Fix shrinking of contiguous bytevectors, as from 'get-bytevector-n'.
[bpt/guile.git] / libguile / bytevectors.c
index a969e3b..b210440 100644 (file)
@@ -1,4 +1,4 @@
-/* Copyright (C) 2009, 2010, 2011 Free Software Foundation, Inc.
+/* Copyright (C) 2009, 2010, 2011, 2012, 2014 Free Software Foundation, Inc.
  *
  * This library is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public License
@@ -31,7 +31,6 @@
 #include "libguile/bytevectors.h"
 #include "libguile/strings.h"
 #include "libguile/validate.h"
-#include "libguile/ieee-754.h"
 #include "libguile/arrays.h"
 #include "libguile/array-handle.h"
 #include "libguile/uniform.h"
 /* Bytevector type.  */
 
 #define SCM_BYTEVECTOR_HEADER_BYTES            \
-  (SCM_BYTEVECTOR_HEADER_SIZE * sizeof (SCM))
+  (SCM_BYTEVECTOR_HEADER_SIZE * sizeof (scm_t_bits))
 
 #define SCM_BYTEVECTOR_SET_LENGTH(_bv, _len)            \
   SCM_SET_CELL_WORD_1 ((_bv), (scm_t_bits) (_len))
@@ -285,7 +284,7 @@ scm_i_make_typed_bytevector (size_t len, scm_t_array_element_type element_type)
 /* Return a bytevector of size LEN made up of CONTENTS.  The area pointed to
    by CONTENTS must have been allocated using `scm_gc_malloc ()'.  */
 SCM
-scm_c_take_bytevector (signed char *contents, size_t len)
+scm_c_take_gc_bytevector (signed char *contents, size_t len)
 {
   return make_bytevector_from_buffer (len, contents, SCM_ARRAY_ELEMENT_TYPE_VU8);
 }
@@ -316,10 +315,16 @@ scm_c_shrink_bytevector (SCM bv, size_t c_new_len)
   SCM_BYTEVECTOR_SET_LENGTH (bv, c_new_len);
 
   if (SCM_BYTEVECTOR_CONTIGUOUS_P (bv))
-    new_bv = PTR2SCM (scm_gc_realloc (SCM2PTR (bv),
-                                     c_len + SCM_BYTEVECTOR_HEADER_BYTES,
-                                     c_new_len + SCM_BYTEVECTOR_HEADER_BYTES,
-                                     SCM_GC_BYTEVECTOR));
+    {
+      signed char *c_bv;
+
+      c_bv = scm_gc_realloc (SCM2PTR (bv),
+                            c_len + SCM_BYTEVECTOR_HEADER_BYTES,
+                            c_new_len + SCM_BYTEVECTOR_HEADER_BYTES,
+                            SCM_GC_BYTEVECTOR);
+      new_bv = PTR2SCM (c_bv);
+      SCM_BYTEVECTOR_SET_CONTENTS (new_bv, c_bv + SCM_BYTEVECTOR_HEADER_BYTES);
+    }
   else
     {
       signed char *c_bv;
@@ -462,7 +467,7 @@ SCM_DEFINE (scm_make_bytevector, "make-bytevector", 1, 1, 0,
   signed char c_fill = '\0';
 
   SCM_VALIDATE_UINT_COPY (1, len, c_len);
-  if (fill != SCM_UNDEFINED)
+  if (!scm_is_eq (fill, SCM_UNDEFINED))
     {
       int value;
 
@@ -473,7 +478,7 @@ SCM_DEFINE (scm_make_bytevector, "make-bytevector", 1, 1, 0,
     }
 
   bv = make_bytevector (c_len, SCM_ARRAY_ELEMENT_TYPE_VU8);
-  if (fill != SCM_UNDEFINED)
+  if (!scm_is_eq (fill, SCM_UNDEFINED))
     {
       unsigned i;
       signed char *contents;
@@ -579,9 +584,9 @@ SCM_DEFINE (scm_bytevector_copy_x, "bytevector-copy!", 5, 0, 0,
   if (SCM_UNLIKELY (c_target_start + c_len > c_target_len))
     scm_out_of_range (FUNC_NAME, target_start);
 
-  memcpy (c_target + c_target_start,
-         c_source + c_source_start,
-         c_len);
+  memmove (c_target + c_target_start,
+          c_source + c_source_start,
+          c_len);
 
   return SCM_UNSPECIFIED;
 }
@@ -1100,20 +1105,18 @@ SCM_DEFINE (scm_bytevector_sint_set_x, "bytevector-sint-set!", 5, 0, 0,
                                                                        \
   SCM_VALIDATE_BYTEVECTOR (1, bv);                                     \
   SCM_VALIDATE_SYMBOL (2, endianness);                                 \
-  c_size = scm_to_uint (size);                                         \
+  c_size = scm_to_unsigned_integer (size, 1, (size_t) -1);             \
                                                                        \
   c_len = SCM_BYTEVECTOR_LENGTH (bv);                                  \
-  if (SCM_UNLIKELY (c_len == 0))                                       \
+  if (SCM_UNLIKELY (c_len < c_size))                                   \
     lst = SCM_EOL;                                                     \
-  else if (SCM_UNLIKELY (c_len < c_size))                              \
-    scm_out_of_range (FUNC_NAME, size);                                        \
   else                                                                 \
     {                                                                  \
       const char *c_bv;                                                        \
                                                                        \
       c_bv = (char *) SCM_BYTEVECTOR_CONTENTS (bv);                    \
                                                                        \
-      lst = scm_make_list (scm_from_uint (c_len / c_size),             \
+      lst = scm_make_list (scm_from_size_t (c_len / c_size),           \
                           SCM_UNSPECIFIED);                            \
       for (i = 0, pair = lst;                                          \
           i <= c_len - c_size;                                         \
@@ -1567,6 +1570,18 @@ SCM_DEFINE (scm_bytevector_s64_native_set_x, "bytevector-s64-native-set!",
    Section 2.1 of R6RS-lib (in response to
    http://www.r6rs.org/formal-comments/comment-187.txt).  */
 
+union scm_ieee754_float
+{
+  float f;
+  scm_t_uint32 i;
+};
+
+union scm_ieee754_double
+{
+  double d;
+  scm_t_uint64 i;
+};
+
 
 /* Convert to/from a floating-point number with different endianness.  This
    method is probably not the most efficient but it should be portable.  */
@@ -1575,20 +1590,10 @@ static inline void
 float_to_foreign_endianness (union scm_ieee754_float *target,
                             float source)
 {
-  union scm_ieee754_float src;
-
-  src.f = source;
+  union scm_ieee754_float input;
 
-#ifdef WORDS_BIGENDIAN
-  /* Assuming little endian for both byte and word order.  */
-  target->little_endian.negative = src.big_endian.negative;
-  target->little_endian.exponent = src.big_endian.exponent;
-  target->little_endian.mantissa = src.big_endian.mantissa;
-#else
-  target->big_endian.negative = src.little_endian.negative;
-  target->big_endian.exponent = src.little_endian.exponent;
-  target->big_endian.mantissa = src.little_endian.mantissa;
-#endif
+  input.f = source;
+  target->i = bswap_32 (input.i);
 }
 
 static inline float
@@ -1596,16 +1601,7 @@ float_from_foreign_endianness (const union scm_ieee754_float *source)
 {
   union scm_ieee754_float result;
 
-#ifdef WORDS_BIGENDIAN
-  /* Assuming little endian for both byte and word order.  */
-  result.big_endian.negative = source->little_endian.negative;
-  result.big_endian.exponent = source->little_endian.exponent;
-  result.big_endian.mantissa = source->little_endian.mantissa;
-#else
-  result.little_endian.negative = source->big_endian.negative;
-  result.little_endian.exponent = source->big_endian.exponent;
-  result.little_endian.mantissa = source->big_endian.mantissa;
-#endif
+  result.i = bswap_32 (source->i);
 
   return (result.f);
 }
@@ -1614,22 +1610,10 @@ static inline void
 double_to_foreign_endianness (union scm_ieee754_double *target,
                              double source)
 {
-  union scm_ieee754_double src;
+  union scm_ieee754_double input;
 
-  src.d = source;
-
-#ifdef WORDS_BIGENDIAN
-  /* Assuming little endian for both byte and word order.  */
-  target->little_little_endian.negative  = src.big_endian.negative;
-  target->little_little_endian.exponent  = src.big_endian.exponent;
-  target->little_little_endian.mantissa0 = src.big_endian.mantissa0;
-  target->little_little_endian.mantissa1 = src.big_endian.mantissa1;
-#else
-  target->big_endian.negative  = src.little_little_endian.negative;
-  target->big_endian.exponent  = src.little_little_endian.exponent;
-  target->big_endian.mantissa0 = src.little_little_endian.mantissa0;
-  target->big_endian.mantissa1 = src.little_little_endian.mantissa1;
-#endif
+  input.d = source;
+  target->i = bswap_64 (input.i);
 }
 
 static inline double
@@ -1637,18 +1621,7 @@ double_from_foreign_endianness (const union scm_ieee754_double *source)
 {
   union scm_ieee754_double result;
 
-#ifdef WORDS_BIGENDIAN
-  /* Assuming little endian for both byte and word order.  */
-  result.big_endian.negative  = source->little_little_endian.negative;
-  result.big_endian.exponent  = source->little_little_endian.exponent;
-  result.big_endian.mantissa0 = source->little_little_endian.mantissa0;
-  result.big_endian.mantissa1 = source->little_little_endian.mantissa1;
-#else
-  result.little_little_endian.negative  = source->big_endian.negative;
-  result.little_little_endian.exponent  = source->big_endian.exponent;
-  result.little_little_endian.mantissa0 = source->big_endian.mantissa0;
-  result.little_little_endian.mantissa1 = source->big_endian.mantissa1;
-#endif
+  result.i = bswap_64 (source->i);
 
   return (result.d);
 }
@@ -1907,7 +1880,7 @@ utf_encoding_name (char *name, size_t utf_width, SCM endianness)
   size_t c_strlen, c_utf_len = 0;                                       \
                                                                         \
   SCM_VALIDATE_STRING (1, str);                                         \
-  if (endianness == SCM_UNDEFINED)                                      \
+  if (scm_is_eq (endianness, SCM_UNDEFINED))                            \
     endianness = scm_sym_big;                                           \
   else                                                                  \
     SCM_VALIDATE_SYMBOL (2, endianness);                                \
@@ -1954,33 +1927,15 @@ SCM_DEFINE (scm_string_to_utf8, "string->utf8",
 #define FUNC_NAME s_scm_string_to_utf8
 {
   SCM utf;
-  uint8_t *c_utf;
-  size_t c_strlen, c_utf_len = 0;
+  scm_t_uint8 *c_utf;
+  size_t c_utf_len = 0;
 
   SCM_VALIDATE_STRING (1, str);
 
-  c_strlen = scm_i_string_length (str);
-  if (scm_i_is_narrow_string (str))
-    c_utf = u8_conv_from_encoding ("ISO-8859-1", iconveh_question_mark,
-                                   scm_i_string_chars (str), c_strlen,
-                                   NULL, NULL, &c_utf_len);
-  else
-    {
-      const scm_t_wchar *wbuf = scm_i_string_wide_chars (str);
-      c_utf = u32_to_u8 ((const uint32_t *) wbuf, c_strlen, NULL, &c_utf_len);
-    }
-  if (SCM_UNLIKELY (c_utf == NULL))
-    scm_syserror (FUNC_NAME);
-  else
-    {
-      scm_dynwind_begin (0);
-      scm_dynwind_free (c_utf);
-
-      utf = make_bytevector (c_utf_len, SCM_ARRAY_ELEMENT_TYPE_VU8);
-      memcpy (SCM_BYTEVECTOR_CONTENTS (utf), c_utf, c_utf_len);
-
-      scm_dynwind_end ();
-    }
+  c_utf = (scm_t_uint8 *) scm_to_utf8_stringn (str, &c_utf_len);
+  utf = make_bytevector (c_utf_len, SCM_ARRAY_ELEMENT_TYPE_VU8);
+  memcpy (SCM_BYTEVECTOR_CONTENTS (utf), c_utf, c_utf_len);
+  free (c_utf);
 
   return (utf);
 }
@@ -1997,6 +1952,14 @@ SCM_DEFINE (scm_string_to_utf16, "string->utf16",
 }
 #undef FUNC_NAME
 
+static void
+swap_u32 (scm_t_wchar *vals, size_t len)
+{
+  size_t n;
+  for (n = 0; n < len; n++)
+    vals[n] = bswap_32 (vals[n]);
+}
+
 SCM_DEFINE (scm_string_to_utf32, "string->utf32",
            1, 1, 0,
            (SCM str, SCM endianness),
@@ -2004,7 +1967,21 @@ SCM_DEFINE (scm_string_to_utf32, "string->utf32",
            "encoding of @var{str}.")
 #define FUNC_NAME s_scm_string_to_utf32
 {
-  STRING_TO_UTF (32);
+  SCM bv;
+  scm_t_wchar *wchars;
+  size_t wchar_len, bytes_len;
+
+  wchars = scm_to_utf32_stringn (str, &wchar_len);
+  bytes_len = wchar_len * sizeof (scm_t_wchar);
+  if (!scm_is_eq (SCM_UNBNDP (endianness) ? scm_endianness_big : endianness,
+                  scm_i_native_endianness))
+    swap_u32 (wchars, wchar_len);
+  
+  bv = make_bytevector (bytes_len, SCM_ARRAY_ELEMENT_TYPE_VU8);
+  memcpy (SCM_BYTEVECTOR_CONTENTS (bv), wchars, bytes_len);
+  free (wchars);
+
+  return bv;
 }
 #undef FUNC_NAME
 
@@ -2020,7 +1997,7 @@ SCM_DEFINE (scm_string_to_utf32, "string->utf32",
   size_t c_strlen = 0, c_utf_len = 0;                                  \
                                                                        \
   SCM_VALIDATE_BYTEVECTOR (1, utf);                                    \
-  if (endianness == SCM_UNDEFINED)                                     \
+  if (scm_is_eq (endianness, SCM_UNDEFINED))                            \
     endianness = scm_sym_big;                                          \
   else                                                                 \
     SCM_VALIDATE_SYMBOL (2, endianness);                               \
@@ -2092,26 +2069,56 @@ SCM_DEFINE (scm_utf32_to_string, "utf32->string",
 \f
 /* Bytevectors as generalized vectors & arrays.  */
 
+#define COMPLEX_ACCESSOR_PROLOGUE(_type)                       \
+  size_t c_len, c_index;                                       \
+  char *c_bv;                                                  \
+                                                               \
+  SCM_VALIDATE_BYTEVECTOR (1, bv);                             \
+  c_index = scm_to_size_t (index);                             \
+                                                               \
+  c_len = SCM_BYTEVECTOR_LENGTH (bv);                          \
+  c_bv = (char *) SCM_BYTEVECTOR_CONTENTS (bv);                        \
+                                                               \
+  if (SCM_UNLIKELY (c_index + 2 * sizeof (_type) - 1 >= c_len))        \
+    scm_out_of_range (FUNC_NAME, index);
+
+/* Template for native access to complex numbers of type TYPE.  */
+#define COMPLEX_NATIVE_REF(_type)                                      \
+  SCM result;                                                          \
+                                                                       \
+  COMPLEX_ACCESSOR_PROLOGUE (_type);                                   \
+                                                                       \
+  {                                                                    \
+    _type real, imag;                                                  \
+                                                                       \
+    memcpy (&real, &c_bv[c_index], sizeof (_type));                    \
+    memcpy (&imag, &c_bv[c_index + sizeof (_type)], sizeof (_type));   \
+                                                                       \
+    result = scm_c_make_rectangular (real, imag);                      \
+  }                                                                    \
+                                                                       \
+  return result;
 
 static SCM
-bytevector_ref_c32 (SCM bv, SCM idx)
-{ /* FIXME add some checks */
-  const float *contents = (const float*)SCM_BYTEVECTOR_CONTENTS (bv);
-  size_t i = scm_to_size_t (idx);
-  return scm_c_make_rectangular (contents[i/4], contents[i/4 + 1]);
+bytevector_ref_c32 (SCM bv, SCM index)
+#define FUNC_NAME "bytevector_ref_c32"
+{
+  COMPLEX_NATIVE_REF (float);
 }
+#undef FUNC_NAME
 
 static SCM
-bytevector_ref_c64 (SCM bv, SCM idx)
-{ /* FIXME add some checks */
-  const double *contents = (const double*)SCM_BYTEVECTOR_CONTENTS (bv);
-  size_t i = scm_to_size_t (idx);
-  return scm_c_make_rectangular (contents[i/8], contents[i/8 + 1]);
+bytevector_ref_c64 (SCM bv, SCM index)
+#define FUNC_NAME "bytevector_ref_c64"
+{
+  COMPLEX_NATIVE_REF (double);
 }
+#undef FUNC_NAME
 
 typedef SCM (*scm_t_bytevector_ref_fn)(SCM, SCM);
 
-const scm_t_bytevector_ref_fn bytevector_ref_fns[SCM_ARRAY_ELEMENT_TYPE_LAST + 1] = 
+static const scm_t_bytevector_ref_fn
+bytevector_ref_fns[SCM_ARRAY_ELEMENT_TYPE_LAST + 1] =
 {
   NULL, /* SCM */
   NULL, /* CHAR */
@@ -2143,24 +2150,36 @@ bv_handle_ref (scm_t_array_handle *h, size_t index)
   return ref_fn (h->array, byte_index);
 }
 
-/* FIXME add checks!!! */
-static SCM
-bytevector_set_c32 (SCM bv, SCM idx, SCM val)
-{ float *contents = (float*)SCM_BYTEVECTOR_CONTENTS (bv);
-  size_t i = scm_to_size_t (idx);
-  contents[i/4] = scm_c_real_part (val);
-  contents[i/4 + 1] = scm_c_imag_part (val);
+/* Template for native modification of complex numbers of type TYPE.  */
+#define COMPLEX_NATIVE_SET(_type)                                      \
+  COMPLEX_ACCESSOR_PROLOGUE (_type);                                   \
+                                                                       \
+  {                                                                    \
+    _type real, imag;                                                  \
+    real = scm_c_real_part (value);                                    \
+    imag = scm_c_imag_part (value);                                    \
+                                                                       \
+    memcpy (&c_bv[c_index], &real, sizeof (_type));                    \
+    memcpy (&c_bv[c_index + sizeof (_type)], &imag, sizeof (_type));   \
+  }                                                                    \
+                                                                       \
   return SCM_UNSPECIFIED;
+
+static SCM
+bytevector_set_c32 (SCM bv, SCM index, SCM value)
+#define FUNC_NAME "bytevector_set_c32"
+{
+  COMPLEX_NATIVE_SET (float);
 }
+#undef FUNC_NAME
 
 static SCM
-bytevector_set_c64 (SCM bv, SCM idx, SCM val)
-{ double *contents = (double*)SCM_BYTEVECTOR_CONTENTS (bv);
-  size_t i = scm_to_size_t (idx);
-  contents[i/8] = scm_c_real_part (val);
-  contents[i/8 + 1] = scm_c_imag_part (val);
-  return SCM_UNSPECIFIED;
+bytevector_set_c64 (SCM bv, SCM index, SCM value)
+#define FUNC_NAME "bytevector_set_c64"
+{
+  COMPLEX_NATIVE_SET (double);
 }
+#undef FUNC_NAME
 
 typedef SCM (*scm_t_bytevector_set_fn)(SCM, SCM, SCM);