Initial commit: Padre Matos Soares (1932) Bible corpus and pms CLIHEADmaster

1 files changed, 89 insertions, 0 deletions

diff --git a/src/pms_util.c b/src/pms_util.c
new file mode 100644
index 0000000..635fa3c
--- /dev/null
+++ b/src/pms_util.c
@@ -0,0 +1,89 @@
+#include "pms_util.h"
+#include <ctype.h>
+#include <string.h>
+
+static char fold_latin1_c3(unsigned char b2) {
+    if ((b2 >= 0x80 && b2 <= 0x85) || (b2 >= 0xA0 && b2 <= 0xA5)) return 'a';
+    if (b2 == 0x87 || b2 == 0xA7) return 'c';
+    if ((b2 >= 0x88 && b2 <= 0x8B) || (b2 >= 0xA8 && b2 <= 0xAB)) return 'e';
+    if ((b2 >= 0x8C && b2 <= 0x8F) || (b2 >= 0xAC && b2 <= 0xAF)) return 'i';
+    if (b2 == 0x91 || b2 == 0xB1) return 'n';
+    if ((b2 >= 0x92 && b2 <= 0x96) || (b2 >= 0xB2 && b2 <= 0xB6)) return 'o';
+    if ((b2 >= 0x99 && b2 <= 0x9C) || (b2 >= 0xB9 && b2 <= 0xBC)) return 'u';
+    if (b2 == 0x9D || b2 == 0xBD || b2 == 0xBF) return 'y';
+    return 0;
+}
+
+size_t pms_fold_utf8(const char *src, char *dst, size_t dst_size, int *byte_map) {
+    if (!src || !dst || dst_size == 0) {
+        return 0;
+    }
+    size_t i = 0;
+    size_t out = 0;
+    while (src[i] != '\0' && out + 1 < dst_size) {
+        unsigned char c = (unsigned char)src[i];
+        if (c < 0x80) {
+            if (byte_map) byte_map[out] = (int)i;
+            dst[out++] = (char)tolower(c);
+            i += 1;
+        } else if (c == 0xC3 && src[i + 1] != '\0') {
+            unsigned char b2 = (unsigned char)src[i + 1];
+            char folded = fold_latin1_c3(b2);
+            if (byte_map) byte_map[out] = (int)i;
+            dst[out++] = folded ? folded : '?';
+            i += 2;
+        } else if ((c & 0xE0) == 0xC0 && src[i + 1] != '\0') {
+            if (byte_map) byte_map[out] = (int)i;
+            dst[out++] = ' ';
+            i += 2;
+        } else if ((c & 0xF0) == 0xE0 && src[i + 1] != '\0' && src[i + 2] != '\0') {
+            /* Handle curly apostrophes U+2018 / U+2019 (E2 80 98 / E2 80 99) and dashes */
+            unsigned char b2 = (unsigned char)src[i + 1];
+            unsigned char b3 = (unsigned char)src[i + 2];
+            char repl = ' ';
+            if (c == 0xE2 && b2 == 0x80 && (b3 == 0x98 || b3 == 0x99)) {
+                repl = '\'';
+            } else if (c == 0xE2 && b2 == 0x80 && (b3 == 0x93 || b3 == 0x94)) {
+                repl = '-';
+            }
+            if (byte_map) byte_map[out] = (int)i;
+            dst[out++] = repl;
+            i += 3;
+        } else if ((c & 0xF8) == 0xF0 && src[i + 1] != '\0' && src[i + 2] != '\0' && src[i + 3] != '\0') {
+            if (byte_map) byte_map[out] = (int)i;
+            dst[out++] = ' ';
+            i += 4;
+        } else {
+            i += 1;
+        }
+    }
+    if (byte_map) {
+        byte_map[out] = (int)strlen(src);
+    }
+    dst[out] = '\0';
+    return out;
+}
+
+size_t pms_normalize_book_key(const char *src, char *dst, size_t dst_size) {
+    char folded[256];
+    pms_fold_utf8(src, folded, sizeof(folded), NULL);
+    size_t out = 0;
+    for (size_t i = 0; folded[i] != '\0' && out + 1 < dst_size; i++) {
+        unsigned char c = (unsigned char)folded[i];
+        if (isalnum(c)) {
+            dst[out++] = (char)c;
+        }
+    }
+    dst[out] = '\0';
+    return out;
+}
+
+size_t pms_utf8_char_len(const char *s, size_t byte_len) {
+    size_t count = 0;
+    for (size_t i = 0; i < byte_len && s[i] != '\0'; i++) {
+        if (((unsigned char)s[i] & 0xC0) != 0x80) {
+            count++;
+        }
+    }
+    return count;
+}