X-Git-Url: http://mj.ucw.cz/gitweb/?a=blobdiff_plain;f=charset%2Fcharconv.c;h=bf431dd1b93569fdf68ade711848801c4e18aedf;hb=d5fdccbecd2acde9a6e067b54fcd69b02f31a820;hp=c5d3717b65cfa8c62281dc21e765cb077ebf83ad;hpb=2926a57424946281a78b9cc5b8098b1cff39e45a;p=libucw.git

diff --git a/charset/charconv.c b/charset/charconv.c
index c5d3717b..bf431dd1 100644
--- a/charset/charconv.c
+++ b/charset/charconv.c
@@ -1,13 +1,16 @@
 /*
- *	Character Set Conversion Library 1.1
+ *	Character Set Conversion Library 1.2
  *
- *	(c) 1998--2001 Martin Mares <mj@ucw.cz>
+ *	(c) 1998--2004 Martin Mares <mj@ucw.cz>
+ *	(c) 2007 Pavel Charvat <pchar@ucw.cz>
  *
  *	This software may be freely distributed and used according to the terms
- *	of the GNU General Public License.
+ *	of the GNU Lesser General Public License.
  */
 
-#include "lib/lib.h"
+#include "ucw/lib.h"
+#include "ucw/unicode.h"
+#include "ucw/unaligned.h"
 #include "charset/charconv.h"
 #include "charset/chartable.h"
 
@@ -26,175 +29,328 @@ conv_none(struct conv_context *c)
   return CONV_SOURCE_END | CONV_DEST_END | CONV_SKIP;
 }
 
+enum state {
+  CLEAN,
+  SINGLE_WRITE,
+  SEQ_WRITE,
+  UTF8_READ,
+  UTF8_WRITE_START,
+  UTF8_WRITE_CONT,
+  UTF16_BE_WRITE,
+  UTF16_LE_WRITE,
+  UTF16_BE_READ,
+  UTF16_BE_READ_1,
+  UTF16_BE_READ_2,
+  UTF16_BE_READ_3,
+  UTF16_LE_READ,
+  UTF16_LE_READ_1,
+  UTF16_LE_READ_2,
+  UTF16_LE_READ_3,
+};
+
 static int
-conv_from_utf8(struct conv_context *c)
+conv_slow(struct conv_context *c)
 {
-  unsigned short *x_to_out = c->x_to_out;
   const unsigned char *s = c->source;
   const unsigned char *se = c->source_end;
   unsigned char *d = c->dest;
   unsigned char *de = c->dest_end;
-  unsigned char *strings = string_table - 0x100;
-  unsigned int counter, code, cc;
-
-  if (c->state)
-    goto go_slow;
 
-  while (s < se)			/* Optimized for speed, beware of spaghetti code */
+  switch (c->state)
     {
-      cc = *s++;
-      if (cc < 0x80)
-	code = cc;
-      else if (cc >= 0xc0)
+    case SINGLE_WRITE:
+      if (d >= de)
+	goto cde;
+      *d++ = c->code;
+      break;
+    case SEQ_WRITE:
+seq:
+      while (c->remains)
 	{
-	  if (s + 6 > se)
-	    goto go_slow_1;
-	  if (cc < 0xe0)
-	    {
-	      if ((s[0] & 0xc0) != 0x80)
-		goto nocode;
-	      code = cc & 0x1f;
-	      code = (code << 6) | (*s++ & 0x3f);
-	    }
-	  else if (cc < 0xf0)
-	    {
-	      if ((s[0] & 0xc0) != 0x80 || (s[1] & 0xc0) != 0x80)
-		goto nocode;
-	      code = cc & 0x0f;
-	      code = (code << 6) | (*s++ & 0x3f);
-	      code = (code << 6) | (*s++ & 0x3f);
-	    }
-	  else if (cc < 0xfc)
+	  if (d >= de)
+	    goto cde;
+	  *d++ = *c->string_at++;
+	  c->remains--;
+	}
+      break;
+
+    case UTF8_READ:
+      while (c->remains)
+	{
+	  if (s >= se)
+	    goto cse;
+	  if ((*s & 0xc0) != 0x80)
 	    {
-	      while (cc & 0x80)
-		{
-		  if ((*s++ & 0xc0) != 0x80)
-		    break;
-		  cc <<= 1;
-		}
-	      goto nocode;
+	      c->code = 0xfffd;
+	      break;
 	    }
-	  else
-	    goto nocode;
+	  c->code = (c->code << 6) | (*s++ & 0x3f);
+	  c->remains--;
+	}
+      if (c->code >= 0x10000)
+	c->code = 0xfffd;
+got_char:
+      c->source = s;
+      c->state = 0;
+      return -1;
+
+    /* Writing of UTF-8 */
+    case UTF8_WRITE_START:
+      if (d >= de)
+	goto cde;
+      if (c->code < 0x80)
+	{
+	  *d++ = c->code;
+	  break;
+	}
+      else if (c->code < 0x800)
+	{
+	  *d++ = 0xc0 | (c->code >> 6);
+	  c->code <<= 10;
+	  c->remains = 1;
 	}
       else
 	{
-	nocode:
-	  code = 0xfffd;
+	  *d++ = 0xe0 | (c->code >> 12);
+	  c->code <<= 4;
+	  c->remains = 2;
 	}
-    uni_again:
-      code = x_to_out[uni_to_x[code >> 8U][code & 0xff]];
-    code_again:
-      if (code < 0x100)
+      c->code &= 0xffff;
+      c->state = UTF8_WRITE_CONT;
+      /* fall-thru */
+    case UTF8_WRITE_CONT:
+      while (c->remains)
 	{
 	  if (d >= de)
-	    goto dend;
-	  *d++ = code;
+	    goto cde;
+	  *d++ = 0x80 | (c->code >> 10);
+	  c->code <<= 6;
+	  c->remains--;
 	}
+      break;
+
+    /* Writing of UTF-16BE */
+    case UTF16_BE_WRITE:
+      {
+	void *p = &c->code;
+	c->string_at = p;
+	uns code = c->code;
+	c->string_at = p;
+	if (code < 0xd800 || code - 0xe000 < 0x2000)
+	  {}
+	else if ((code -= 0x10000) < 0x100000)
+	  {
+	    put_u16_be(p, 0xd800 | (code >> 10));
+	    put_u16_be(p + 2, 0xdc00 | (code & 0x3ff));
+	    c->remains = 4;
+	    c->state = SEQ_WRITE;
+	    goto seq;
+	  }
+	else
+	  code = UNI_REPLACEMENT;
+	put_u16_be(p, code);
+	c->remains = 2;
+        c->state = SEQ_WRITE;
+	goto seq;
+      }
+
+    /* Writing of UTF-16LE */
+    case UTF16_LE_WRITE:
+      {
+	void *p = &c->code;
+	c->string_at = p;
+	uns code = c->code;
+	c->string_at = p;
+	if (code < 0xd800 || code - 0xe000 < 0x2000)
+	  {}
+	else if ((code -= 0x10000) < 0x100000)
+	  {
+	    put_u16_le(p, 0xd800 | (code >> 10));
+	    put_u16_le(p + 2, 0xdc00 | (code & 0x3ff));
+	    c->remains = 4;
+            c->state = SEQ_WRITE;
+	  }
+	else
+	  code = UNI_REPLACEMENT;
+        put_u16_le(p, code);
+        c->remains = 2;
+        c->state = SEQ_WRITE;
+	goto seq;
+      }
+
+    /* Reading of UTF16-BE */
+    case UTF16_BE_READ:
+      if (s >= se)
+	goto cse;
+      c->code = *s++;
+      c->state = UTF16_BE_READ_1;
+      /* fall-thru */
+    case UTF16_BE_READ_1:
+      if (s >= se)
+	goto cse;
+      c->code = (c->code << 8) | *s++;
+      if (c->code - 0xd800 >= 0x800)
+        goto got_char;
+      c->code = (c->code - 0xd800) << 10;
+      c->state = UTF16_BE_READ_2;
+      /* fall-thru */
+    case UTF16_BE_READ_2:
+      if (s >= se)
+	goto cse;
+      if (*s - 0xdc >= 4)
+	c->code = ~0U;
       else
-	{
-	  unsigned char *k = strings + code;
-	  unsigned int len = *k++;
+	c->code |= (*s - 0xdc) << 8;
+      s++;
+      c->state = UTF16_BE_READ_3;
+      /* fall-thru */
+    case UTF16_BE_READ_3:
+      if (s >= se)
+	goto cse;
+      if ((int)c->code >= 0)
+	c->code += 0x10000 + *s;
+      else
+	c->code = UNI_REPLACEMENT;
+      s++;
+      goto got_char;
 
-	  if (d + len > de)
-	    goto dend;
-	  while (len--)
-	    *d++ = *k++;
-	}
+    /* Reading of UTF16-LE */
+    case UTF16_LE_READ:
+      if (s >= se)
+	goto cse;
+      c->code = *s++;
+      c->state = UTF16_LE_READ_1;
+      /* fall-thru */
+    case UTF16_LE_READ_1:
+      if (s >= se)
+	goto cse;
+      c->code |= *s++ << 8;
+      if (c->code - 0xd800 >= 0x800)
+        goto got_char;
+      c->code = (c->code - 0xd800) << 10;
+      c->state = UTF16_LE_READ_2;
+      /* fall-thru */
+    case UTF16_LE_READ_2:
+      if (s >= se)
+	goto cse;
+      c->code |= *s++;
+      c->state = UTF16_LE_READ_3;
+      /* fall-thru */
+    case UTF16_LE_READ_3:
+      if (s >= se)
+	goto cse;
+      if (*s - 0xdc < 4)
+	c->code += 0x10000 + ((*s - 0xdc) << 8);
+      else
+	c->code = UNI_REPLACEMENT;
+      s++;
+      goto got_char;
+
+    default:
+      ASSERT(0);
     }
-  c->state = 0;
-send_noreset:
   c->source = s;
   c->dest = d;
-  return CONV_SOURCE_END;
+  c->state = 0;
+  return 0;
 
-dend:
-  c->state = ~0;
-  c->value = code;
+ cse:
   c->source = s;
+  return CONV_SOURCE_END;
+
+ cde:
   c->dest = d;
   return CONV_DEST_END;
+}
 
-go_slow:
-  code = c->value;
-  counter = c->state;
-  if (counter == ~0U)
-    goto code_again;
-  goto go_slow_2;
+/* Generate inlined routines */
 
-go_slow_1:
-  if (cc < 0xe0) { code = cc & 0x1f; counter = 1; }
-  else if (cc < 0xf0) { code = cc & 0x0f; counter = 2; }
-  else
-    {
-      code = ~0;
-      if (cc < 0xf8) counter = 3;
-      else if (cc < 0xfc) counter = 4;
-      else if (cc < 0xfe) counter = 5;
-      else goto nocode;
-    }
-go_slow_2:
-  while (counter)
-    {
-      if (s >= se)
-	{
-	  c->state = counter;
-	  c->value = code;
-	  goto send_noreset;
-	}
-      if ((*s & 0xc0) != 0x80)
-	goto nocode;
-      code = (code << 6) | (*s++ & 0x3f);
-      counter--;
-    }
-  if (code >= 0x10000)
-    goto nocode;
-  goto uni_again;
+static int
+conv_std_to_utf8(struct conv_context *c)
+{
+#define CONV_READ_STD
+#define CONV_WRITE_UTF8
+#include "charset/charconv-gen.h"
 }
 
 static int
-conv_to_utf8(struct conv_context *c)
+conv_utf8_to_std(struct conv_context *c)
 {
-  unsigned short *in_to_x = c->in_to_x;
-  const unsigned char *s = c->source;
-  const unsigned char *se = c->source_end;
-  unsigned char *d = c->dest;
-  unsigned char *de = c->dest_end;
+#define CONV_READ_UTF8
+#define CONV_WRITE_STD
+#include "charset/charconv-gen.h"
+}
 
-  while (s < se)
-    {
-      unsigned int code = x_to_uni[in_to_x[*s]];
-      if (code < 0x80)
-	{
-	  if (d >= de)
-	    goto dend;
-	  *d++ = code;
-	}
-      else if (code < 0x800)
-	{
-	  if (d + 2 > de)
-	    goto dend;
-	  *d++ = 0xc0 | (code >> 6);
-	  *d++ = 0x80 | (code & 0x3f);
-	}
-      else
-	{
-	  if (d + 3 > de)
-	    goto dend;
-	  *d++ = 0xc0 | (code >> 12);
-	  *d++ = 0x80 | ((code >> 6) & 0x3f);
-	  *d++ = 0x80 | (code & 0x3f);
-	}
-      s++;
-    }
-  c->source = s;
-  c->dest = d;
-  return CONV_SOURCE_END;
+static int
+conv_std_to_utf16_be(struct conv_context *c)
+{
+#define CONV_READ_STD
+#define CONV_WRITE_UTF16_BE
+#include "charset/charconv-gen.h"
+}
 
-dend:
-  c->source = s;
-  c->dest = d;
-  return CONV_DEST_END;
+static int
+conv_utf16_be_to_std(struct conv_context *c)
+{
+#define CONV_READ_UTF16_BE
+#define CONV_WRITE_STD
+#include "charset/charconv-gen.h"
+}
+
+static int
+conv_std_to_utf16_le(struct conv_context *c)
+{
+#define CONV_READ_STD
+#define CONV_WRITE_UTF16_LE
+#include "charset/charconv-gen.h"
+}
+
+static int
+conv_utf16_le_to_std(struct conv_context *c)
+{
+#define CONV_READ_UTF16_LE
+#define CONV_WRITE_STD
+#include "charset/charconv-gen.h"
+}
+
+static int
+conv_utf8_to_utf16_be(struct conv_context *c)
+{
+#define CONV_READ_UTF8
+#define CONV_WRITE_UTF16_BE
+#include "charset/charconv-gen.h"
+}
+
+static int
+conv_utf16_be_to_utf8(struct conv_context *c)
+{
+#define CONV_READ_UTF16_BE
+#define CONV_WRITE_UTF8
+#include "charset/charconv-gen.h"
+}
+
+static int
+conv_utf8_to_utf16_le(struct conv_context *c)
+{
+#define CONV_READ_UTF8
+#define CONV_WRITE_UTF16_LE
+#include "charset/charconv-gen.h"
+}
+
+static int
+conv_utf16_le_to_utf8(struct conv_context *c)
+{
+#define CONV_READ_UTF16_LE
+#define CONV_WRITE_UTF8
+#include "charset/charconv-gen.h"
+}
+
+static int
+conv_utf16_be_to_utf16_le(struct conv_context *c)
+{
+#define CONV_READ_UTF16_BE
+#define CONV_WRITE_UTF16_LE
+#include "charset/charconv-gen.h"
 }
 
 static int
@@ -202,28 +358,33 @@ conv_standard(struct conv_context *c)
 {
   unsigned short *in_to_x = c->in_to_x;
   unsigned short *x_to_out = c->x_to_out;
-  const unsigned char *s = c->source;
-  const unsigned char *se = c->source_end;
-  unsigned char *d = c->dest;
-  unsigned char *de = c->dest_end;
-  unsigned char *strings = string_table - 0x100;
+  const unsigned char *s, *se;
+  unsigned char *d, *de, *k;
+  unsigned int len, e;
+
+  if (unlikely(c->state))
+    goto slow;
 
+main:
+  s = c->source;
+  se = c->source_end;
+  d = c->dest;
+  de = c->dest_end;
   while (s < se)
     {
       unsigned int code = x_to_out[in_to_x[*s]];
       if (code < 0x100)
 	{
-	  if (d >= de)
+	  if (unlikely(d >= de))
 	    goto dend;
 	  *d++ = code;
 	}
       else
 	{
-	  unsigned char *k = strings + code;
-	  unsigned int len = *k++;
-
-	  if (d + len > de)
-	    goto dend;
+	  k = string_table + code - 0x100;
+	  len = *k++;
+	  if (unlikely(d + len > de))
+	    goto dend_str;
 	  while (len--)
 	    *d++ = *k++;
 	}
@@ -237,24 +398,49 @@ dend:
   c->source = s;
   c->dest = d;
   return CONV_DEST_END;
+
+dend_str:
+  c->source = s;
+  c->dest = d;
+  c->state = SEQ_WRITE;
+  c->string_at = k;
+  c->remains = len;
+slow:
+  e = conv_slow(c);
+  if (e)
+    return e;
+  goto main;
 }
 
 void
 conv_set_charset(struct conv_context *c, int src, int dest)
 {
+  c->source_charset = src;
+  c->dest_charset = dest;
   if (src == dest)
-    c->convert = conv_none;
+    {
+      c->convert = conv_none;
+      c->in_to_x = NULL;
+      c->x_to_out = NULL;
+    }
   else
     {
-      c->convert = conv_standard;
-      if (src == CONV_CHARSET_UTF8)
-	c->convert = conv_from_utf8;
-      else
-	c->in_to_x = input_to_x[src];
-      if (dest == CONV_CHARSET_UTF8)
-	c->convert = conv_to_utf8;
-      else
-	c->x_to_out = x_to_output[dest];
+      static uns lookup[] = {
+	[CONV_CHARSET_UTF8] = 1,
+	[CONV_CHARSET_UTF16_BE] = 2,
+	[CONV_CHARSET_UTF16_LE] = 3,
+      };
+      static int (*tab[4][4])(struct conv_context *c) = {
+	{ conv_standard,	conv_std_to_utf8,	conv_std_to_utf16_be,	conv_std_to_utf16_le },
+	{ conv_utf8_to_std,	conv_none,		conv_utf8_to_utf16_be,	conv_utf8_to_utf16_le },
+	{ conv_utf16_be_to_std,	conv_utf16_be_to_utf8,	conv_none,		conv_utf16_be_to_utf16_le },
+	{ conv_utf16_le_to_std,	conv_utf16_le_to_utf8,	conv_utf16_be_to_utf16_le,	conv_none },
+      };
+      uns src_idx = ((uns)src < ARRAY_SIZE(lookup)) ? lookup[src] : 0;
+      uns dest_idx = ((uns)dest < ARRAY_SIZE(lookup)) ? lookup[dest] : 0;
+      c->convert = tab[src_idx][dest_idx];
+      c->in_to_x = src_idx ? NULL : input_to_x[src];
+      c->x_to_out = dest_idx ? NULL : x_to_output[dest];
     }
   c->state = 0;
 }
@@ -276,3 +462,18 @@ conv_x_count(void)
 {
   return sizeof(x_to_uni) / sizeof(x_to_uni[0]);
 }
+
+int
+conv_in_to_ucs(struct conv_context *c, unsigned int y)
+{
+  return x_to_uni[c->in_to_x[y]];
+}
+
+int conv_ucs_to_out(struct conv_context *c, unsigned int ucs)
+{
+  uns x = uni_to_x[ucs >> 8U][ucs & 0xff];
+  if (x == 256 || c->x_to_out[x] >= 256)
+    return -1;
+  else
+    return c->x_to_out[x];
+}