/* Copyright (C) 1995, 1996, 1997, 1999, 2000, 2001, 2003, 2004, 2006,
- * 2007, 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc.
+ * 2007, 2008, 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
#endif
#include <stdio.h>
-#include <ctype.h>
#include <string.h>
#include <unistd.h>
#include <unicase.h>
#include <unictype.h>
+#include <c-strcase.h>
+#include <c-ctype.h>
#include "libguile/_scm.h"
#include "libguile/bytevectors.h"
#include "libguile/hashtab.h"
#include "libguile/hash.h"
#include "libguile/ports.h"
+#include "libguile/ports-internal.h"
#include "libguile/fports.h"
#include "libguile/root.h"
#include "libguile/strings.h"
SCM_SYMBOL (scm_keyword_postfix, "postfix");
SCM_SYMBOL (sym_nil, "nil");
-scm_t_option scm_read_opts[] = {
- { SCM_OPTION_BOOLEAN, "copy", 0,
- "Copy source code expressions." },
- { SCM_OPTION_BOOLEAN, "positions", 1,
- "Record positions of source code expressions." },
- { SCM_OPTION_BOOLEAN, "case-insensitive", 0,
- "Convert symbols to lower case."},
- { SCM_OPTION_SCM, "keywords", (scm_t_bits) SCM_BOOL_F_BITS,
- "Style of keyword recognition: #f, 'prefix or 'postfix."},
- { SCM_OPTION_BOOLEAN, "r6rs-hex-escapes", 0,
- "Use R6RS variable-length character and string hex escapes."},
- { SCM_OPTION_BOOLEAN, "square-brackets", 1,
- "Treat `[' and `]' as parentheses, for R6RS compatibility."},
- { SCM_OPTION_BOOLEAN, "hungry-eol-escapes", 0,
- "In strings, consume leading whitespace after an escaped end-of-line."},
- { 0, },
-};
+/* SRFI-105 curly infix expression support */
+SCM_SYMBOL (sym_nfx, "$nfx$");
+SCM_SYMBOL (sym_bracket_list, "$bracket-list$");
+SCM_SYMBOL (sym_bracket_apply, "$bracket-apply$");
+
+scm_t_option scm_read_opts[] =
+ {
+ { SCM_OPTION_BOOLEAN, "copy", 0,
+ "Copy source code expressions." },
+ { SCM_OPTION_BOOLEAN, "positions", 1,
+ "Record positions of source code expressions." },
+ { SCM_OPTION_BOOLEAN, "case-insensitive", 0,
+ "Convert symbols to lower case."},
+ { SCM_OPTION_SCM, "keywords", (scm_t_bits) SCM_BOOL_F_BITS,
+ "Style of keyword recognition: #f, 'prefix or 'postfix."},
+ { SCM_OPTION_BOOLEAN, "r6rs-hex-escapes", 0,
+ "Use R6RS variable-length character and string hex escapes."},
+ { SCM_OPTION_BOOLEAN, "square-brackets", 1,
+ "Treat `[' and `]' as parentheses, for R6RS compatibility."},
+ { SCM_OPTION_BOOLEAN, "hungry-eol-escapes", 0,
+ "In strings, consume leading whitespace after an escaped end-of-line."},
+ { SCM_OPTION_BOOLEAN, "curly-infix", 0,
+ "Support SRFI-105 curly infix expressions."},
+ { SCM_OPTION_BOOLEAN, "r7rs-symbols", 0,
+ "Support R7RS |...| symbol notation."},
+ { 0, },
+ };
/* Internal read options structure. This is initialized by 'scm_read'
- from the global read options, and a pointer is passed down to all
- helper functions. */
-enum t_keyword_style {
- KEYWORD_STYLE_HASH_PREFIX,
- KEYWORD_STYLE_PREFIX,
- KEYWORD_STYLE_POSTFIX
-};
+ from the global and per-port read options, and a pointer is passed
+ down to all helper functions. */
-struct t_read_opts {
+enum t_keyword_style
+ {
+ KEYWORD_STYLE_HASH_PREFIX,
+ KEYWORD_STYLE_PREFIX,
+ KEYWORD_STYLE_POSTFIX
+ };
+
+struct t_read_opts
+{
enum t_keyword_style keyword_style;
unsigned int copy_source_p : 1;
unsigned int record_positions_p : 1;
unsigned int r6rs_escapes_p : 1;
unsigned int square_brackets_p : 1;
unsigned int hungry_eol_escapes_p : 1;
+ unsigned int curly_infix_p : 1;
+ unsigned int neoteric_p : 1;
+ unsigned int r7rs_symbols_p : 1;
};
typedef struct t_read_opts scm_t_read_opts;
-/* Initialize OPTS from the global read options. */
-static void
-init_read_options (scm_t_read_opts *opts)
-{
- SCM val;
- int x;
-
- val = SCM_PACK (SCM_KEYWORD_STYLE);
- if (scm_is_eq (val, scm_keyword_prefix))
- x = KEYWORD_STYLE_PREFIX;
- else if (scm_is_eq (val, scm_keyword_postfix))
- x = KEYWORD_STYLE_POSTFIX;
- else
- x = KEYWORD_STYLE_HASH_PREFIX;
- opts->keyword_style = x;
-
-#define RESOLVE_BOOLEAN_OPTION(NAME, name) \
- (opts->name = !!SCM_ ## NAME)
-
- RESOLVE_BOOLEAN_OPTION (COPY_SOURCE_P, copy_source_p);
- RESOLVE_BOOLEAN_OPTION (RECORD_POSITIONS_P, record_positions_p);
- RESOLVE_BOOLEAN_OPTION (CASE_INSENSITIVE_P, case_insensitive_p);
- RESOLVE_BOOLEAN_OPTION (R6RS_ESCAPES_P, r6rs_escapes_p);
- RESOLVE_BOOLEAN_OPTION (SQUARE_BRACKETS_P, square_brackets_p);
- RESOLVE_BOOLEAN_OPTION (HUNGRY_EOL_ESCAPES_P, hungry_eol_escapes_p);
-
-#undef RESOLVE_BOOLEAN_OPTION
-}
-
/*
Give meaningful error messages for errors
/* The maximum size of Scheme character names. */
#define READER_CHAR_NAME_MAX_SIZE 50
+/* The maximum size of reader directive names. */
+#define READER_DIRECTIVE_NAME_MAX_SIZE 50
+
/* `isblank' is only in C99. */
#define CHAR_IS_BLANK_(_chr) \
#define CHAR_IS_DELIMITER(c) \
(CHAR_IS_R5RS_DELIMITER (c) \
- || (((c) == ']' || (c) == '[') && opts->square_brackets_p))
+ || (((c) == ']' || (c) == '[') && (opts->square_brackets_p \
+ || opts->curly_infix_p)) \
+ || (((c) == '}' || (c) == '{') && opts->curly_infix_p))
/* Exponent markers, as defined in section 7.1.1 of R5RS, ``Lexical
Structure''. */
{
int chr;
- chr = scm_get_byte_or_eof (port);
+ chr = scm_get_byte_or_eof_unlocked (port);
if (chr == EOF)
return 0;
else if (CHAR_IS_DELIMITER (chr))
{
- scm_unget_byte (chr, port);
+ scm_unget_byte_unlocked (chr, port);
return 0;
}
else
{
scm_t_wchar c;
while (1)
- switch (c = scm_getc (port))
+ switch (c = scm_getc_unlocked (port))
{
case EOF:
goteof:
case ';':
lp:
- switch (c = scm_getc (port))
+ switch (c = scm_getc_unlocked (port))
{
case EOF:
goto goteof;
break;
case '#':
- switch (c = scm_getc (port))
+ switch (c = scm_getc_unlocked (port))
{
case EOF:
eoferr = "read_sharp";
}
/* fall through */
default:
- scm_ungetc (c, port);
+ scm_ungetc_unlocked (c, port);
return '#';
}
break;
{
int c;
SCM tmp, tl, ans = SCM_EOL;
- const int terminating_char = ((chr == '[') ? ']' : ')');
+ const int curly_list_p = (chr == '{') && opts->curly_infix_p;
+ const int terminating_char = ((chr == '{') ? '}'
+ : ((chr == '[') ? ']'
+ : ')'));
/* Need to capture line and column numbers here. */
long line = SCM_LINUM (port);
if (terminating_char == c)
return SCM_EOL;
- scm_ungetc (c, port);
+ scm_ungetc_unlocked (c, port);
tmp = scm_read_expression (port, opts);
/* Note that it is possible for scm_read_expression to return
{
SCM new_tail;
- if (c == ')' || (c == ']' && opts->square_brackets_p))
+ if (c == ')' || (c == ']' && opts->square_brackets_p)
+ || ((c == '}' || c == ']') && opts->curly_infix_p))
scm_i_input_error (FUNC_NAME, port,
"in pair: mismatched close paren: ~A",
scm_list_1 (SCM_MAKE_CHAR (c)));
- scm_ungetc (c, port);
+ scm_ungetc_unlocked (c, port);
tmp = scm_read_expression (port, opts);
/* See above note about scm_sym_dot. */
if (terminating_char != c)
scm_i_input_error (FUNC_NAME, port,
"in pair: missing close paren", SCM_EOL);
- goto exit;
+ break;
}
new_tail = scm_cons (tmp, SCM_EOL);
tl = new_tail;
}
- exit:
+ if (curly_list_p)
+ {
+ /* In addition to finding the length, 'scm_ilength' checks for
+ improper or circular lists, in which case it returns -1. */
+ int len = scm_ilength (ans);
+
+ /* The (len == 0) case is handled above */
+ if (len == 1)
+ /* Return directly to avoid re-annotating the element's source
+ location with the position of the outer brace. Also, it
+ might not be possible to annotate the element. */
+ return scm_car (ans); /* {e} => e */
+ else if (len == 2)
+ ; /* Leave the list unchanged: {e1 e2} => (e1 e2) */
+ else if (len >= 3 && (len & 1))
+ {
+ /* It's a proper list whose length is odd and at least 3. If
+ the elements at odd indices (the infix operator positions)
+ are all 'equal?', then it's a simple curly-infix list.
+ Otherwise it's a mixed curly-infix list. */
+ SCM op = scm_cadr (ans);
+
+ /* Check to see if the elements at odd indices are 'equal?' */
+ for (tl = scm_cdddr (ans); ; tl = scm_cddr (tl))
+ {
+ if (scm_is_null (tl))
+ {
+ /* Convert simple curly-infix list to prefix:
+ {a <op> b <op> ...} => (<op> a b ...) */
+ tl = ans;
+ while (scm_is_pair (scm_cdr (tl)))
+ {
+ tmp = scm_cddr (tl);
+ SCM_SETCDR (tl, tmp);
+ tl = tmp;
+ }
+ ans = scm_cons (op, ans);
+ break;
+ }
+ else if (scm_is_false (scm_equal_p (op, scm_car (tl))))
+ {
+ /* Mixed curly-infix list: {e ...} => ($nfx$ e ...) */
+ ans = scm_cons (sym_nfx, ans);
+ break;
+ }
+ }
+ }
+ else
+ /* Mixed curly-infix (possibly improper) list:
+ {e . tail} => ($nfx$ e . tail) */
+ ans = scm_cons (sym_nfx, ans);
+ }
+
return maybe_annotate_source (ans, port, opts, line, column);
}
#undef FUNC_NAME
c = 0; \
while (i < ndigits) \
{ \
- a = scm_getc (port); \
+ a = scm_getc_unlocked (port); \
if (a == EOF) \
goto str_eof; \
if (terminator \
do
{
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
if (c == EOF)
return;
}
while (c == '\t' || uc_is_general_category (c, UC_SPACE_SEPARATOR));
- scm_ungetc (c, port);
+ scm_ungetc_unlocked (c, port);
}
+/* Read either a double-quoted string or an R7RS-style symbol delimited
+ by vertical lines, depending on the value of 'chr' ('"' or '|').
+ Regardless, the result is always returned as a string. */
static SCM
-scm_read_string (int chr, SCM port, scm_t_read_opts *opts)
+scm_read_string_like_syntax (int chr, SCM port, scm_t_read_opts *opts)
#define FUNC_NAME "scm_lreadr"
{
/* For strings smaller than C_STR, this function creates only one Scheme
long line = SCM_LINUM (port);
int column = SCM_COL (port) - 1;
- while ('"' != (c = scm_getc (port)))
+ while (chr != (c = scm_getc_unlocked (port)))
{
if (c == EOF)
{
str_eof:
scm_i_input_error (FUNC_NAME, port,
- "end of file in string constant", SCM_EOL);
+ (chr == '|'
+ ? "end of file in symbol"
+ : "end of file in string constant"),
+ SCM_EOL);
}
if (c_str_len + 1 >= READER_STRING_BUFFER_SIZE)
if (c == '\\')
{
- switch (c = scm_getc (port))
+ switch (c = scm_getc_unlocked (port))
{
case EOF:
goto str_eof;
- case '"':
+ case '|':
case '\\':
break;
case '\n':
c = '\010';
break;
case 'x':
- if (opts->r6rs_escapes_p)
+ if (opts->r6rs_escapes_p || chr == '|')
SCM_READ_HEX_ESCAPE (10, ';');
else
SCM_READ_HEX_ESCAPE (2, '\0');
break;
}
default:
+ if (c == chr)
+ break;
bad_escaped:
scm_i_input_error (FUNC_NAME, port,
"illegal character in escape sequence: ~S",
}
#undef FUNC_NAME
+static SCM
+scm_read_string (int chr, SCM port, scm_t_read_opts *opts)
+{
+ return scm_read_string_like_syntax (chr, port, opts);
+}
+
+static SCM
+scm_read_r7rs_symbol (int chr, SCM port, scm_t_read_opts *opts)
+{
+ return scm_string_to_symbol (scm_read_string_like_syntax (chr, port, opts));
+}
static SCM
scm_read_number (scm_t_wchar chr, SCM port, scm_t_read_opts *opts)
SCM result, str = SCM_EOL;
char local_buffer[READER_BUFFER_SIZE], *buffer;
size_t bytes_read;
- scm_t_port *pt = SCM_PTAB_ENTRY (port);
/* Need to capture line and column numbers here. */
long line = SCM_LINUM (port);
int column = SCM_COL (port) - 1;
- scm_ungetc (chr, port);
+ scm_ungetc_unlocked (chr, port);
buffer = read_complete_token (port, opts, local_buffer, sizeof local_buffer,
&bytes_read);
- str = scm_from_stringn (buffer, bytes_read, pt->encoding, pt->ilseq_handler);
+ str = scm_from_port_stringn (buffer, bytes_read, port);
result = scm_string_to_number (str, SCM_UNDEFINED);
if (scm_is_false (result))
size_t bytes_read;
int postfix = (opts->keyword_style == KEYWORD_STYLE_POSTFIX);
char local_buffer[READER_BUFFER_SIZE], *buffer;
- scm_t_port *pt = SCM_PTAB_ENTRY (port);
SCM str;
- scm_ungetc (chr, port);
+ scm_ungetc_unlocked (chr, port);
buffer = read_complete_token (port, opts, local_buffer, sizeof local_buffer,
&bytes_read);
if (bytes_read > 0)
if (postfix && ends_with_colon && (bytes_read > 1))
{
- str = scm_from_stringn (buffer, bytes_read - 1,
- pt->encoding, pt->ilseq_handler);
+ str = scm_from_port_stringn (buffer, bytes_read - 1, port);
if (opts->case_insensitive_p)
str = scm_string_downcase_x (str);
}
else
{
- str = scm_from_stringn (buffer, bytes_read,
- pt->encoding, pt->ilseq_handler);
+ str = scm_from_port_stringn (buffer, bytes_read, port);
if (opts->case_insensitive_p)
str = scm_string_downcase_x (str);
char local_buffer[READER_BUFFER_SIZE], *buffer;
unsigned int radix;
SCM str;
- scm_t_port *pt;
switch (chr)
{
break;
default:
- scm_ungetc (chr, port);
- scm_ungetc ('#', port);
+ scm_ungetc_unlocked (chr, port);
+ scm_ungetc_unlocked ('#', port);
radix = 10;
}
buffer = read_complete_token (port, opts, local_buffer, sizeof local_buffer,
&read);
- pt = SCM_PTAB_ENTRY (port);
- str = scm_from_stringn (buffer, read, pt->encoding, pt->ilseq_handler);
+ str = scm_from_port_stringn (buffer, read, port);
result = scm_string_to_number (str, scm_from_uint (radix));
{
scm_t_wchar c;
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
if ('@' == c)
p = scm_sym_uq_splicing;
else
{
- scm_ungetc (c, port);
+ scm_ungetc_unlocked (c, port);
p = scm_sym_unquote;
}
break;
{
int c;
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
if ('@' == c)
p = sym_unsyntax_splicing;
else
{
- scm_ungetc (c, port);
+ scm_ungetc_unlocked (c, port);
p = sym_unsyntax;
}
break;
/* We use the get_byte here because there is no need to get the
locale correct with comment input. This presumes that newline
always represents itself no matter what the encoding is. */
- for (c = scm_get_byte_or_eof (port);
+ for (c = scm_get_byte_or_eof_unlocked (port);
(c != EOF) && (c != '\n');
- c = scm_get_byte_or_eof (port));
+ c = scm_get_byte_or_eof_unlocked (port));
return SCM_UNSPECIFIED;
}
+/* If the EXPECTED_CHARS are the next ones available from PORT, then
+ consume them and return 1. Otherwise leave the port position where
+ it was and return 0. EXPECTED_CHARS should be all lowercase, and
+ will be matched case-insensitively against the characters read from
+ PORT. */
+static int
+try_read_ci_chars (SCM port, const char *expected_chars)
+{
+ int num_chars_wanted = strlen (expected_chars);
+ int num_chars_read = 0;
+ char *chars_read = alloca (num_chars_wanted);
+ int c;
+
+ while (num_chars_read < num_chars_wanted)
+ {
+ c = scm_getc_unlocked (port);
+ if (c == EOF)
+ break;
+ else if (c_tolower (c) != expected_chars[num_chars_read])
+ {
+ scm_ungetc_unlocked (c, port);
+ break;
+ }
+ else
+ chars_read[num_chars_read++] = c;
+ }
+
+ if (num_chars_read == num_chars_wanted)
+ return 1;
+ else
+ {
+ while (num_chars_read > 0)
+ scm_ungetc_unlocked (chars_read[--num_chars_read], port);
+ return 0;
+ }
+}
+
\f
/* Sharp readers, i.e. readers called after a `#' sign has been read. */
{
case 't':
case 'T':
+ try_read_ci_chars (port, "rue");
return SCM_BOOL_T;
case 'f':
case 'F':
+ try_read_ci_chars (port, "alse");
return SCM_BOOL_F;
}
size_t charname_len, bytes_read;
scm_t_wchar cp;
int overflow;
- scm_t_port *pt;
+ scm_t_port_internal *pti;
overflow = read_token (port, opts, buffer, READER_CHAR_NAME_MAX_SIZE,
&bytes_read);
if (bytes_read == 0)
{
- chr = scm_getc (port);
+ chr = scm_getc_unlocked (port);
if (chr == EOF)
scm_i_input_error (FUNC_NAME, port, "unexpected end of file "
"while reading character", SCM_EOL);
return (SCM_MAKE_CHAR (chr));
}
- pt = SCM_PTAB_ENTRY (port);
+ pti = SCM_PORT_GET_INTERNAL (port);
/* Simple ASCII characters can be processed immediately. Also, simple
ISO-8859-1 characters can be processed immediately if the encoding for this
port is ISO-8859-1. */
- if (bytes_read == 1 && ((unsigned char) buffer[0] <= 127 || pt->encoding == NULL))
+ if (bytes_read == 1 &&
+ ((unsigned char) buffer[0] <= 127
+ || pti->encoding_mode == SCM_PORT_ENCODING_MODE_LATIN1))
{
SCM_COL (port) += 1;
return SCM_MAKE_CHAR (buffer[0]);
/* Otherwise, convert the buffer into a proper scheme string for
processing. */
- charname = scm_from_stringn (buffer, bytes_read, pt->encoding,
- pt->ilseq_handler);
+ charname = scm_from_port_stringn (buffer, bytes_read, port);
charname_len = scm_i_string_length (charname);
SCM_COL (port) += charname_len;
cp = scm_i_string_ref (charname, 0);
if (c == '-')
{
sign = -1;
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
}
while ('0' <= c && c <= '9')
{
+ if (((SSIZE_MAX - (c-'0')) / 10) <= res)
+ scm_i_input_error ("read_decimal_integer", port,
+ "number too large", SCM_EOL);
res = 10*res + c-'0';
got_it = 1;
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
}
if (got_it)
vectors. Also, the conflict between '#f' and '#f32' and '#f64' is
handled here.
- C is the first character read after the '#'.
-*/
+ C is the first character read after the '#'. */
static SCM
scm_read_array (int c, SCM port, scm_t_read_opts *opts, long line, int column)
{
/* XXX - shortcut for ordinary vectors. Shouldn't be necessary but
the array code can not deal with zero-length dimensions yet, and
- we want to allow zero-length vectors, of course.
- */
+ we want to allow zero-length vectors, of course. */
if (c == '(')
return scm_read_vector (c, port, opts, line, column);
- /* Disambiguate between '#f' and uniform floating point vectors.
- */
+ /* Disambiguate between '#f' and uniform floating point vectors. */
if (c == 'f')
{
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
if (c != '3' && c != '6')
{
- if (c != EOF)
- scm_ungetc (c, port);
+ if (c == 'a' && try_read_ci_chars (port, "lse"))
+ return SCM_BOOL_F;
+ else if (c != EOF)
+ scm_ungetc_unlocked (c, port);
return SCM_BOOL_F;
}
rank = 1;
&& tag_len < sizeof tag_buf / sizeof tag_buf[0])
{
tag_buf[tag_len++] = c;
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
}
if (tag_len == 0)
tag = SCM_BOOL_T;
if (c == '@')
{
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
c = read_decimal_integer (port, c, &lbnd);
}
if (c == ':')
{
- c = scm_getc (port);
+ c = scm_getc_unlocked (port);
c = read_decimal_integer (port, c, &len);
if (len < 0)
scm_i_input_error (NULL, port,
scm_read_bytevector (scm_t_wchar chr, SCM port, scm_t_read_opts *opts,
long line, int column)
{
- chr = scm_getc (port);
+ chr = scm_getc_unlocked (port);
if (chr != 'u')
goto syntax;
- chr = scm_getc (port);
+ chr = scm_getc_unlocked (port);
if (chr != '8')
goto syntax;
- chr = scm_getc (port);
+ chr = scm_getc_unlocked (port);
if (chr != '(')
goto syntax;
terribly inefficient but who cares? */
SCM s_bits = SCM_EOL;
- for (chr = scm_getc (port);
+ for (chr = scm_getc_unlocked (port);
(chr != EOF) && ((chr == '0') || (chr == '1'));
- chr = scm_getc (port))
+ chr = scm_getc_unlocked (port))
{
s_bits = scm_cons ((chr == '0') ? SCM_BOOL_F : SCM_BOOL_T, s_bits);
}
if (chr != EOF)
- scm_ungetc (chr, port);
+ scm_ungetc_unlocked (chr, port);
return maybe_annotate_source
(scm_bitvector (scm_reverse_x (s_bits, SCM_EOL)),
for (;;)
{
- int c = scm_getc (port);
+ int c = scm_getc_unlocked (port);
if (c == EOF)
scm_i_input_error ("skip_block_comment", port,
return SCM_UNSPECIFIED;
}
+static void set_port_case_insensitive_p (SCM port, scm_t_read_opts *opts,
+ int value);
+static void set_port_square_brackets_p (SCM port, scm_t_read_opts *opts,
+ int value);
+static void set_port_curly_infix_p (SCM port, scm_t_read_opts *opts,
+ int value);
+
static SCM
scm_read_shebang (scm_t_wchar chr, SCM port, scm_t_read_opts *opts)
{
- int c = 0;
- if ((c = scm_get_byte_or_eof (port)) != 'r')
- {
- scm_ungetc (c, port);
- return scm_read_scsh_block_comment (chr, port);
- }
- if ((c = scm_get_byte_or_eof (port)) != '6')
- {
- scm_ungetc (c, port);
- scm_ungetc ('r', port);
- return scm_read_scsh_block_comment (chr, port);
- }
- if ((c = scm_get_byte_or_eof (port)) != 'r')
- {
- scm_ungetc (c, port);
- scm_ungetc ('6', port);
- scm_ungetc ('r', port);
- return scm_read_scsh_block_comment (chr, port);
- }
- if ((c = scm_get_byte_or_eof (port)) != 's')
+ char name[READER_DIRECTIVE_NAME_MAX_SIZE + 1];
+ int c;
+ int i = 0;
+
+ while (i <= READER_DIRECTIVE_NAME_MAX_SIZE)
{
- scm_ungetc (c, port);
- scm_ungetc ('r', port);
- scm_ungetc ('6', port);
- scm_ungetc ('r', port);
- return scm_read_scsh_block_comment (chr, port);
+ c = scm_getc_unlocked (port);
+ if (c == EOF)
+ scm_i_input_error ("skip_block_comment", port,
+ "unterminated `#! ... !#' comment", SCM_EOL);
+ else if (('a' <= c && c <= 'z') || ('0' <= c && c <= '9') || c == '-')
+ name[i++] = c;
+ else if (CHAR_IS_DELIMITER (c))
+ {
+ scm_ungetc_unlocked (c, port);
+ name[i] = '\0';
+ if (0 == strcmp ("r6rs", name))
+ ; /* Silently ignore */
+ else if (0 == strcmp ("fold-case", name))
+ set_port_case_insensitive_p (port, opts, 1);
+ else if (0 == strcmp ("no-fold-case", name))
+ set_port_case_insensitive_p (port, opts, 0);
+ else if (0 == strcmp ("curly-infix", name))
+ set_port_curly_infix_p (port, opts, 1);
+ else if (0 == strcmp ("curly-infix-and-bracket-lists", name))
+ {
+ set_port_curly_infix_p (port, opts, 1);
+ set_port_square_brackets_p (port, opts, 0);
+ }
+ else
+ break;
+
+ return SCM_UNSPECIFIED;
+ }
+ else
+ {
+ scm_ungetc_unlocked (c, port);
+ break;
+ }
}
-
- return SCM_UNSPECIFIED;
+ while (i > 0)
+ scm_ungetc_unlocked (name[--i], port);
+ return scm_read_scsh_block_comment (chr, port);
}
static SCM
nested. So care must be taken. */
int nesting_level = 1;
- int a = scm_getc (port);
+ int a = scm_getc_unlocked (port);
if (a == EOF)
scm_i_input_error ("scm_read_r6rs_block_comment", port,
while (nesting_level > 0)
{
- int b = scm_getc (port);
+ int b = scm_getc_unlocked (port);
if (b == EOF)
scm_i_input_error ("scm_read_r6rs_block_comment", port,
if (EOF == c)
scm_i_input_error ("read_commented_expression", port,
"no expression after #; comment", SCM_EOL);
- scm_ungetc (c, port);
+ scm_ungetc_unlocked (c, port);
scm_read_expression (port, opts);
return SCM_UNSPECIFIED;
}
buf = scm_i_string_start_writing (buf);
- while ((chr = scm_getc (port)) != EOF)
+ while ((chr = scm_getc_unlocked (port)) != EOF)
{
if (saw_brace)
{
that the extended read syntax would never put a `\' before
an `x'. For now, we just ignore other instances of
backslash in the string. */
- switch ((chr = scm_getc (port)))
+ switch ((chr = scm_getc_unlocked (port)))
{
case EOF:
goto done;
got = scm_call_2 (proc, SCM_MAKE_CHAR (chr), port);
- if (scm_is_pair (got) && !scm_i_has_source_properties (got))
+ if (opts->record_positions_p && SCM_NIMP (got)
+ && !scm_i_has_source_properties (got))
scm_i_set_source_properties_x (got, line, column, SCM_FILENAME (port));
return got;
{
SCM result;
- chr = scm_getc (port);
+ chr = scm_getc_unlocked (port);
result = scm_read_sharp_extension (chr, port, opts);
if (!scm_is_eq (result, SCM_UNSPECIFIED))
case '0': case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9':
case '@':
-#if SCM_ENABLE_DEPRECATED
- /* See below for 'i' and 'e'. */
- case 'a':
- case 'y':
- case 'h':
- case 'l':
-#endif
return (scm_read_array (chr, port, opts, line, column));
case 'i':
case 'e':
-#if SCM_ENABLE_DEPRECATED
- {
- /* When next char is '(', it really is an old-style
- uniform array. */
- scm_t_wchar next_c = scm_getc (port);
- if (next_c != EOF)
- scm_ungetc (next_c, port);
- if (next_c == '(')
- return scm_read_array (chr, port, opts, line, column);
- /* Fall through. */
- }
-#endif
case 'b':
case 'B':
case 'o':
#undef FUNC_NAME
static SCM
-scm_read_expression (SCM port, scm_t_read_opts *opts)
-#define FUNC_NAME "scm_read_expression"
+read_inner_expression (SCM port, scm_t_read_opts *opts)
+#define FUNC_NAME "read_inner_expression"
{
while (1)
{
scm_t_wchar chr;
- chr = scm_getc (port);
+ chr = scm_getc_unlocked (port);
switch (chr)
{
case ';':
(void) scm_read_semicolon_comment (chr, port);
break;
+ case '{':
+ if (opts->curly_infix_p)
+ {
+ if (opts->neoteric_p)
+ return scm_read_sexp (chr, port, opts);
+ else
+ {
+ SCM expr;
+
+ /* Enable neoteric expressions within curly braces */
+ opts->neoteric_p = 1;
+ expr = scm_read_sexp (chr, port, opts);
+ opts->neoteric_p = 0;
+ return expr;
+ }
+ }
+ else
+ return scm_read_mixed_case_symbol (chr, port, opts);
case '[':
- if (!opts->square_brackets_p)
- return (scm_read_mixed_case_symbol (chr, port, opts));
- /* otherwise fall through */
+ if (opts->square_brackets_p)
+ return scm_read_sexp (chr, port, opts);
+ else if (opts->curly_infix_p)
+ {
+ /* The syntax of neoteric expressions requires that '[' be
+ a delimiter when curly-infix is enabled, so it cannot
+ be part of an unescaped symbol. We might as well do
+ something useful with it, so we adopt Kawa's convention:
+ [...] => ($bracket-list$ ...) */
+ long line = SCM_LINUM (port);
+ int column = SCM_COL (port) - 1;
+ return maybe_annotate_source
+ (scm_cons (sym_bracket_list, scm_read_sexp (chr, port, opts)),
+ port, opts, line, column);
+ }
+ else
+ return scm_read_mixed_case_symbol (chr, port, opts);
case '(':
return (scm_read_sexp (chr, port, opts));
case '"':
return (scm_read_string (chr, port, opts));
+ case '|':
+ if (opts->r7rs_symbols_p)
+ return scm_read_r7rs_symbol (chr, port, opts);
+ else
+ return scm_read_mixed_case_symbol (chr, port, opts);
case '\'':
case '`':
case ',':
case ')':
scm_i_input_error (FUNC_NAME, port, "unexpected \")\"", SCM_EOL);
break;
+ case '}':
+ if (opts->curly_infix_p)
+ scm_i_input_error (FUNC_NAME, port, "unexpected \"}\"", SCM_EOL);
+ else
+ return scm_read_mixed_case_symbol (chr, port, opts);
case ']':
if (opts->square_brackets_p)
scm_i_input_error (FUNC_NAME, port, "unexpected \"]\"", SCM_EOL);
}
#undef FUNC_NAME
+static SCM
+scm_read_expression (SCM port, scm_t_read_opts *opts)
+#define FUNC_NAME "scm_read_expression"
+{
+ if (!opts->neoteric_p)
+ return read_inner_expression (port, opts);
+ else
+ {
+ long line = 0;
+ int column = 0;
+ SCM expr;
+
+ if (opts->record_positions_p)
+ {
+ /* We need to get the position of the first non-whitespace
+ character in order to correctly annotate neoteric
+ expressions. For example, for the expression 'f(x)', the
+ first call to 'read_inner_expression' reads the 'f' (which
+ cannot be annotated), and then we later read the '(x)' and
+ use it to construct the new list (f x). */
+ int c = flush_ws (port, opts, (char *) NULL);
+ if (c == EOF)
+ return SCM_EOF_VAL;
+ scm_ungetc_unlocked (c, port);
+ line = SCM_LINUM (port);
+ column = SCM_COL (port);
+ }
+
+ expr = read_inner_expression (port, opts);
+
+ /* 'expr' is the first component of the neoteric expression. Now
+ we loop, and as long as the next character is '(', '[', or '{',
+ (without any intervening whitespace), we use it to construct a
+ new expression. For example, f{n - 1}(x) => ((f (- n 1)) x). */
+ for (;;)
+ {
+ int chr = scm_getc_unlocked (port);
+
+ if (chr == '(')
+ /* e(...) => (e ...) */
+ expr = scm_cons (expr, scm_read_sexp (chr, port, opts));
+ else if (chr == '[')
+ /* e[...] => ($bracket-apply$ e ...) */
+ expr = scm_cons (sym_bracket_apply,
+ scm_cons (expr,
+ scm_read_sexp (chr, port, opts)));
+ else if (chr == '{')
+ {
+ SCM arg = scm_read_sexp (chr, port, opts);
+
+ if (scm_is_null (arg))
+ expr = scm_list_1 (expr); /* e{} => (e) */
+ else
+ expr = scm_list_2 (expr, arg); /* e{...} => (e {...}) */
+ }
+ else
+ {
+ if (chr != EOF)
+ scm_ungetc_unlocked (chr, port);
+ break;
+ }
+ maybe_annotate_source (expr, port, opts, line, column);
+ }
+ return expr;
+ }
+}
+#undef FUNC_NAME
+
\f
/* Actual reader. */
+static void init_read_options (SCM port, scm_t_read_opts *opts);
+
SCM_DEFINE (scm_read, "read", 0, 1, 0,
(SCM port),
"Read an s-expression from the input port @var{port}, or from\n"
port = scm_current_input_port ();
SCM_VALIDATE_OPINPORT (1, port);
- init_read_options (&opts);
+ init_read_options (port, &opts);
c = flush_ws (port, &opts, (char *) NULL);
if (EOF == c)
return SCM_EOF_VAL;
- scm_ungetc (c, port);
+ scm_ungetc_unlocked (c, port);
return (scm_read_expression (port, &opts));
}
}
}
-#define SCM_ENCODING_SEARCH_SIZE (500)
+static int
+is_encoding_char (char c)
+{
+ if (c >= 'a' && c <= 'z') return 1;
+ if (c >= 'A' && c <= 'Z') return 1;
+ if (c >= '0' && c <= '9') return 1;
+ return strchr ("_-.:/,+=()", c) != NULL;
+}
+
+/* Maximum size of an encoding name. This is a bit more than the
+ longest name listed at
+ <http://www.iana.org/assignments/character-sets> ("ISO-2022-JP-2", 13
+ characters.) */
+#define ENCODING_NAME_MAX_SIZE 20
+
+/* Number of bytes at the beginning or end of a file that are scanned
+ for a "coding:" declaration. */
+#define SCM_ENCODING_SEARCH_SIZE (500 + ENCODING_NAME_MAX_SIZE)
-/* Search the first few hundred characters of a file for an Emacs-like coding
- declaration. Returns either NULL or a string whose storage has been
- allocated with `scm_gc_malloc ()'. */
+
+/* Search the SCM_ENCODING_SEARCH_SIZE bytes of a file for an Emacs-like
+ coding declaration. Returns either NULL or a string whose storage
+ has been allocated with `scm_gc_malloc'. */
char *
scm_i_scan_for_encoding (SCM port)
{
char header[SCM_ENCODING_SEARCH_SIZE+1];
size_t bytes_read, encoding_length, i;
char *encoding = NULL;
- int utf8_bom = 0;
char *pos, *encoding_start;
int in_comment;
pt = SCM_PTAB_ENTRY (port);
if (pt->rw_active == SCM_PORT_WRITE)
- scm_flush (port);
+ scm_flush_unlocked (port);
if (pt->rw_random)
pt->rw_active = SCM_PORT_READ;
if (pt->read_pos == pt->read_end)
{
/* We can use the read buffer, and thus avoid a seek. */
- if (scm_fill_input (port) == EOF)
+ if (scm_fill_input_unlocked (port) == EOF)
return NULL;
bytes_read = pt->read_end - pt->read_pos;
if (SCM_FPORTP (port) && !SCM_FDES_RANDOM_P (SCM_FPORT_FDES (port)))
return NULL;
- bytes_read = scm_c_read (port, header, SCM_ENCODING_SEARCH_SIZE);
+ bytes_read = scm_c_read_unlocked (port, header, SCM_ENCODING_SEARCH_SIZE);
header[bytes_read] = '\0';
scm_seek (port, scm_from_int (0), scm_from_int (SEEK_SET));
}
- if (bytes_read > 3
- && header[0] == '\xef' && header[1] == '\xbb' && header[2] == '\xbf')
- utf8_bom = 1;
-
/* search past "coding[:=]" */
pos = header;
while (1)
if ((pos = strstr(pos, "coding")) == NULL)
return NULL;
- pos += strlen("coding");
- if (pos - header >= SCM_ENCODING_SEARCH_SIZE ||
+ pos += strlen ("coding");
+ if (pos - header >= SCM_ENCODING_SEARCH_SIZE ||
(*pos == ':' || *pos == '='))
{
pos ++;
}
/* skip spaces */
- while (pos - header <= SCM_ENCODING_SEARCH_SIZE &&
+ while (pos - header <= SCM_ENCODING_SEARCH_SIZE &&
(*pos == ' ' || *pos == '\t'))
pos ++;
+ if (pos - header >= SCM_ENCODING_SEARCH_SIZE - ENCODING_NAME_MAX_SIZE)
+ /* We found the "coding:" string, but there is probably not enough
+ room to store an encoding name in its entirety, so ignore it.
+ This makes sure we do not end up returning a truncated encoding
+ name. */
+ return NULL;
+
/* grab the next token */
encoding_start = pos;
i = 0;
while (encoding_start + i - header <= SCM_ENCODING_SEARCH_SIZE
&& encoding_start + i - header < bytes_read
- && (isalnum ((int) encoding_start[i])
- || strchr ("_-.:/,+=()", encoding_start[i]) != NULL))
+ && is_encoding_char (encoding_start[i]))
i++;
encoding_length = i;
return NULL;
encoding = scm_gc_strndup (encoding_start, encoding_length, "encoding");
- for (i = 0; i < encoding_length; i++)
- encoding[i] = toupper ((int) encoding[i]);
/* push backwards to make sure we were in a comment */
in_comment = 0;
/* This wasn't in a comment */
return NULL;
- if (utf8_bom && strcmp(encoding, "UTF-8"))
- scm_misc_error (NULL,
- "the port input declares the encoding ~s but is encoded as UTF-8",
- scm_list_1 (scm_from_locale_string (encoding)));
-
return encoding;
}
return SCM_BOOL_F;
else
{
- s_enc = scm_from_locale_string (enc);
+ s_enc = scm_string_upcase (scm_from_locale_string (enc));
return s_enc;
}
}
#undef FUNC_NAME
+\f
+/* Per-port read options.
+
+ We store per-port read options in the 'port-read-options' port
+ property, which is stored in the internal port structure. The value
+ stored is a single integer that contains a two-bit field for each
+ read option.
+
+ If a bit field contains READ_OPTION_INHERIT (3), that indicates that
+ the applicable value should be inherited from the corresponding
+ global read option. Otherwise, the bit field contains the value of
+ the read option. For boolean read options that have been set
+ per-port, the possible values are 0 or 1. If the 'keyword_style'
+ read option has been set per-port, its possible values are those in
+ 'enum t_keyword_style'. */
+
+/* Key to read options in port properties. */
+SCM_SYMBOL (sym_port_read_options, "port-read-options");
+
+/* Offsets of bit fields for each per-port override */
+#define READ_OPTION_COPY_SOURCE_P 0
+#define READ_OPTION_RECORD_POSITIONS_P 2
+#define READ_OPTION_CASE_INSENSITIVE_P 4
+#define READ_OPTION_KEYWORD_STYLE 6
+#define READ_OPTION_R6RS_ESCAPES_P 8
+#define READ_OPTION_SQUARE_BRACKETS_P 10
+#define READ_OPTION_HUNGRY_EOL_ESCAPES_P 12
+#define READ_OPTION_CURLY_INFIX_P 14
+#define READ_OPTION_R7RS_SYMBOLS_P 16
+
+/* The total width in bits of the per-port overrides */
+#define READ_OPTIONS_NUM_BITS 18
+
+#define READ_OPTIONS_INHERIT_ALL ((1UL << READ_OPTIONS_NUM_BITS) - 1)
+#define READ_OPTIONS_MAX_VALUE READ_OPTIONS_INHERIT_ALL
+
+#define READ_OPTION_MASK 3
+#define READ_OPTION_INHERIT 3
+
+static void
+set_port_read_option (SCM port, int option, int new_value)
+{
+ SCM scm_read_options;
+ unsigned int read_options;
+
+ new_value &= READ_OPTION_MASK;
+
+ scm_dynwind_begin (0);
+ scm_dynwind_lock_port (port);
+
+ scm_read_options = scm_i_port_property (port, sym_port_read_options);
+ if (scm_is_unsigned_integer (scm_read_options, 0, READ_OPTIONS_MAX_VALUE))
+ read_options = scm_to_uint (scm_read_options);
+ else
+ read_options = READ_OPTIONS_INHERIT_ALL;
+ read_options &= ~(READ_OPTION_MASK << option);
+ read_options |= new_value << option;
+ scm_read_options = scm_from_uint (read_options);
+ scm_i_set_port_property_x (port, sym_port_read_options, scm_read_options);
+
+ scm_dynwind_end ();
+}
+
+/* Set OPTS and PORT's case-insensitivity according to VALUE. */
+static void
+set_port_case_insensitive_p (SCM port, scm_t_read_opts *opts, int value)
+{
+ value = !!value;
+ opts->case_insensitive_p = value;
+ set_port_read_option (port, READ_OPTION_CASE_INSENSITIVE_P, value);
+}
+
+/* Set OPTS and PORT's square_brackets_p option according to VALUE. */
+static void
+set_port_square_brackets_p (SCM port, scm_t_read_opts *opts, int value)
+{
+ value = !!value;
+ opts->square_brackets_p = value;
+ set_port_read_option (port, READ_OPTION_SQUARE_BRACKETS_P, value);
+}
+
+/* Set OPTS and PORT's curly_infix_p option according to VALUE. */
+static void
+set_port_curly_infix_p (SCM port, scm_t_read_opts *opts, int value)
+{
+ value = !!value;
+ opts->curly_infix_p = value;
+ set_port_read_option (port, READ_OPTION_CURLY_INFIX_P, value);
+}
+
+/* Initialize OPTS based on PORT's read options and the global read
+ options. */
+static void
+init_read_options (SCM port, scm_t_read_opts *opts)
+{
+ SCM val, scm_read_options;
+ unsigned int read_options, x;
+
+ scm_read_options = scm_i_port_property (port, sym_port_read_options);
+
+ if (scm_is_unsigned_integer (scm_read_options, 0, READ_OPTIONS_MAX_VALUE))
+ read_options = scm_to_uint (scm_read_options);
+ else
+ read_options = READ_OPTIONS_INHERIT_ALL;
+
+ x = READ_OPTION_MASK & (read_options >> READ_OPTION_KEYWORD_STYLE);
+ if (x == READ_OPTION_INHERIT)
+ {
+ val = SCM_PACK (SCM_KEYWORD_STYLE);
+ if (scm_is_eq (val, scm_keyword_prefix))
+ x = KEYWORD_STYLE_PREFIX;
+ else if (scm_is_eq (val, scm_keyword_postfix))
+ x = KEYWORD_STYLE_POSTFIX;
+ else
+ x = KEYWORD_STYLE_HASH_PREFIX;
+ }
+ opts->keyword_style = x;
+
+#define RESOLVE_BOOLEAN_OPTION(NAME, name) \
+ do \
+ { \
+ x = READ_OPTION_MASK & (read_options >> READ_OPTION_ ## NAME); \
+ if (x == READ_OPTION_INHERIT) \
+ x = !!SCM_ ## NAME; \
+ opts->name = x; \
+ } \
+ while (0)
+
+ RESOLVE_BOOLEAN_OPTION (COPY_SOURCE_P, copy_source_p);
+ RESOLVE_BOOLEAN_OPTION (RECORD_POSITIONS_P, record_positions_p);
+ RESOLVE_BOOLEAN_OPTION (CASE_INSENSITIVE_P, case_insensitive_p);
+ RESOLVE_BOOLEAN_OPTION (R6RS_ESCAPES_P, r6rs_escapes_p);
+ RESOLVE_BOOLEAN_OPTION (SQUARE_BRACKETS_P, square_brackets_p);
+ RESOLVE_BOOLEAN_OPTION (HUNGRY_EOL_ESCAPES_P, hungry_eol_escapes_p);
+ RESOLVE_BOOLEAN_OPTION (CURLY_INFIX_P, curly_infix_p);
+ RESOLVE_BOOLEAN_OPTION (R7RS_SYMBOLS_P, r7rs_symbols_p);
+
+#undef RESOLVE_BOOLEAN_OPTION
+
+ opts->neoteric_p = 0;
+}
+
void
scm_init_read ()
{
HCoop / bpt / guile.git / blobdiff