\f
/* Standard ports --- current input, output, error, and more(!). */
-static SCM cur_inport_fluid = 0;
-static SCM cur_outport_fluid = 0;
-static SCM cur_errport_fluid = 0;
-static SCM cur_loadport_fluid = 0;
+static SCM cur_inport_fluid = SCM_BOOL_F;
+static SCM cur_outport_fluid = SCM_BOOL_F;
+static SCM cur_errport_fluid = SCM_BOOL_F;
+static SCM cur_loadport_fluid = SCM_BOOL_F;
SCM_DEFINE (scm_current_input_port, "current-input-port", 0, 0, 0,
(),
"returns the @dfn{standard input} in Unix and C terminology.")
#define FUNC_NAME s_scm_current_input_port
{
- if (cur_inport_fluid)
+ if (scm_is_true (cur_inport_fluid))
return scm_fluid_ref (cur_inport_fluid);
else
return SCM_BOOL_F;
"Unix and C terminology.")
#define FUNC_NAME s_scm_current_output_port
{
- if (cur_outport_fluid)
+ if (scm_is_true (cur_outport_fluid))
return scm_fluid_ref (cur_outport_fluid);
else
return SCM_BOOL_F;
"@dfn{standard error} in Unix and C terminology).")
#define FUNC_NAME s_scm_current_error_port
{
- if (cur_errport_fluid)
+ if (scm_is_true (cur_errport_fluid))
return scm_fluid_ref (cur_errport_fluid);
else
return SCM_BOOL_F;
"have no effect as far as @var{port-for-each} is concerned.")
#define FUNC_NAME s_scm_port_for_each
{
+ SCM ports;
+
SCM_VALIDATE_PROC (1, proc);
- scm_c_port_for_each ((void (*)(void*,SCM))scm_call_1, proc);
+ /* Copy out the port table as a list so that we get strong references
+ to all the values. */
+ scm_i_pthread_mutex_lock (&scm_i_port_table_mutex);
+ ports = scm_internal_hash_fold (collect_keys, NULL,
+ SCM_EOL, scm_i_port_weak_hash);
+ scm_i_pthread_mutex_unlock (&scm_i_port_table_mutex);
+
+ for (; scm_is_pair (ports); ports = scm_cdr (ports))
+ if (SCM_PORTP (SCM_CAR (ports)))
+ scm_call_1 (proc, SCM_CAR (ports));
+
return SCM_UNSPECIFIED;
}
#undef FUNC_NAME
switch (c)
{
case '\a':
+ case EOF:
break;
case '\b':
SCM_DECCOL (port);
return codepoint;
}
-/* Read a codepoint from PORT and return it in *CODEPOINT. Fill BUF
- with the byte representation of the codepoint in PORT's encoding, and
- set *LEN to the length in bytes of that representation. Return 0 on
- success and an errno value on error. */
+/* Read a UTF-8 sequence from PORT. On success, return 0 and set
+ *CODEPOINT to the codepoint that was read, fill BUF with its UTF-8
+ representation, and set *LEN to the length in bytes. Return
+ `EILSEQ' on error. */
static int
-get_codepoint (SCM port, scm_t_wchar *codepoint,
- char buf[SCM_MBCHAR_BUF_SIZE], size_t *len)
+get_utf8_codepoint (SCM port, scm_t_wchar *codepoint,
+ scm_t_uint8 buf[SCM_MBCHAR_BUF_SIZE], size_t *len)
{
+#define ASSERT_NOT_EOF(b) \
+ if (SCM_UNLIKELY ((b) == EOF)) \
+ goto invalid_seq
+#define CONSUME_PEEKED_BYTE() \
+ pt->read_pos++
+
+ int byte;
+ scm_t_port *pt;
+
+ *len = 0;
+ pt = SCM_PTAB_ENTRY (port);
+
+ byte = scm_get_byte_or_eof (port);
+ if (byte == EOF)
+ {
+ *codepoint = EOF;
+ return 0;
+ }
+
+ buf[0] = (scm_t_uint8) byte;
+ *len = 1;
+
+ if (buf[0] <= 0x7f)
+ /* 1-byte form. */
+ *codepoint = buf[0];
+ else if (buf[0] >= 0xc2 && buf[0] <= 0xdf)
+ {
+ /* 2-byte form. */
+ byte = scm_peek_byte_or_eof (port);
+ ASSERT_NOT_EOF (byte);
+
+ if (SCM_UNLIKELY ((byte & 0xc0) != 0x80))
+ goto invalid_seq;
+
+ CONSUME_PEEKED_BYTE ();
+ buf[1] = (scm_t_uint8) byte;
+ *len = 2;
+
+ *codepoint = ((scm_t_wchar) buf[0] & 0x1f) << 6UL
+ | (buf[1] & 0x3f);
+ }
+ else if ((buf[0] & 0xf0) == 0xe0)
+ {
+ /* 3-byte form. */
+ byte = scm_peek_byte_or_eof (port);
+ ASSERT_NOT_EOF (byte);
+
+ if (SCM_UNLIKELY ((byte & 0xc0) != 0x80
+ || (buf[0] == 0xe0 && byte < 0xa0)
+ || (buf[0] == 0xed && byte > 0x9f)))
+ goto invalid_seq;
+
+ CONSUME_PEEKED_BYTE ();
+ buf[1] = (scm_t_uint8) byte;
+ *len = 2;
+
+ byte = scm_peek_byte_or_eof (port);
+ ASSERT_NOT_EOF (byte);
+
+ if (SCM_UNLIKELY ((byte & 0xc0) != 0x80))
+ goto invalid_seq;
+
+ CONSUME_PEEKED_BYTE ();
+ buf[2] = (scm_t_uint8) byte;
+ *len = 3;
+
+ *codepoint = ((scm_t_wchar) buf[0] & 0x0f) << 12UL
+ | ((scm_t_wchar) buf[1] & 0x3f) << 6UL
+ | (buf[2] & 0x3f);
+ }
+ else if (buf[0] >= 0xf0 && buf[0] <= 0xf4)
+ {
+ /* 4-byte form. */
+ byte = scm_peek_byte_or_eof (port);
+ ASSERT_NOT_EOF (byte);
+
+ if (SCM_UNLIKELY (((byte & 0xc0) != 0x80)
+ || (buf[0] == 0xf0 && byte < 0x90)
+ || (buf[0] == 0xf4 && byte > 0x8f)))
+ goto invalid_seq;
+
+ CONSUME_PEEKED_BYTE ();
+ buf[1] = (scm_t_uint8) byte;
+ *len = 2;
+
+ byte = scm_peek_byte_or_eof (port);
+ ASSERT_NOT_EOF (byte);
+
+ if (SCM_UNLIKELY ((byte & 0xc0) != 0x80))
+ goto invalid_seq;
+
+ CONSUME_PEEKED_BYTE ();
+ buf[2] = (scm_t_uint8) byte;
+ *len = 3;
+
+ byte = scm_peek_byte_or_eof (port);
+ ASSERT_NOT_EOF (byte);
+
+ if (SCM_UNLIKELY ((byte & 0xc0) != 0x80))
+ goto invalid_seq;
+
+ CONSUME_PEEKED_BYTE ();
+ buf[3] = (scm_t_uint8) byte;
+ *len = 4;
+
+ *codepoint = ((scm_t_wchar) buf[0] & 0x07) << 18UL
+ | ((scm_t_wchar) buf[1] & 0x3f) << 12UL
+ | ((scm_t_wchar) buf[2] & 0x3f) << 6UL
+ | (buf[3] & 0x3f);
+ }
+ else
+ goto invalid_seq;
+
+ return 0;
+
+ invalid_seq:
+ /* Here we could choose the consume the faulty byte when it's not a
+ valid starting byte, but it's not a requirement. What Section 3.9
+ of Unicode 6.0.0 mandates, though, is to not consume a byte that
+ would otherwise be a valid starting byte. */
+
+ return EILSEQ;
+
+#undef CONSUME_PEEKED_BYTE
+#undef ASSERT_NOT_EOF
+}
+
+/* Likewise, read a byte sequence from PORT, passing it through its
+ input conversion descriptor. */
+static int
+get_iconv_codepoint (SCM port, scm_t_wchar *codepoint,
+ char buf[SCM_MBCHAR_BUF_SIZE], size_t *len)
+{
+ scm_t_port *pt;
int err, byte_read;
size_t bytes_consumed, output_size;
char *output;
scm_t_uint8 utf8_buf[SCM_MBCHAR_BUF_SIZE];
- scm_t_port *pt = SCM_PTAB_ENTRY (port);
- if (SCM_UNLIKELY (pt->input_cd == (iconv_t) -1))
- /* Initialize the conversion descriptors. */
- scm_i_set_port_encoding_x (port, pt->encoding);
+ pt = SCM_PTAB_ENTRY (port);
for (output_size = 0, output = (char *) utf8_buf,
bytes_consumed = 0, err = 0;
output_size = sizeof (utf8_buf) - output_left;
}
- if (SCM_UNLIKELY (err != 0))
+ if (SCM_UNLIKELY (output_size == 0))
+ /* An unterminated sequence. */
+ err = EILSEQ;
+ else if (SCM_LIKELY (err == 0))
{
- /* Reset the `iconv' state. */
- iconv (pt->input_cd, NULL, NULL, NULL, NULL);
+ /* Convert the UTF8_BUF sequence to a Unicode code point. */
+ *codepoint = utf8_to_codepoint (utf8_buf, output_size);
+ *len = bytes_consumed;
+ }
- if (pt->ilseq_handler == SCM_ICONVEH_QUESTION_MARK)
- {
- *codepoint = '?';
- err = 0;
- }
+ return err;
+}
- /* Fail when the strategy is SCM_ICONVEH_ERROR or
- SCM_ICONVEH_ESCAPE_SEQUENCE (the latter doesn't make sense for
- input encoding errors.) */
- }
+/* Read a codepoint from PORT and return it in *CODEPOINT. Fill BUF
+ with the byte representation of the codepoint in PORT's encoding, and
+ set *LEN to the length in bytes of that representation. Return 0 on
+ success and an errno value on error. */
+static int
+get_codepoint (SCM port, scm_t_wchar *codepoint,
+ char buf[SCM_MBCHAR_BUF_SIZE], size_t *len)
+{
+ int err;
+ scm_t_port *pt = SCM_PTAB_ENTRY (port);
+
+ if (pt->input_cd == (iconv_t) -1)
+ /* Initialize the conversion descriptors, if needed. */
+ scm_i_set_port_encoding_x (port, pt->encoding);
+
+ /* FIXME: In 2.1, add a flag to determine whether a port is UTF-8. */
+ if (pt->input_cd == (iconv_t) -1)
+ err = get_utf8_codepoint (port, codepoint, (scm_t_uint8 *) buf, len);
else
- /* Convert the UTF8_BUF sequence to a Unicode code point. */
- *codepoint = utf8_to_codepoint (utf8_buf, output_size);
+ err = get_iconv_codepoint (port, codepoint, buf, len);
if (SCM_LIKELY (err == 0))
update_port_lf (*codepoint, port);
-
- *len = bytes_consumed;
+ else if (pt->ilseq_handler == SCM_ICONVEH_QUESTION_MARK)
+ {
+ *codepoint = '?';
+ err = 0;
+ update_port_lf (*codepoint, port);
+ }
return err;
}
if (encoding == NULL)
encoding = "ISO-8859-1";
- pt->encoding = scm_gc_strdup (encoding, "port");
+ if (pt->encoding != encoding)
+ pt->encoding = scm_gc_strdup (encoding, "port");
- if (SCM_CELL_WORD_0 (port) & SCM_RDNG)
+ /* If ENCODING is UTF-8, then no conversion descriptor is opened
+ because we do I/O ourselves. This saves 100+ KiB for each
+ descriptor. */
+ if (strcmp (encoding, "UTF-8"))
{
- /* Open an input iconv conversion descriptor, from ENCODING
- to UTF-8. We choose UTF-8, not UTF-32, because iconv
- implementations can typically convert from anything to
- UTF-8, but not to UTF-32 (see
- <http://lists.gnu.org/archive/html/bug-libunistring/2010-09/msg00007.html>). */
- new_input_cd = iconv_open ("UTF-8", encoding);
- if (new_input_cd == (iconv_t) -1)
- goto invalid_encoding;
- }
+ if (SCM_CELL_WORD_0 (port) & SCM_RDNG)
+ {
+ /* Open an input iconv conversion descriptor, from ENCODING
+ to UTF-8. We choose UTF-8, not UTF-32, because iconv
+ implementations can typically convert from anything to
+ UTF-8, but not to UTF-32 (see
+ <http://lists.gnu.org/archive/html/bug-libunistring/2010-09/msg00007.html>). */
+ new_input_cd = iconv_open ("UTF-8", encoding);
+ if (new_input_cd == (iconv_t) -1)
+ goto invalid_encoding;
+ }
- if (SCM_CELL_WORD_0 (port) & SCM_WRTNG)
- {
- new_output_cd = iconv_open (encoding, "UTF-8");
- if (new_output_cd == (iconv_t) -1)
+ if (SCM_CELL_WORD_0 (port) & SCM_WRTNG)
{
- if (new_input_cd != (iconv_t) -1)
- iconv_close (new_input_cd);
- goto invalid_encoding;
+ new_output_cd = iconv_open (encoding, "UTF-8");
+ if (new_output_cd == (iconv_t) -1)
+ {
+ if (new_input_cd != (iconv_t) -1)
+ iconv_close (new_input_cd);
+ goto invalid_encoding;
+ }
}
}
if (scm_is_false (port))
{
/* Set the default encoding for future ports. */
- if (!scm_conversion_strategy
+ if (!scm_conversion_strategy_init
|| !scm_is_fluid (SCM_VARIABLE_REF (scm_conversion_strategy)))
scm_misc_error (NULL, "tried to set conversion strategy fluid before it is initialized",
SCM_EOL);
#include "libguile/ports.x"
/* Use Latin-1 as the default port encoding. */
- SCM_VARIABLE_SET (default_port_encoding_var, scm_make_fluid ());
- scm_fluid_set_x (SCM_VARIABLE_REF (default_port_encoding_var), SCM_BOOL_F);
+ SCM_VARIABLE_SET (default_port_encoding_var,
+ scm_make_fluid_with_default (SCM_BOOL_F));
scm_port_encoding_init = 1;
- SCM_VARIABLE_SET (scm_conversion_strategy, scm_make_fluid ());
- scm_fluid_set_x (SCM_VARIABLE_REF (scm_conversion_strategy),
- scm_from_int ((int) SCM_FAILED_CONVERSION_QUESTION_MARK));
+ SCM_VARIABLE_SET (scm_conversion_strategy,
+ scm_make_fluid_with_default
+ (scm_from_int ((int) SCM_FAILED_CONVERSION_QUESTION_MARK)));
scm_conversion_strategy_init = 1;
}
HCoop / bpt / guile.git / blobdiff