Numerous updates:

* Add makefile to install data * Reference text is now formatted * UTF-8 string processing
2015-08-06 14:31:38 -04:00
parent af38b1eafc
commit cfc0ba7e9a
13 changed files with 269 additions and 7 deletions
--- a/src/utf8.c
+++ b/src/utf8.c
@@ -0,0 +1,124 @@
+#include <stdlib.h>
+#include <string.h>
+#include "utf8.h"
+
+utf8iterator*
+utf8NewIterator(const char* txt) {
+    utf8iterator* result = malloc(sizeof(utf8iterator));
+    memset(result, 0, sizeof(utf8iterator));
+    result->txt = txt;
+    return result;
+}
+
+uint32_t utf8CharAt(const utf8iterator* iter) {
+    int byteIndex = iter->byteIndex;
+    if ((iter->txt[byteIndex] & 0x80) == 0) {
+        return iter->txt[byteIndex];
+    } else if (((iter->txt[byteIndex] & 0xe0) == 0xc0) &&
+               ((iter->txt[byteIndex + 1] & 0xc0) == 0x80)) {
+        return (uint32_t)(iter->txt[byteIndex + 1] & 0x3f) |
+               ((uint32_t)(iter->txt[byteIndex] & 0x1f) << 6);
+    } else if (((iter->txt[byteIndex] & 0xf0) == 0xe0) && 
+               ((iter->txt[byteIndex + 1] & 0xc0) == 0x80) && 
+               ((iter->txt[byteIndex + 2] & 0xc0) == 0x80)) {
+        return (uint32_t)(iter->txt[byteIndex + 2] & 0x3f) |
+               ((uint32_t)(iter->txt[byteIndex + 1] & 0x3f) << 6) |
+               ((uint32_t)(iter->txt[byteIndex] & 0xf) << 12);
+    } else if (((iter->txt[byteIndex] & 0xf8) == 0xf0) && 
+               ((iter->txt[byteIndex + 1] & 0xc0) == 0x80) && 
+               ((iter->txt[byteIndex + 2] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 3] & 0xc0) == 0x80)) {
+        return (uint32_t)(iter->txt[byteIndex + 3] & 0x3f) |
+               ((uint32_t)(iter->txt[byteIndex + 2] & 0x3f) << 6) |
+               ((uint32_t)(iter->txt[byteIndex + 1] & 0x3f) << 12) |
+               ((uint32_t)(iter->txt[byteIndex] & 0x07) << 18);
+    } else if (((iter->txt[byteIndex] & 0xfc) == 0xf8) &&
+               ((iter->txt[byteIndex + 1] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 2] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 3] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 4] & 0xc0) == 0x80)) {
+        return (uint32_t)(iter->txt[byteIndex + 4] & 0x3f) |
+               ((uint32_t)(iter->txt[byteIndex + 3] & 0x3f) << 6) |
+               ((uint32_t)(iter->txt[byteIndex + 2] & 0x3f) << 12) |
+               ((uint32_t)(iter->txt[byteIndex + 1] & 0x3f) << 18) |
+               ((uint32_t)(iter->txt[byteIndex] & 0x03) << 24);
+
+    } else if (((iter->txt[byteIndex] & 0xf7) == 0xfc) &&
+               ((iter->txt[byteIndex + 1]) & 0xc0 == 0x80) &&
+               ((iter->txt[byteIndex + 2] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 3] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 4] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 5] & 0xc0) == 0x80)) {
+        return (uint32_t)(iter->txt[byteIndex + 5] & 0x3f) |
+               ((uint32_t)(iter->txt[byteIndex + 4] & 0x3f) << 6) |
+               ((uint32_t)(iter->txt[byteIndex + 3] & 0x3f) << 12) |
+               ((uint32_t)(iter->txt[byteIndex + 2] & 0x3f) << 18) |
+               ((uint32_t)(iter->txt[byteIndex + 1] & 0x3f) << 24) |
+               ((uint32_t)(iter->txt[byteIndex] & 0x01) << 30);
+    }
+    return 0;
+}
+
+static int
+_next_offset(const utf8iterator* iter) {
+    int byteIndex = iter->byteIndex;
+    if ((iter->txt[byteIndex] & 0x80) == 0) {
+        return 1;
+    } else if (((iter->txt[byteIndex] & 0xe0) == 0xc0) &&
+               ((iter->txt[byteIndex + 1] & 0xc0) == 0x80)) {
+        return 2;
+    } else if (((iter->txt[byteIndex] & 0xf0) == 0xe0) && 
+               ((iter->txt[byteIndex + 1] & 0xc0) == 0x80) && 
+               ((iter->txt[byteIndex + 2] & 0xc0) == 0x80)) {
+        return 3;
+    } else if (((iter->txt[byteIndex] & 0xf8) == 0xf0) && 
+               ((iter->txt[byteIndex + 1] & 0xc0) == 0x80) && 
+               ((iter->txt[byteIndex + 2] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 3] & 0xc0) == 0x80)) {
+        return 4;
+    } else if (((iter->txt[byteIndex] & 0xfc) == 0xf8) &&
+               ((iter->txt[byteIndex + 1] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 2] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 3] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 4] & 0xc0) == 0x80)) {
+        return 5;
+    } else if (((iter->txt[byteIndex] & 0xf7) == 0xfc) &&
+               ((iter->txt[byteIndex + 1]) & 0xc0 == 0x80) &&
+               ((iter->txt[byteIndex + 2] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 3] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 4] & 0xc0) == 0x80) &&
+               ((iter->txt[byteIndex + 5] & 0xc0) == 0x80)) {
+        return 6;
+    } else return -1;
+}
+
+int
+utf8Advance(utf8iterator* iter) {
+    int next_offset = _next_offset(iter);
+    if (next_offset > 0) {
+        iter->byteIndex += next_offset;
+        return ++iter->logicalIndex;
+    } else return -1;
+}
+
+int
+utf8CopyEncodedCharAt(const utf8iterator* iter, size_t bufsize, char* dest) {
+    size_t sz = _next_offset(iter);
+    if (sz + 1 > bufsize) {
+        return 0;
+    } else {
+        memset(dest, 0, bufsize);
+        memcpy(dest, iter->txt + iter->byteIndex, sz);
+        return 1;
+    }
+}
+
+void
+utf8ResetIterator(utf8iterator* iter) {
+    iter->logicalIndex = iter->byteIndex = 0;
+}
+
+void
+utf8FreeIterator(utf8iterator* iter) {
+    free(iter);
+}