cii: Add atom

cii/.gitignore

@@ -0,0 +1,2 @@
+/cmake-build-debug
+/cmake-build-release

cii/CMakeLists.txt

@@ -0,0 +1,19 @@
+cmake_minimum_required(VERSION 3.20)
+project(cii C)
+
+set(CMAKE_C_STANDARD 11)
+
+set(SOURCE_FILES
+    src/atom.c
+    )
+set(HEADER_FILES
+    include/cii/atom.h
+    )
+
+add_library(cii STATIC
+    ${SOURCE_FILES}
+    ${HEADER_FILES}
+    )
+target_include_directories(cii
+    PUBLIC include
+    )

cii/include/cii/atom.h

@@ -0,0 +1,70 @@
+// Copyright (c) 2023 Xu Shaohua <[email protected]>. All rights reserved.
+// Use of this source is governed by GNU General Public License
+// that can be found in the LICENSE file.
+
+#ifndef CII_ATOM_H_
+#define CII_ATOM_H_
+
+/**
+ * An atom is a pointer to a unique, immutable sequence of zero or more
+ * arbitrary bytes. Most atoms are null-terminated strings, but a pointer to
+ * any sequence of bytes can be an atom.
+ *
+ * There is only a single occurrence(instance) of any atom, which is why it's
+ * called an atom.
+ *
+ * Two atoms are identical if they point to the same location. Comparing two byte
+ * sequences for equality by simply comparing pointers is one of the advantages
+ * of atoms. Another advantage is that using atoms saves space because there's
+ * only one occurrence of each sequence.
+ *
+ * Once an atom is created, it exists for the duration of the client's execution.
+ * An atom is always terminated with a null character.
+ */
+
+#include <stdint.h>
+#include <stddef.h>
+
+/**
+ * It adds a copy of the sequence to the table of atoms, if necessary, and
+ * returns the pointer to that atom.
+ *
+ * Never returns null pointer.
+ *
+ * @param str sequence of bytes, shall not be null pointer.
+ * @param len number of bytes
+ * @return the atom, which is a pointer to the copy of the sequence in the atom
+ * table.
+ */
+const char* atom_new(const char* str, size_t len);
+
+/**
+ * Like |atom_new|, it caters to the common use of character strings as atoms.
+ *
+ * Adds a copy of that string to the atom table, if necessary, and returns the
+ * pointer to that atom.
+ * @param str null-terminated string
+ * @return the atom, which is a pointer to the copy of the sequence in the atom
+ * table.
+ */
+const char* atom_string(const char* str);
+
+/**
+ * Returns the atom for the string representation of the integer |n|.
+ * @param n integer value to be represented as string.
+ * @return the atom, which is a pointer to the copy of the integer in the atom
+ * table.
+ */
+const char* atom_int(int64_t n);
+
+/**
+ * Returns length of its atom argument.
+ *
+ * Will raise error if |str| is not a pointer to an atom.
+ *
+ * @param str pointer to an atom.
+ * @return the string length of an atom.
+ */
+size_t atom_length(const char* str);
+
+#endif  // CII_ATOM_H_

cii/src/atom.c

@@ -0,0 +1,56 @@
+// Copyright (c) 2023 Xu Shaohua <[email protected]>. All rights reserved.
+// Use of this source is governed by GNU General Public License
+// that can be found in the LICENSE file.
+
+#include "cii/atom.h"
+
+#include <assert.h>
+#include <limits.h>
+#include <string.h>
+
+struct atom_s {
+  struct atom_s* link;
+  size_t len;
+  char* str;
+};
+
+typedef struct atom_s atom_t;
+
+const size_t kBucketCap = 2048;
+static atom_t g_buckets[kBucketCap];
+
+const char* atom_string(const char* str) {
+  assert(str != NULL);
+  return atom_new(str, strlen(str));
+}
+
+const char* atom_int(int64_t n) {
+  uint64_t m;
+  if (n == INT64_MIN) {
+    m = INT64_MAX + 1UL;
+  } else if (n < 0) {
+    m = -n;
+  } else {
+    m = n;
+  }
+
+  // 43 characters can hold decimal representation of 128-bit integers.
+  const size_t kBufLen = 43;
+  char buf[kBufLen];
+  char* end_buf = buf + kBufLen;
+  char* s = end_buf;
+  while (m > 0) {
+    s -= 1;
+    *s = m % 10 + '0';
+    m /= 10;
+  }
+  if (n < 0) {
+    s -= 1;
+    *s = '-';
+  }
+  assert(s >= buf);
+
+  return atom_new(s, end_buf - s);
+}
+
+