + 2 * SZ_CACHE_LINE_WIDTH;
     std::vector<char> haystack_buffer(haystack_buffer_length, 'x');
     char *haystack = haystack_buffer.data();