/* * read.c - reader * * Copyright (c) 2000-2024 Shiro Kawai * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * 3. Neither the name of the authors nor the names of its contributors * may be used to endorse or promote products derived from this * software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include #include #include #define LIBGAUCHE_BODY #include "gauche.h" #include "gauche/priv/configP.h" #include "gauche/char_attr.h" #include "gauche/priv/portP.h" #include "gauche/priv/builtin-syms.h" #include "gauche/priv/readerP.h" /* * READ */ static void read_context_flush(ScmReadContext *ctx); static ScmObj read_internal(ScmPort *port, ScmReadContext *ctx); static ScmObj read_item(ScmPort *port, ScmReadContext *ctx); static ScmObj read_list(ScmPort *port, ScmChar closer, ScmReadContext *ctx); static ScmObj read_vector(ScmPort *port, ScmChar closer, ScmReadContext *ctx); static ScmObj read_string(ScmPort *port, int incompletep, ScmReadContext *ctx); static ScmObj read_quoted(ScmPort *port, ScmObj quoter, ScmReadContext *ctx); static ScmObj read_char(ScmPort *port, ScmReadContext *ctx); static ScmObj read_word(ScmPort *port, ScmChar initial, ScmReadContext *ctx, int temp_case_fold, int include_hash_sign); static ScmObj read_symbol(ScmPort *port, ScmChar initial, ScmReadContext *ctx); static ScmObj read_immediate_symbol(ScmPort *port, ScmChar initial, int interned, const char *preceding, ScmReadContext *ctx); static ScmObj read_number(ScmPort *port, ScmChar initial, int radix, /* #r case */ ScmReadContext *ctx); static ScmObj read_symbol_or_number(ScmPort *port, ScmChar initial, ScmReadContext *ctx); static ScmObj read_escaped_symbol(ScmPort *port, ScmChar delim, int interned, ScmReadContext *ctx); static ScmObj read_keyword(ScmPort *port, ScmReadContext *ctx); static ScmObj read_regexp(ScmPort *port); static ScmObj read_charset(ScmPort *port); static ScmObj read_sharp_comma(ScmPort *port, ScmReadContext *ctx); static ScmObj read_sharp_asterisk(ScmPort *port, ScmReadContext *ctx); static ScmObj process_sharp_comma(ScmPort *port, int line, ScmObj key, ScmObj args, ScmReadContext *ctx, int has_ref); static ScmObj read_shebang(ScmPort *port, ScmReadContext *ctx); static ScmObj read_num_prefixed(ScmPort *port, ScmChar ch, ScmReadContext *ctx); static ScmObj read_sharp_word(ScmPort *port, char c, ScmReadContext *ctx); /* Table of 'read-time constructor' in SRFI-10 */ static struct { ScmHashTable *table; ScmInternalMutex mutex; } readCtorData; /* Table of hash-bang directive */ static struct { ScmHashTable *table; ScmInternalMutex mutex; } hashBangData; /* Parameter location for default reader mode */ static ScmPrimitiveParameter *defaultReadContext; /*---------------------------------------------------------------- * Entry points * Note: Entire read operation are done while locking the input port. * So we can use 'unsafe' version of port operations inside this file. * The lock is removed if reader routine signals an error. It is OK * to call read routine recursively. */ ScmObj Scm_ReadWithContext(ScmObj port, ScmReadContext *ctx) { ScmVM *vm = Scm_VM(); volatile ScmObj r = SCM_NIL; if (!SCM_PORTP(port) || SCM_PORT_DIR(port) != SCM_PORT_INPUT) { Scm_Error("input port required: %S", port); } if (!(ctx->flags & RCTX_RECURSIVELY)) { ctx->table = NULL; ctx->pending = SCM_NIL; } if (PORT_LOCKED(SCM_PORT(port), vm)) { r = read_item(SCM_PORT(port), ctx); } else { PORT_LOCK(SCM_PORT(port), vm); PORT_SAFE_CALL(SCM_PORT(port), r = read_item(SCM_PORT(port), ctx), /*no cleanup*/); PORT_UNLOCK(SCM_PORT(port)); } if (!(ctx->flags & RCTX_RECURSIVELY)) { read_context_flush(ctx); } return r; } ScmObj Scm_Read(ScmObj port) { return Scm_ReadWithContext(port, Scm_MakeReadContext(NULL)); } /* Convenience functions */ ScmObj Scm_ReadFromString(ScmString *str) { ScmObj inp = Scm_MakeInputStringPort(str, TRUE); ScmReadContext *ctx = Scm_MakeReadContext(NULL); ScmObj r = read_item(SCM_PORT(inp), ctx); read_context_flush(ctx); return r; } ScmObj Scm_ReadFromCString(const char *cstr) { ScmObj s = SCM_MAKE_STR_IMMUTABLE(cstr); ScmObj inp = Scm_MakeInputStringPort(SCM_STRING(s), TRUE); ScmReadContext *ctx = Scm_MakeReadContext(NULL); ScmObj r = read_item(SCM_PORT(inp), ctx); read_context_flush(ctx); return r; } ScmObj Scm_ReadListWithContext(ScmObj port, ScmChar closer, ScmReadContext *ctx) { ScmVM *vm = Scm_VM(); volatile ScmObj r = SCM_NIL; if (!SCM_PORTP(port) || SCM_PORT_DIR(port) != SCM_PORT_INPUT) { Scm_Error("input port required: %S", port); } if (!(ctx->flags & RCTX_RECURSIVELY)) { ctx->table = NULL; ctx->pending = SCM_NIL; } if (PORT_LOCKED(SCM_PORT(port), vm)) { r = read_list(SCM_PORT(port), closer, ctx); } else { PORT_LOCK(SCM_PORT(port), vm); PORT_SAFE_CALL(SCM_PORT(port), r = read_list(SCM_PORT(port), closer, ctx), /*no cleanup*/); PORT_UNLOCK(SCM_PORT(port)); } if (!(ctx->flags & RCTX_RECURSIVELY)) { read_context_flush(ctx); } return r; } ScmObj Scm_ReadList(ScmObj port, ScmChar closer) { ScmReadContext *ctx = Scm_MakeReadContext(NULL); return Scm_ReadListWithContext(port, closer, ctx); } /*---------------------------------------------------------------- * Read context */ ScmReadContext *Scm_CurrentReadContext() { ScmObj c = Scm_PrimitiveParameterRef(Scm_VM(), defaultReadContext); SCM_ASSERT(SCM_READ_CONTEXT_P(c)); return SCM_READ_CONTEXT(c); } ScmReadContext *Scm_SetCurrentReadContext(ScmReadContext *ctx) { ScmObj p = Scm_PrimitiveParameterSet(Scm_VM(), defaultReadContext, SCM_OBJ(ctx)); SCM_ASSERT(SCM_READ_CONTEXT_P(p)); return SCM_READ_CONTEXT(p); } static ScmReadContext *make_read_context(ScmReadContext *proto) { ScmReadContext *ctx = SCM_NEW(ScmReadContext); SCM_SET_CLASS(ctx, SCM_CLASS_READ_CONTEXT); ctx->flags = proto ? proto->flags : RCTX_SOURCE_INFO; ctx->table = NULL; ctx->pending = SCM_NIL; return ctx; } ScmReadContext *Scm_MakeReadContext(ScmReadContext *proto) { return make_read_context(proto? proto : Scm_CurrentReadContext()); } static void read_context_print(ScmObj obj, ScmPort *port, ScmWriteContext *ctx SCM_UNUSED) { Scm_Printf(port, "#", obj); } SCM_DEFINE_BUILTIN_CLASS_SIMPLE(Scm_ReadContextClass, read_context_print); /* public api */ int Scm_ReadContextLiteralImmutable(ScmReadContext *ctx) { return (ctx->flags & RCTX_LITERAL_IMMUTABLE); } /*---------------------------------------------------------------- * Error */ void Scm_ReadError(ScmPort *port, const char *msg, ...) { ScmObj ostr = Scm_MakeOutputStringPort(TRUE); ScmObj name = Scm_PortName(port); ScmSize line = Scm_PortLine(port); Scm_Printf(SCM_PORT(ostr), "Read error at %S:", SCM_STRINGP(name)? name : SCM_OBJ(SCM_MAKE_STR("??"))); if (line >= 0) { Scm_Printf(SCM_PORT(ostr), "line %d: ", line); } va_list ap; va_start(ap, msg); Scm_Vprintf(SCM_PORT(ostr), msg, ap, TRUE); va_end(ap); ScmObj rerr = Scm_MakeReadError(Scm_GetOutputString(SCM_PORT(ostr), 0), port, line); Scm_Raise(rerr, 0); } /*---------------------------------------------------------------- * Read reference */ /* Read reference is a proxy object to for referenced object (#N=). */ static void read_reference_print(ScmObj obj, ScmPort *port, ScmWriteContext *ctx); SCM_DEFINE_BUILTIN_CLASS_SIMPLE(Scm_ReadReferenceClass, read_reference_print); ScmObj Scm_MakeReadReference(void) { ScmReadReference *a = SCM_NEW(ScmReadReference); SCM_SET_CLASS(a, SCM_CLASS_READ_REFERENCE); a->value = SCM_UNBOUND; return SCM_OBJ(a); } static void read_reference_print(ScmObj obj SCM_UNUSED, ScmPort *port, ScmWriteContext *ctx SCM_UNUSED) { Scm_Printf(port, "#"); } static void ref_push(ScmReadContext *ctx, ScmObj obj, ScmObj finisher) { ctx->pending = Scm_Acons(obj, finisher, ctx->pending); } static ScmObj ref_val(ScmObj ref) { if (!SCM_READ_REFERENCE_REALIZED(ref)) { Scm_Error("reader encontered unresolved read reference. Implementation error?"); } return SCM_READ_REFERENCE(ref)->value; } static ScmObj ref_register(ScmReadContext *ctx, ScmObj obj, int refnum) { SCM_ASSERT(ctx->table); Scm_HashTableSet(ctx->table, SCM_MAKE_INT(refnum), obj, 0); return obj; } /* ctx->pending contains an assoc list of objects who contains read reference which should be resolved. The car of each entry is the object that needs to be fixed, and the cdr of eacy entry may contain a finisher procedure (if the object is created by read-time constructor. */ static void read_context_flush(ScmReadContext *ctx) { ScmObj cp, ep; SCM_FOR_EACH(cp, ctx->pending) { ScmObj entry = SCM_CAR(cp); SCM_ASSERT(SCM_PAIRP(entry)); ScmObj obj = SCM_CAR(entry); ScmObj finisher = SCM_CDR(entry); if (!SCM_FALSEP(finisher)) { Scm_ApplyRec(finisher, SCM_LIST1(obj)); } else if (SCM_PAIRP(obj)) { SCM_FOR_EACH(ep, obj) { if (SCM_READ_REFERENCE_P(SCM_CAR(ep))) { SCM_SET_CAR_UNCHECKED(ep, ref_val(SCM_CAR(ep))); } if (SCM_READ_REFERENCE_P(SCM_CDR(ep))) { /* in case we have (... . #N#) */ SCM_SET_CDR_UNCHECKED(ep, ref_val(SCM_CDR(ep))); break; } } } else if (SCM_VECTORP(obj)) { int i, len = SCM_VECTOR_SIZE(obj); for (i=0; i ? */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, /* @ A B C D E F G H I J K L M N O */ 1, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, /* P Q R S T U V W X Y Z [ \ ] ^ _ */ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 0, 0, 0, 1, 1, /* ` a b c d e f g h i j k l m n o */ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* p q r s t u v w x y z { | } ~ ^? */ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 0, }; /* For characters >= 0x80, we follow R[67]RS. */ inline static int char_word_constituent(int c, int include_hash_sign) { if (c < 128) { return ((c >= 0 && (ctypes[(unsigned char)c]&1)) || (c == '#' && include_hash_sign)); } else if (c == 0x200c || c == 0x200d) { return TRUE; } else { switch (Scm_CharGeneralCategory(c)) { case SCM_CHAR_CATEGORY_Lu: case SCM_CHAR_CATEGORY_Ll: case SCM_CHAR_CATEGORY_Lt: case SCM_CHAR_CATEGORY_Lm: case SCM_CHAR_CATEGORY_Lo: case SCM_CHAR_CATEGORY_Mn: case SCM_CHAR_CATEGORY_Nl: case SCM_CHAR_CATEGORY_No: case SCM_CHAR_CATEGORY_Pd: case SCM_CHAR_CATEGORY_Pc: case SCM_CHAR_CATEGORY_Po: case SCM_CHAR_CATEGORY_Sc: case SCM_CHAR_CATEGORY_Sm: case SCM_CHAR_CATEGORY_Sk: case SCM_CHAR_CATEGORY_So: case SCM_CHAR_CATEGORY_Co: return TRUE; default: return FALSE; } } } inline static int char_word_case_fold(int c) { /*NB: Should we adopt full-fledged case folding for #!fold-case mode? It seems overkill to me. For now, we only handles ASCII range.*/ return (c >= 0 && c < 128 && (ctypes[(unsigned char)c]&2)); } /* For external API * We may use flags for include_hash_sign in future. */ int Scm_CharWordConstituent(ScmChar ch, u_long flags SCM_UNUSED) { return char_word_constituent((int)ch, FALSE); } /* R7RS 7.1.1 */ static int char_is_delimiter(ScmChar ch) { return ((ch < 0x80 && strchr("()\";| \t\n\r", ch) != NULL) || SCM_CHAR_EXTRA_WHITESPACE(ch)); } static void read_nested_comment(ScmPort *port, ScmReadContext *ctx SCM_UNUSED) { int nesting = 0; ScmSize line = Scm_PortLine(port); for (ScmChar c = Scm_GetcUnsafe(port);;) { switch (c) { case '#': { ScmChar c1 = Scm_GetcUnsafe(port); if (c1 == '|') { nesting++; break; } else if (c1 == EOF) goto eof; else c = c1; continue; } case '|': { ScmChar c1 = Scm_GetcUnsafe(port); if (c1 == '#') { if (nesting-- == 0) { return; } break; } else if (c1 == EOF) goto eof; else c = c1; continue; } case EOF: eof: Scm_ReadError(port, "encountered EOF inside nested multi-line comment (comment begins at line %d)", line); default: break; } c = Scm_GetcUnsafe(port); } } static void read_comment(ScmPort *port) /* leading semicolon is already read */ { for (;;) { /* NB: comment may contain unexpected character code. for the safety, we read bytes here. */ int c = Scm_GetbUnsafe(port); if (c == '\n' || c == EOF) break; } } static int skipws(ScmPort *port, ScmReadContext *ctx SCM_UNUSED) { for (;;) { int c = Scm_GetcUnsafe(port); if (c == EOF) return c; if (c <= 127) { if (isspace(c)) continue; if (c == ';') { read_comment(port); continue; } return c; } else if (!SCM_CHAR_EXTRA_WHITESPACE(c)) return c; } } static void reject_in_r7(ScmPort *port, ScmReadContext *ctx SCM_UNUSED, const char *token) { if (SCM_EQ(Scm_GetPortReaderLexicalMode(port), SCM_SYM_STRICT_R7)) { Scm_ReadError(port, "lexical syntax %s isn't allowed in strict R7RS mode", token); } } static ScmObj read_internal(ScmPort *port, ScmReadContext *ctx) { int c = skipws(port, ctx); switch (c) { case '(': return read_list(port, ')', ctx); case '"': return read_string(port, FALSE, ctx); case '#': { int c1 = Scm_GetcUnsafe(port); switch (c1) { case EOF: Scm_ReadError(port, "premature #-sequence at EOF"); return SCM_UNDEFINED; /* dummy */ case 't':; case 'T': return read_sharp_word(port, 't', ctx); case 'f':; case 'F': return read_sharp_word(port, 'f', ctx); case 's':; case 'S': return read_sharp_word(port, 's', ctx); case 'u':; case 'U': return read_sharp_word(port, 'u', ctx); case 'c':; case 'C': return read_sharp_word(port, 'c', ctx); case '(': return read_vector(port, ')', ctx); case '\\': return read_char(port, ctx); case 'x':; case 'X':; case 'o':; case 'O':; case 'b':; case 'B':; case 'd':; case 'D':; case 'e':; case 'E':; case 'i':; case 'I':; Scm_UngetcUnsafe(c1, port); return read_number(port, c, 0, ctx); /* let StringToNumber handle radix prefix */ case '!': /* #! is either a script shebang or a reader directive */ return read_shebang(port, ctx); case '"': { /* #"..." - string interpolation */ reject_in_r7(port, ctx, "#\"...\""); Scm_UngetcUnsafe(c1, port); int line = Scm_PortLine(port); ScmObj form = read_item(port, ctx); return process_sharp_comma(port, line, SCM_SYM_STRING_INTERPOLATE, SCM_LIST4(form, SCM_FALSE, Scm_PortName(port), SCM_MAKE_INT(line)), ctx, FALSE); } case '/': /* #/.../ literal regexp */ reject_in_r7(port, ctx, "#/.../"); return read_regexp(port); case '[': /* #[...] literal charset */ reject_in_r7(port, ctx, "#[...]"); return read_charset(port); case ',': /* #,(form) - SRFI-10 read-time macro */ return read_sharp_comma(port, ctx); case '|': /* #| - block comment (SRFI-30) it is equivalent to whitespace, so we return # */ read_nested_comment(port, ctx); return SCM_UNDEFINED; case '`': { /* #`"..." - Legacy string interpolation syntax */ reject_in_r7(port, ctx, "#`\"...\""); int line = Scm_PortLine(port); ScmObj form = read_item(port, ctx); return process_sharp_comma(port, line, SCM_SYM_STRING_INTERPOLATE, SCM_LIST4(form, SCM_TRUE, Scm_PortName(port), SCM_MAKE_INT(line)), ctx, FALSE); } case '?': { /* #? - debug directives */ reject_in_r7(port, ctx, "#?"); int c2 = Scm_GetcUnsafe(port); switch (c2) { case '=': { /* #?=form - debug print */ ScmObj form = read_item(port, ctx); return SCM_LIST2(SCM_SYM_DEBUG_PRINT, form); } case ',': { /* #?,form - debug funcall */ ScmObj form = read_item(port, ctx); return SCM_LIST2(SCM_SYM_DEBUG_FUNCALL, form); } case '?': { /* #??= or #??, - conditinal debug stub */ ScmObj sym = SCM_FALSE; int c3 = Scm_GetcUnsafe(port); switch (c3) { case '=': sym = SCM_SYM_DEBUG_PRINT_CONDITIONALLY; break; case ',': sym = SCM_SYM_DEBUG_FUNCALL_CONDITIONALLY; break; default: Scm_ReadError(port, "unsupported #?-syntax: #??%C", c3); return SCM_UNDEFINED; /* dummy */ } ScmObj test = read_item(port, ctx); ScmObj form = read_item(port, ctx); return SCM_LIST3(sym, test, form); } case '@': { int c3 = Scm_GetcUnsafe(port); switch (c3) { case '=': { /* #?@=form - debug thread log */ ScmObj form = read_item(port, ctx); return SCM_LIST2(SCM_SYM_DEBUG_THREAD_LOG, form); } default: Scm_ReadError(port, "unsupported #?-syntax: #?@%C", c3); return SCM_UNDEFINED; /* dummy */ } } case EOF: return SCM_EOF; default: Scm_ReadError(port, "unsupported #?-syntax: #?%C", c2); return SCM_UNDEFINED; /* dummy */ } } case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': /* #N#, #N= or #Nr form */ return read_num_prefixed(port, c1, ctx); case '*': { reject_in_r7(port, ctx, "#*"); /* #**"...." byte string #*01001001 for bit vector. */ return read_sharp_asterisk(port, ctx); } case ':': { reject_in_r7(port, ctx, "#:"); /* #:name - uninterned symbol */ return read_immediate_symbol(port, Scm_GetcUnsafe(port), FALSE, "#:", ctx); } case ';': { /* #;expr - comment out sexpr */ int orig = ctx->flags; ctx->flags |= RCTX_DISABLE_CTOR; read_item(port, ctx); /* read and discard */ ctx->flags = orig; return SCM_UNDEFINED; /* indicate this is a comment */ } default: Scm_ReadError(port, "unsupported #-syntax: #%C", c1); return SCM_UNDEFINED; /* dummy */ } } case '\'': return read_quoted(port, SCM_SYM_QUOTE, ctx); case '`': return read_quoted(port, SCM_SYM_QUASIQUOTE, ctx); case ':': /* Would be removed once we make keywords a subtype of symbols. */ if (SCM_EQ(Scm_GetPortReaderLexicalMode(port), SCM_SYM_STRICT_R7)) { return read_symbol(port, c, ctx); } else { return read_keyword(port, ctx); } case ',': { int c1 = Scm_GetcUnsafe(port); if (c1 == EOF) { Scm_ReadError(port, "unterminated unquote"); return SCM_UNDEFINED; /* dummy */ } else if (c1 == '@') { return read_quoted(port, SCM_SYM_UNQUOTE_SPLICING, ctx); } else { Scm_UngetcUnsafe(c1, port); return read_quoted(port, SCM_SYM_UNQUOTE, ctx); } } case '|': return read_escaped_symbol(port, '|', TRUE, ctx); case '[': /* TODO: make it customizable */ reject_in_r7(port, ctx, "[]"); return read_list(port, ']', ctx); case '{': reject_in_r7(port, ctx, "{}"); return read_list(port, '}', ctx); case '+':; case '-': /* Note: R5RS doesn't permit identifiers beginning with '+' or '-', but some Scheme programs use such identifiers. */ return read_symbol_or_number(port, c, ctx); case '.': { int c1 = Scm_GetcUnsafe(port); if (!char_word_constituent(c1, FALSE)) { Scm_ReadError(port, "dot in wrong context"); } Scm_UngetcUnsafe(c1, port); return read_symbol_or_number(port, c, ctx); } case '0':; case '1':; case '2':; case '3':; case '4':; case '5':; case '6':; case '7':; case '8':; case '9':; /* Note: R5RS doesn't permit identifiers beginning with digits, but some Scheme programs use such identifiers. */ return read_symbol_or_number(port, c, ctx); case ')':; case ']':; case '}':; Scm_ReadError(port, "extra close parenthesis `%c'", c); return SCM_UNDEFINED; /* dummy */ case EOF: return SCM_EOF; default: return read_symbol(port, c, ctx); } } static ScmObj read_item(ScmPort *port, ScmReadContext *ctx) { for (;;) { ScmObj obj = read_internal(port, ctx); if (!SCM_UNDEFINEDP(obj)) return obj; } } /*-------------------------------------------------------- * Common routine to handle hex-digit escape \xNN etc. */ /* TODO: Redesign these for clearer and simpler API. */ /* Hex-escape sequence can appear in the following places: - character literals #\x1b - string literals "\x1b;(" - symbol literals |\x1b;| - character set literals #[\x1b;-\x1f;] - regexp literals #/\x1b;+ / Legacy Gauche supports fixed-digit syntax. \xN{2} - char code up to 256 in _native encoding_ \uN{4} - unicode codepoint in BMP \UN{8} - unicode codepoint New (R7RS) syntax has variable number of digits, terminated by ';' \xN{1,}; - unicode codepoint There's an ambiguity in '\x' case - we first try to read it as R7RS syntax, then falls back to legacy syntax, by default. Because of this backtrack, the API is a bit complicated. Scm_ReadXdigitsFromString reads from char* buffer, up to buflen octets. The preceding backslash and a character 'x', 'u' or 'U' should already be read. The character is passed to KEY. It also receives ScmReaderLexicalMode. Character literal doesn't take terminating ';'. TERMINATOR flag indicates whether we expect terminator or not. NEXTBUF is an output variable points to the next character on success. Returns the retrieved decoded character, or SCM_CHAR_INVALID on error. Scm_ReadXdigitsFromPort is a convenience routine on top of Scm_ReadXdigitsFromString, when the caller is reading from a port. Since we need look ahead arbitrary number of characters, we can't just return a retrieved character. The procedure takes ScmDString to which the read characters (the decoded character, plus remaining hexdigit characters) are accumulated. When the escape syntax is invalid, it returns # of prefetched characters. On success, it returns #t. */ ScmChar Scm_ReadXdigitsFromString(const char *buf, int buflen, ScmChar key, /* x, u or U */ ScmObj mode, /* Reader lexical mode */ int terminator, /* TRUE expecting ';' */ const char **nextbuf) { int legacy_fallback = FALSE; if (key == 'x' && !SCM_EQ(mode, SCM_SYM_LEGACY)) { int val = 0, i; int overflow = FALSE; for (i=0; i SCM_CHAR_MAX) overflow = TRUE; } else if (terminator && buf[i] == ';' && i > 0) { /* R7RS syntax */ *nextbuf = buf+i+1; return overflow? SCM_CHAR_INVALID : Scm_UcsToChar(val); } else if (terminator && i < 2) { return SCM_CHAR_INVALID; } else { break; } } if (!terminator && i == buflen) { *nextbuf = buf+i; return overflow? SCM_CHAR_INVALID:Scm_UcsToChar(val); } /* Fallback to legacy syntax */ legacy_fallback = TRUE; } if (SCM_EQ(mode, SCM_SYM_STRICT_R7)) return SCM_CHAR_INVALID; if (key == 'x' && SCM_EQ(mode, SCM_SYM_WARN_LEGACY)) { Scm_Warn("Legacy \\x hex-escape: \\x%c%c", buf[0], buf[1]); } { int val = 0; int ndigits = (key == 'u')? 4 : (key == 'x')? 2 : 8; if (ndigits > buflen) return SCM_CHAR_INVALID; for (int i=0; i SCM_CHAR_MAX) return SCM_CHAR_INVALID; } *nextbuf = buf + ndigits; if (!legacy_fallback && key != 'x') val = Scm_UcsToChar(val); return val; } } /* On success, parsed char and other prefetched hexdigits are added to BUF and returns #t. Otherwise the prefetched string is returned. */ ScmObj Scm_ReadXdigitsFromPort(ScmPort *port, int key, ScmObj mode, int incompletep, ScmDString *buf) { ScmDString ds; Scm_DStringInit(&ds); for (;;) { int ch = Scm_GetcUnsafe(port); if (ch == ';') { Scm_DStringPutc(&ds, ch); break; } if (ch == EOF || ch >= 0x80 || !isxdigit(ch)) { Scm_UngetcUnsafe(ch, port); break; } Scm_DStringPutc(&ds, ch); } ScmSmallInt numchars; const char *chars = Scm_DStringPeek(&ds, &numchars, NULL); const char *next; int r = Scm_ReadXdigitsFromString(chars, numchars, key, mode, TRUE, &next); if (r != SCM_CHAR_INVALID) { if (incompletep) Scm_DStringPutb(buf, r); else Scm_DStringPutc(buf, r); if (next - chars < numchars) { Scm_DStringPutz(buf, next, numchars - (next-chars)); } return SCM_TRUE; } else { return Scm_MakeString(chars, numchars, -1, SCM_STRING_COPYING); } } /* Another utility procedure. Read sequence of digits from PORT and made it long and return. CH can be prefetched digit char; if you don't have one, pass SCM_CHAR_INVALID. The first nondigit char read is stored in NEXT, and number of digits read is stored in NUMREAD. If the number overflows LONG_MAX or there's no digits to be read, returns -1 (you can check NUMREAD to see which is the case). NB: This may throw an error if ch is neither a valid digit char nor SCM_CHAR_INVALID. */ /* NB: Call for refactoring, using Scm_ParseDigitsAsLong */ long Scm_ReadDigitsAsLong(ScmPort *port, ScmChar ch, int radix, ScmChar *next, /*out*/ ScmSize *numread /*out*/) { long val = 0; ScmSize nchars = 0; if (ch != SCM_CHAR_INVALID) { int v = Scm_DigitToInt(ch, radix, FALSE); if (v < 0) { /* This is for safety net. The caller should filter out this case. */ Scm_ReadError(port, "Digit char expected, but got %C", ch); } val = v; } for (;;) { ch = Scm_Getc(port); nchars++; int v = Scm_DigitToInt(ch, radix, FALSE); if (v < 0) { /* EOF case is covered here */ *next = ch; *numread = nchars; return val; } if (val >= (LONG_MAX/radix+1)) { /* we'll overflow */ *next = ch; *numread = nchars; return -1; } val = val*radix+v; } /* NOTREACHED */ } /* Read long digits from BUF up to LEN. BUF must only contain single-byte chars. */ long Scm_ParseDigitsAsLong(const char *buf, ScmSize len, int radix, ScmSize *numread) /*out*/ { long val = 0; ScmSize nchars = 0; for (; nchars < len; nchars++, buf++) { int v = Scm_DigitToInt((ScmChar)*buf, radix, FALSE); if (v < 0) { *numread = nchars; return val; } if (val >= (LONG_MAX/radix+1)) { /* we'll overflow */ *numread = nchars; return -1; } val = val*radix+v; } *numread = nchars; return (nchars == 0)? -1 : val; } /*---------------------------------------------------------------- * List */ /* Internal read_list. returns whether the list contains unresolved reference or not within the flag has_ref */ static ScmObj read_list_int(ScmPort *port, ScmChar closer, ScmReadContext *ctx, int *has_ref, int start_line) { ScmObj start = SCM_NIL, last = SCM_NIL; int dot_seen = FALSE, ref_seen = FALSE; for (;;) { int c = skipws(port, ctx); if (c == EOF) goto eoferr; if (c == closer) { *has_ref = !!ref_seen; return start; } /* NB: We need to keep reading after we see the dotted pair, since there can be a case like (a . b #;c), which is valid syntax. We only reject when we see a next item after reading dotted pair. */ ScmObj item; if (c == '.') { if (dot_seen) goto baddot; int c2 = Scm_GetcUnsafe(port); if (c2 == closer) { goto baddot; } else if (c2 == EOF) { goto eoferr; } else if (!char_word_constituent(c2, FALSE)) { /* can be a dot pair at the end */ if (start == SCM_NIL) goto baddot; Scm_UngetcUnsafe(c2, port); item = read_item(port, ctx); if (SCM_READ_REFERENCE_P(item)) ref_seen = TRUE; SCM_SET_CDR_UNCHECKED(last, item); dot_seen = TRUE; continue; } Scm_UngetcUnsafe(c2, port); item = read_symbol_or_number(port, c, ctx); } else { Scm_UngetcUnsafe(c, port); item = read_internal(port, ctx); if (SCM_UNDEFINEDP(item)) continue; /* it was just a comment */ if (dot_seen) goto baddot; if (SCM_READ_REFERENCE_P(item)) ref_seen = TRUE; } SCM_APPEND1(start, last, item); } eoferr: if (start_line >= 0) { Scm_ReadError(port, "EOF inside a list (starting from line %d)", start_line); } else { Scm_ReadError(port, "EOF inside a list"); } baddot: Scm_ReadError(port, "bad dot syntax"); return SCM_NIL; /* dummy */ } static ScmObj read_list(ScmPort *port, ScmChar closer, ScmReadContext *ctx) { int has_ref; ScmSize line = -1; if (ctx->flags & RCTX_SOURCE_INFO) line = Scm_PortLine(port); ScmObj r = read_list_int(port, closer, ctx, &has_ref, line); if (SCM_PAIRP(r) && (ctx->flags & RCTX_SOURCE_INFO) && line >= 0) { /* Swap the head of the list for an extended pair to record source-code info.*/ r = Scm_ExtendedCons(SCM_CAR(r), SCM_CDR(r)); Scm_PairAttrSet(SCM_PAIR(r), SCM_SYM_SOURCE_INFO, SCM_LIST2(Scm_PortName(port), SCM_MAKE_INT(line))); } if (has_ref) ref_push(ctx, r, SCM_FALSE); return r; } static ScmObj read_vector(ScmPort *port, ScmChar closer, ScmReadContext *ctx) { int has_ref; ScmSize line = -1; ScmObj r; if (ctx->flags & RCTX_SOURCE_INFO) line = Scm_PortLine(port); r = read_list_int(port, closer, ctx, &has_ref, line); r = Scm_ListToVector(r, 0, -1); if (has_ref) ref_push(ctx, r, SCM_FALSE); return r; } static ScmObj read_quoted(ScmPort *port, ScmObj quoter, ScmReadContext *ctx) { ScmSize line = -1; ScmObj r; if (ctx->flags & RCTX_SOURCE_INFO) line = Scm_PortLine(port); ScmObj item = read_item(port, ctx); if (SCM_EOFP(item)) Scm_ReadError(port, "unterminated quote"); if (line >= 0) { r = Scm_ExtendedCons(quoter, Scm_Cons(item, SCM_NIL)); Scm_PairAttrSet(SCM_PAIR(r), SCM_SYM_SOURCE_INFO, SCM_LIST2(Scm_PortName(port), SCM_MAKE_INT(line))); } else { r = Scm_Cons(quoter, Scm_Cons(item, SCM_NIL)); } if (SCM_READ_REFERENCE_P(item)) ref_push(ctx, SCM_CDR(r), SCM_FALSE); return r; } /*---------------------------------------------------------------- * String */ /* Utility API to read string-literal like thingy. * The caller already read and consumed the opening double-quote. * * FLAGS are ScmStringFlags. Only SCM_STRING_INCOMPLETE is recognized. */ ScmObj Scm_ReadStringLiteral(ScmPort *port, ScmReadContext *ctx, /* can be NULL */ u_long flags, ScmChar terminator SCM_UNUSED) /* reserved */ { int incompletep = flags & SCM_STRING_INCOMPLETE; return read_string(port, incompletep, ctx); } /* Handling \xNN;, \uNNNN, \UNNNNNNNN escapes. To make it easier to support both legacy \xN{2} syntax and new \xN{1,} syntax, we read-ahead as many hexdigits as possible first, then parse it, and append the result to BUF, along with any unused digits. */ static void read_string_xdigits(ScmPort *port, int key, int incompletep, ScmDString *buf) { ScmObj bad = Scm_ReadXdigitsFromPort(port, key, Scm_GetPortReaderLexicalMode(port), incompletep, buf); if (SCM_STRINGP(bad)) { /* skip chars to the end of string, so that the reader will read after the erroneous string */ for (;;) { int c; if (incompletep) c = Scm_GetbUnsafe(port); else c = Scm_GetcUnsafe(port); if (c == EOF || c == '"') break; if (c == '\\') { if (incompletep) c = Scm_GetbUnsafe(port); else c = Scm_GetcUnsafe(port); } } Scm_ReadError(port, "Bad \\%c escape sequence in a string literal: `\\%c%A'", key, key, bad); } } static ScmObj read_string(ScmPort *port, int incompletep, ScmReadContext *ctx SCM_UNUSED) { int c = 0; ScmDString ds; Scm_DStringInit(&ds); #define FETCH(var) \ if (incompletep) { var = Scm_GetbUnsafe(port); } \ else { var = Scm_GetcUnsafe(port); } #define ACCUMULATE(var) \ if (incompletep) { SCM_DSTRING_PUTB(&ds, var); } \ else { SCM_DSTRING_PUTC(&ds, var); } #define INTRALINE_WS(var) \ ((var)==' ' || (var)=='\t' || SCM_CHAR_EXTRA_WHITESPACE_INTRALINE(var)) for (;;) { FETCH(c); switch (c) { case EOF: goto eof_exit; case '"': goto finish; backslash: case '\\': { FETCH(c); switch (c) { case EOF: goto eof_exit; case 'n': ACCUMULATE('\n'); break; case 'r': ACCUMULATE('\r'); break; case 'f': ACCUMULATE('\f'); break; case 't': ACCUMULATE('\t'); break; case 'a': ACCUMULATE(0x07); break; /* alarm */ case 'b': ACCUMULATE(0x08); break; /* backspace */ case '\\': ACCUMULATE('\\'); break; case '0': ACCUMULATE(0); break; case 'x': { read_string_xdigits(port, 'x', incompletep, &ds); break; } case 'u': case 'U': { if (SCM_EQ(Scm_GetPortReaderLexicalMode(port), SCM_SYM_STRICT_R7)) { Scm_ReadError(port, "\\%c in string literal isn't allowed " "in strinct-r7rs mode", c); } read_string_xdigits(port, c, incompletep, &ds); break; } /* R6RS-style line continuation handling*/ line_continuation: case ' ': case '\t': { for (;;) { FETCH(c); if (c == EOF) goto eof_exit; if (c == '\r') goto cont_cr; if (c == '\n') goto cont_lf; if (!INTRALINE_WS(c)) goto cont_err; } } /*FALLTHROUGH*/ cont_cr: case '\r': { FETCH(c); if (c == EOF) goto eof_exit; if (c == '"') goto finish; if (c == '\\') goto backslash; if (c != '\n' && !INTRALINE_WS(c)) { ACCUMULATE(c); break; } } /*FALLTHROUGH*/ cont_lf: case '\n': { for (;;) { FETCH(c); if (c == EOF) goto eof_exit; if (c == '"') goto finish; if (c == '\\') goto backslash; if (!INTRALINE_WS(c)) { ACCUMULATE(c); break; } } break; } default: if (SCM_CHAR_EXTRA_WHITESPACE_INTRALINE(c)) goto line_continuation; else ACCUMULATE(c); } break; } default: ACCUMULATE(c); break; } } eof_exit: Scm_ReadError(port, "EOF encountered in a string literal: %S", Scm_DStringGet(&ds, 0)); cont_err: Scm_ReadError(port, "Invalid line continuation sequence in a string literal: %S...", Scm_DStringGet(&ds, 0)); finish:; int flags = ((incompletep? SCM_STRING_INCOMPLETE:0) | SCM_STRING_IMMUTABLE); return Scm_DStringGet(&ds, flags); } /*---------------------------------------------------------------- * Character */ static struct char_name { const char *name; u_int size; ScmObj ch; } char_names[] = { #define DEFCHAR(name, char) \ { #name, sizeof(#name)-1, SCM_MAKE_CHAR(char) } DEFCHAR(alarm, 0x07), DEFCHAR(backspace, 0x08), DEFCHAR(space, ' '), DEFCHAR(newline, '\n'), DEFCHAR(nl, '\n'), DEFCHAR(lf, '\n'), DEFCHAR(return, '\r'), DEFCHAR(cr, '\r'), DEFCHAR(tab, '\t'), DEFCHAR(ht, '\t'), DEFCHAR(page, '\f'), DEFCHAR(escape, 0x1b), DEFCHAR(esc, 0x1b), DEFCHAR(delete, 0x7f), DEFCHAR(del, 0x7f), DEFCHAR(null, 0), { NULL, 0, 0 } }; static ScmObj read_char(ScmPort *port, ScmReadContext *ctx) { ScmString *name; int c = Scm_GetcUnsafe(port); switch (c) { case EOF: Scm_ReadError(port, "EOF encountered in character literal"); break; case '(':; case ')':; case '[':; case ']':; case '{':; case '}':; case '"':; case ' ':; case '\\':; case '|':; case ';':; case '#':; return SCM_MAKE_CHAR(c); default: { /* need to read word to see if it is a character name */ name = SCM_STRING(read_word(port, c, ctx, TRUE, FALSE)); if (SCM_EQ(Scm_GetPortReaderLexicalMode(port), SCM_SYM_STRICT_R7)) { ScmChar following = Scm_GetcUnsafe(port); if (!char_is_delimiter(following)) { Scm_Error("Character literal isn't delimited: #\\%A%C ...", name, following); } Scm_UngetcUnsafe(following, port); } ScmSmallInt namelen, namesize; const char *cname = Scm_GetStringContent(name, &namesize, &namelen, NULL); if (namelen == 1) { return SCM_MAKE_CHAR(c); } if (namelen != namesize) { /* no character name contains multibyte chars */ goto unknown; } ScmObj lexmode = Scm_GetPortReaderLexicalMode(port); /* handle #\x1f etc. */ if (cname[0] == 'x' && isxdigit(cname[1])) { const char *nextptr; ScmChar code = Scm_ReadXdigitsFromString(cname+1, namesize-1, 'x', lexmode, FALSE, &nextptr); if (code == SCM_CHAR_INVALID || *nextptr != '\0') goto unknown; return SCM_MAKE_CHAR(code); } /* handle legacy #\uxxxx or #\uxxxxxxxx */ if ((cname[0] == 'u') && isxdigit(cname[1]) && (!SCM_EQ(lexmode, SCM_SYM_STRICT_R7))) { if (namesize >= 5 && namesize <= 9) { const char *nextptr; /* NB: We want to allow variable number of digits, so we pass 'x' as key (instead of 'u') here. */ ScmChar code = Scm_ReadXdigitsFromString(cname+1, namesize-1, 'x', lexmode, FALSE, &nextptr); if (code >= 0 && *nextptr == '\0') { return SCM_MAKE_CHAR(code); } } /* if we come here, it's an error. */ Scm_ReadError(port, "Bad UCS character code: #\\%s", cname); } struct char_name *cntab = char_names; while (cntab->name) { if (cntab->size == (unsigned)namesize && strncmp(cntab->name, cname, namesize) == 0) { return cntab->ch; } cntab++; } unknown: Scm_ReadError(port, "Unknown character name: #\\%A", name); } } return SCM_UNDEFINED; /* dummy */ } /*---------------------------------------------------------------- * Symbols and Numbers */ /* Reads a sequence of word-constituent characters from PORT, and returns ScmString. INITIAL may be a readahead character, or SCM_CHAR_INVALID if there's none. TEMP_CASE_FOLD turns on case-fold mode regardless of the read context setting. INCLUDE_HASH_SIGN allows '#' to appear in the word. */ static ScmObj read_word(ScmPort *port, ScmChar initial, ScmReadContext *ctx SCM_UNUSED, int temp_case_fold, int include_hash_sign) { int case_fold = temp_case_fold || SCM_PORT_CASE_FOLDING(port); ScmDString ds; Scm_DStringInit(&ds); if (initial != SCM_CHAR_INVALID) { if (case_fold && char_word_case_fold(initial)) initial = tolower(initial); SCM_DSTRING_PUTC(&ds, initial); } for (;;) { int c = Scm_GetcUnsafe(port); if (c == EOF || !char_word_constituent(c, include_hash_sign)) { Scm_UngetcUnsafe(c, port); return Scm_DStringGet(&ds, 0); } if (case_fold && char_word_case_fold(c)) c = tolower(c); SCM_DSTRING_PUTC(&ds, c); } } /* Kaveat: We don't allow '#' in symbols, but we need to read '#' for numbers. To allow weird identifers like '1+', we need to read the word as a number fist and convert it to a symbol when the read word can't be interpreted as a number. For the consistency, we read with '#' and then check it in read_symbol, too. */ static void check_valid_symbol(ScmString *s) { ScmObj r = Scm_StringScanChar(s, SCM_CHAR('#'), SCM_STRING_SCAN_INDEX); if (!SCM_FALSEP(r)) { Scm_Error("invalid symbol name: %S", SCM_OBJ(s)); } } /* Read a symbol starting with INITIAL (assuming unescaped), interned. */ static ScmObj read_symbol(ScmPort *port, ScmChar initial, ScmReadContext *ctx) { ScmString *s = SCM_STRING(read_word(port, initial, ctx, FALSE, TRUE)); check_valid_symbol(s); return Scm_Intern(s); } /* This is called when a symbol (either bare or escaped) is expected starting with INITIAL char; e.g. after #: or #!. An error is thrown if a symbol isn't immediately followed. PRECEDING is used in the error message. */ static ScmObj read_immediate_symbol(ScmPort *port, ScmChar initial, int interned, const char *preceding, ScmReadContext *ctx) { if (initial == '|') { return read_escaped_symbol(port, initial, interned, ctx); } else if (char_word_constituent(initial, FALSE)) { ScmString *s = SCM_STRING(read_word(port, initial, ctx, FALSE, TRUE)); check_valid_symbol(s); /* we have to exclude numbers. a bit ugly - call for cleanup */ if (!(isdigit(initial) || initial == '+' || initial == '-') || SCM_FALSEP(Scm_StringToNumber(s, 10, 0))) { return Scm_MakeSymbol(s, interned); } } /* If we come here, we have invalid syntax. */ if (initial == EOF) { Scm_ReadError(port, "'%s' followed by nothing", preceding); } else { Scm_ReadError(port, "invalid %s syntax near '%s%C'", preceding, preceding, initial); } return SCM_UNDEFINED; /* dummy */ } static ScmObj read_number(ScmPort *port, ScmChar initial, int radix, ScmReadContext *ctx) { ScmString *s = SCM_STRING(read_word(port, initial, ctx, FALSE, TRUE)); u_long flags = radix >=2 ? SCM_NUMBER_FORMAT_ALT_RADIX : 0; if (SCM_EQ(Scm_GetPortReaderLexicalMode(port), SCM_SYM_STRICT_R7)) { flags |= SCM_NUMBER_FORMAT_STRICT_R7RS; } int default_radix = radix >= 2? radix : 10; ScmObj num = Scm_StringToNumber(s, default_radix, flags); if (num == SCM_FALSE) { if (radix >= 2) { /* In this case, we've read #r syntax */ Scm_ReadError(port, "bad numeric format: \"#%dr%A\"", radix, s); } else { Scm_ReadError(port, "bad numeric format: %S", s); } } return num; } static ScmObj read_symbol_or_number(ScmPort *port, ScmChar initial, ScmReadContext *ctx) { ScmString *s = SCM_STRING(read_word(port, initial, ctx, FALSE, TRUE)); u_long flags = 0; if (SCM_EQ(Scm_GetPortReaderLexicalMode(port), SCM_SYM_STRICT_R7)) { flags |= SCM_NUMBER_FORMAT_STRICT_R7RS; } ScmObj num = Scm_StringToNumber(s, 10, flags); if (num != SCM_FALSE) return num; check_valid_symbol(s); return Scm_Intern(s); } static ScmObj read_keyword(ScmPort *port, ScmReadContext *ctx) { int c2 = Scm_GetcUnsafe(port); if (c2 == '|') { ScmObj name = read_escaped_symbol(port, c2, FALSE, ctx); /* read as uninterned */ return Scm_MakeKeyword(SCM_SYMBOL_NAME(name)); } else { Scm_UngetcUnsafe(c2, port); ScmObj name = read_word(port, SCM_CHAR_INVALID, ctx, FALSE, FALSE); return Scm_MakeKeyword(SCM_STRING(name)); } } static ScmObj read_escaped_symbol(ScmPort *port, ScmChar delim, int interned, ScmReadContext *ctx SCM_UNUSED) { ScmDString ds; Scm_DStringInit(&ds); ScmObj xmode = Scm_GetPortReaderLexicalMode(port); for (;;) { int c = Scm_GetcUnsafe(port); if (c == EOF) { goto err; } else if (c == delim) { ScmString *s = SCM_STRING(Scm_DStringGet(&ds, 0)); return Scm_MakeSymbol(s, interned); } else if (c == '\\') { /* CL-style single escape */ c = Scm_GetcUnsafe(port); if (c == EOF) goto err; if (SCM_EQ(xmode, SCM_SYM_LEGACY)) { SCM_DSTRING_PUTC(&ds, c); } else { switch (c) { case 'x': { /* R7RS-style hex escape. */ ScmObj bad = Scm_ReadXdigitsFromPort(port, 'x', xmode, FALSE, &ds); if (SCM_STRINGP(bad)) { Scm_ReadError(port, "invalid hex escape in a symbol literal: \\x%C", bad); } break; } case '\\': case '|': SCM_DSTRING_PUTC(&ds, c); break; case 'a': SCM_DSTRING_PUTC(&ds, '\a'); break; case 'b': SCM_DSTRING_PUTC(&ds, '\b'); break; case 't': SCM_DSTRING_PUTC(&ds, '\t'); break; case 'n': SCM_DSTRING_PUTC(&ds, '\n'); break; case 'r': SCM_DSTRING_PUTC(&ds, '\r'); break; default: if (SCM_EQ(xmode, SCM_SYM_STRICT_R7)) { Scm_ReadError(port, "invalid backslash-escape in a symbol literal: \\%A", SCM_MAKE_CHAR(c)); } else { SCM_DSTRING_PUTC(&ds, c); } } } } else { SCM_DSTRING_PUTC(&ds, c); } } err: Scm_ReadError(port, "unterminated escaped symbol: |%s ...", Scm_DStringGetz(&ds)); return SCM_UNDEFINED; /* dummy */ } /*---------------------------------------------------------------- * Regexp & charset */ /* gauche extension : #/regexp/ */ static ScmObj read_regexp(ScmPort *port) { ScmDString ds; Scm_DStringInit(&ds); for (;;) { ScmChar c = Scm_GetcUnsafe(port); if (c == SCM_CHAR_INVALID) { Scm_ReadError(port, "unterminated literal regexp"); } if (c == '\\') { /* NB: We "eat" a backslash before '/', since it is only dealt with the reader and nothing to do with regexp parser itself. */ ScmChar c1 = Scm_GetcUnsafe(port); if (c1 == SCM_CHAR_INVALID) { Scm_ReadError(port, "unterminated literal regexp"); } if (c1 != '/') SCM_DSTRING_PUTC(&ds, c); SCM_DSTRING_PUTC(&ds, c1); } else if (c == '/') { /* Read one more char to see if we have a flag */ int flags = 0; c = Scm_GetcUnsafe(port); if (c == 'i') flags |= SCM_REGEXP_CASE_FOLD; else Scm_UngetcUnsafe(c, port); return Scm_RegComp(SCM_STRING(Scm_DStringGet(&ds, 0)), flags); } else { SCM_DSTRING_PUTC(&ds, c); } } } /* gauche extension : #[charset] */ static ScmObj read_charset(ScmPort *port) { ScmCharSet *cs = SCM_CHAR_SET(Scm_CharSetRead(port, NULL, TRUE, FALSE)); return Scm_CharSetFreezeX(cs); /* literal is immutable */ } /*---------------------------------------------------------------- * Numeric prefixed (#N#, #N=, #Nr) */ static ScmObj read_num_prefixed(ScmPort *port, ScmChar ch, ScmReadContext *ctx) { ScmObj e = SCM_UNBOUND; ScmChar ch2; ScmSize nread; long prefix = Scm_ReadDigitsAsLong(port, ch, 10, &ch2, &nread); if (ch2 == EOF) Scm_ReadError(port, "unterminated reference form (#digits)"); if (prefix < 0) Scm_ReadError(port, "#-prefix number overflow"); switch (ch2) { case '#': /* #digit# - back reference */ if (ctx->table == NULL || SCM_UNBOUNDP(e = Scm_HashTableRef(ctx->table, Scm_MakeInteger(prefix), SCM_UNBOUND))) { Scm_ReadError(port, "invalid reference number in #%d#", prefix); } if (SCM_READ_REFERENCE_P(e) && SCM_READ_REFERENCE_REALIZED(e)) { return SCM_READ_REFERENCE(e)->value; } else { return e; } case '=': /* #digit= - register */ { ScmObj ref = Scm_MakeReadReference(); if (ctx->table == NULL) { ctx->table = SCM_HASH_TABLE(Scm_MakeHashTableSimple(SCM_HASH_EQV, 0)); } if (!SCM_UNBOUNDP(Scm_HashTableRef(ctx->table, Scm_MakeInteger(prefix), SCM_UNBOUND))) { Scm_ReadError(port, "duplicate back-reference number in #%d=", prefix); } ref_register(ctx, ref, prefix); ScmObj val = read_item(port, ctx); if (ref == val) { /* an edge case: #0=#0# */ Scm_ReadError(port, "indeterminate read reference: #%d=#%d#", prefix, prefix); } SCM_READ_REFERENCE(ref)->value = val; return val; } case 'r': case 'R': /* #digitR - radix */ if (SCM_EQ(Scm_GetPortReaderLexicalMode(port), SCM_SYM_STRICT_R7)) { Scm_ReadError(port, "Radix prefix isn't allowed in strict R7RS mode."); } if (prefix < SCM_RADIX_MIN || prefix > SCM_RADIX_MAX) { Scm_ReadError(port, "Radix prefix out of range: radix in #R must be between %d and %d inclusive, but got: %d", SCM_RADIX_MIN, SCM_RADIX_MAX, prefix); return SCM_UNDEFINED; /* dummy */ } else { int ch = Scm_GetcUnsafe(port); if (ch == EOF) { Scm_ReadError(port, "Premature end of radix-prefixed number: #%dr", prefix); } return read_number(port, ch, prefix, ctx); } default: Scm_ReadError(port, "invalid numeric prefix (#, =, r or R is expected) : #%d%A", prefix, SCM_MAKE_CHAR(ch2)); return SCM_UNDEFINED; } } /*---------------------------------------------------------------- * SRFI-10 support */ /* NB: The 'module' argument of DefineReaderCtor and GetReaderCtor may be used in future to make reader-ctor binding associated with modules. For now, it is not used and the caller should pass SCM_FALSE. */ ScmObj Scm_DefineReaderCtor(ScmObj symbol, ScmObj proc, ScmObj finisher, ScmObj module SCM_UNUSED /*reserved*/) { if (!SCM_PROCEDUREP(proc)) { Scm_Error("procedure required, but got %S", proc); } ScmObj pair = Scm_Cons(proc, finisher); (void)SCM_INTERNAL_MUTEX_LOCK(readCtorData.mutex); Scm_HashTableSet(readCtorData.table, symbol, pair, 0); (void)SCM_INTERNAL_MUTEX_UNLOCK(readCtorData.mutex); return SCM_UNDEFINED; } ScmObj Scm_GetReaderCtor(ScmObj symbol, ScmObj module SCM_UNUSED/*reserved*/) { (void)SCM_INTERNAL_MUTEX_LOCK(readCtorData.mutex); ScmObj r = Scm_HashTableRef(readCtorData.table, symbol, SCM_FALSE); (void)SCM_INTERNAL_MUTEX_UNLOCK(readCtorData.mutex); return r; } static ScmObj read_sharp_comma(ScmPort *port, ScmReadContext *ctx) { ScmChar next = Scm_GetcUnsafe(port); if (next != '(') { Scm_ReadError(port, "bad #,-form: '(' should be followed, but got %C", next); } int has_ref; ScmSize line = -1; if (ctx->flags & RCTX_SOURCE_INFO) line = Scm_PortLine(port); ScmObj form = read_list_int(port, ')', ctx, &has_ref, line); int len = Scm_Length(form); if (len <= 0) { Scm_ReadError(port, "bad #,-form: #,%S", form); } ScmObj r = process_sharp_comma(port, line, SCM_CAR(form), SCM_CDR(form), ctx, has_ref); return r; } static ScmObj process_sharp_comma(ScmPort *port, int line, ScmObj key, ScmObj args, ScmReadContext *ctx, int has_ref) { if (ctx->flags & RCTX_DISABLE_CTOR) return SCM_FALSE; (void)SCM_INTERNAL_MUTEX_LOCK(readCtorData.mutex); ScmObj e = Scm_HashTableRef(readCtorData.table, key, SCM_FALSE); (void)SCM_INTERNAL_MUTEX_UNLOCK(readCtorData.mutex); if (!SCM_PAIRP(e)) Scm_ReadError(port, "unknown #,-key: %S", key); ScmObj r = Scm_ApplyRec(SCM_CAR(e), args); if (SCM_PAIRP(r) && (ctx->flags & RCTX_SOURCE_INFO) && line >= 0) { r = Scm_ExtendedCons(SCM_CAR(r), SCM_CDR(r)); Scm_PairAttrSet(SCM_PAIR(r), SCM_SYM_SOURCE_INFO, SCM_LIST2(Scm_PortName(port), SCM_MAKE_INT(line))); } if (has_ref) ref_push(ctx, r, SCM_CDR(e)); return r; } static ScmObj reader_ctor(ScmObj *args, int nargs, void *data SCM_UNUSED) { ScmObj optarg = (nargs > 2? args[2] : SCM_FALSE); return Scm_DefineReaderCtor(args[0], args[1], optarg, SCM_FALSE); } /*---------------------------------------------------------------- * #!-support */ ScmObj Scm_DefineReaderDirective(ScmObj symbol, ScmObj proc) { (void)SCM_INTERNAL_MUTEX_LOCK(hashBangData.mutex); Scm_HashTableSet(hashBangData.table, symbol, proc, 0); (void)SCM_INTERNAL_MUTEX_UNLOCK(hashBangData.mutex); return SCM_UNDEFINED; } static ScmObj read_shebang(ScmPort *port, ScmReadContext *ctx) { /* If '#!' appears at the beginning of the input port, and it is followed by '/' or a space or a newline, then we consider it as the beginning of shebang and discard the entire line. Otherwise, we take this as #! directive as specified in R6RS, and calls appropriate handler. R6RS is actually not very clean at this corner. It requires distinct modes to parse a script (which always begins with shebang) and to parse R6RS program (in which '#!' always introduces #! token). There's no way for just one parser that strictly covers both situation. */ int c2 = Scm_GetcUnsafe(port); if (Scm_PortBytes(port) == 3 && (c2 == '/' || c2 == ' ' || c2 == '\n')) { /* shebang */ for (;;) { if (c2 == '\n') return SCM_UNDEFINED; if (c2 == EOF) return SCM_EOF; c2 = Scm_GetcUnsafe(port); } /*NOTREACHED*/ } else { ScmObj id = read_immediate_symbol(port, c2, TRUE, "#!", ctx); (void)SCM_INTERNAL_MUTEX_LOCK(hashBangData.mutex); ScmObj e = Scm_HashTableRef(hashBangData.table, id, SCM_FALSE); (void)SCM_INTERNAL_MUTEX_UNLOCK(hashBangData.mutex); if (SCM_FALSEP(e)) { Scm_Warn("Ignoring unrecognized hash-bang directive: #!%S", id); return SCM_UNDEFINED; } /* Reader directive may return zero or one value. When it returns no values, we call Scm_VMSetResult to adjust the number of values. */ ScmObj r = Scm_ApplyRec3(e, id, SCM_OBJ(port), SCM_OBJ(ctx)); if (Scm_VMGetNumResults(Scm_VM()) == 1) return r; else { Scm_VMSetResult(SCM_UNDEFINED); return SCM_UNDEFINED; } } } /*---------------------------------------------------------------- * #f, #t, #false, #true, and UVector literals */ /* Pre-0.9.4 reader. #t and #f delimit themselves (except '1', '3' or '6' follows '#f'.) I doubt any code breaks if we change that, but there may be a data files around that somehow relies on this behavior. So we keep this in 'legacy' reader mode. */ static ScmObj read_sharp_word_legacy(ScmPort *port, char ch, ScmReadContext *ctx) { ScmChar c1, c2 = SCM_CHAR_INVALID; char *tag = NULL; if (ch == 't') return SCM_TRUE; c1 = Scm_GetcUnsafe(port); if (ch == 'f') { if (c1 != '1' && c1 != '3' && c1 != '6') { Scm_UngetcUnsafe(c1, port); return SCM_FALSE; } c2 = Scm_GetcUnsafe(port); if (c1 == '3' && c2 == '2') tag = "f32"; else if (c1 == '6' && c2 == '4') tag = "f64"; else if (c1 == '1' && c2 == '6') tag = "f16"; } else { if (c1 == '8') tag = (ch == 's')? "s8" : "u8"; else if (c1 == '1') { c2 = Scm_GetcUnsafe(port); if (c2 == '6') tag = (ch == 's')? "s16" : "u16"; } else if (c1 == '3') { c2 = Scm_GetcUnsafe(port); if (c2 == '2') tag = (ch == 's')? "s32" : "u32"; } else if (c1 == '6') { c2 = Scm_GetcUnsafe(port); if (c2 == '4') tag = (ch == 's')? "s64" : "u64"; } } if (tag == NULL) { char buf[SCM_CHAR_MAX_BYTES*4], *bufp = buf; *bufp++ = ch; SCM_CHAR_PUT(bufp, c1); bufp += SCM_CHAR_NBYTES(c1); if (c2 != SCM_CHAR_INVALID) { SCM_CHAR_PUT(bufp, c2); bufp += SCM_CHAR_NBYTES(c2); } *bufp = '\0'; Scm_ReadError(port, "invalid uniform vector tag: %s", buf); } return Scm_ReadUVector(port, tag, ctx); } /* A 'new' version, friendly to R7RS */ static ScmObj read_sharp_word_1(ScmPort *port, char ch, ScmReadContext *ctx) { ScmString *s = SCM_STRING(read_word(port, ch, ctx, TRUE, FALSE)); const char *w = Scm_GetStringConst(s); const char *tag = NULL; switch (ch) { case 'c': if (strcmp(w, "c32") == 0 || strcmp(w, "c64") == 0 || strcmp(w, "c128") == 0) { tag = w; } break; case 'f': if (strcmp(w, "f16") == 0 || strcmp(w, "f32") == 0 || strcmp(w, "f64") == 0) { tag = w; } else if (w[1] == '\0' || strcmp(w, "false") == 0) { return SCM_FALSE; } break; case 's': if (strcmp(w, "s8") == 0 || strcmp(w, "s16") == 0 || strcmp(w, "s32") == 0 || strcmp(w, "s64") == 0) { tag = w; } break; case 'u': if (strcmp(w, "u8") == 0 || strcmp(w, "u16") == 0 || strcmp(w, "u32") == 0 || strcmp(w, "u64") == 0) { tag = w; } break; case 't': if (w[1] == '\0' || strcmp(w, "true") == 0) { return SCM_TRUE; } } if (tag == NULL) { Scm_ReadError(port, "invalid #-token: #%s", w); } return Scm_ReadUVector(port, tag, ctx); } static ScmObj read_sharp_word(ScmPort *port, char ch, ScmReadContext *ctx) { if (SCM_EQ(Scm_GetPortReaderLexicalMode(port), SCM_SYM_LEGACY)) { return read_sharp_word_legacy(port, ch, ctx); } else { return read_sharp_word_1(port, ch, ctx); } } /*---------------------------------------------------------------- * #**"....", #*1010010... */ static ScmObj read_sharp_asterisk(ScmPort *port, ScmReadContext *ctx) { /* We used to use #*"..." for incomplete string literals, but it turned out it can be read as an zero-length bitvector #* followed by an ordinary string (since a double quote is a in R7RS). So we changed the literal to #**"..." (yeah, ugly, but incomplete strings are anomalies and shouldn't be used often). The reader behavior can be customized by reader-lexical-mode: permissive (default) - For the next release, #*"..." is read as an incomplete string, without warning. In future, it warns. legacy - #*"..." is read as an incomplete string, without warning. warn-legacy - #*"..." is read as an incomplete string, but warns. strict-r7 - #*"..." is read as a zero-length bitvector followed by a string. #**"..." is rejected. */ int c = Scm_GetcUnsafe(port); if (c == '*') { c = Scm_GetcUnsafe(port); if (c == '"') return read_string(port, TRUE, ctx); Scm_ReadError(port, "Invalid #* syntax: #**%C", c); } if (c == '"') { ScmObj m = Scm_GetPortReaderLexicalMode(port); if (SCM_EQ(m, SCM_SYM_PERMISSIVE)) { /* TRANSIENT: We switch this default behavior to warn in future */ return read_string(port, TRUE, ctx); } if (SCM_EQ(m, SCM_SYM_LEGACY)) { return read_string(port, TRUE, ctx); } if (SCM_EQ(m, SCM_SYM_WARN_LEGACY)) { Scm_Warn("Deprecated incomplete string syntax #*\"...\" at %A:%ld", Scm_PortName(port), Scm_PortLine(port)); return read_string(port, TRUE, ctx); } if (SCM_EQ(m, SCM_SYM_STRICT_R7)) { Scm_UngetcUnsafe(c, port); return Scm_MakeBitvector(0, SCM_FALSE); } Scm_Panic("invalid reader-lexical-mode"); } ScmDString ds; Scm_DStringInit(&ds); for (;;) { switch (c) { case EOF: goto finish; case '(': case ')': case '"': case '|': case ';': Scm_UngetcUnsafe(c, port); goto finish; case '0': case '1': Scm_DStringPutc(&ds, c); break; default: if ((SCM_CHAR_ASCII_P(c) && isspace(c)) || (!SCM_CHAR_ASCII_P(c) && SCM_CHAR_EXTRA_WHITESPACE(c))) { Scm_UngetcUnsafe(c, port); goto finish; } Scm_ReadError(port, "Invalid char in bitvector literal: %C", c); } c = Scm_GetcUnsafe(port); } finish:; ScmObj s = Scm_DStringGet(&ds, 0); return Scm_StringToBitvector(SCM_STRING(s), FALSE); } /* OBSOLETED: gauche.uvector used to call this to set up reader pointer. Now it is read in src/vector.c. We keep this entry for ABI compatibility. Remove on 1.0 release. */ void Scm__InstallReadUvectorHook(ScmObj (*f)(ScmPort*, const char*, ScmReadContext *) SCM_UNUSED) { } /*---------------------------------------------------------------- * Initialization */ void Scm__InitRead(void) { readCtorData.table = SCM_HASH_TABLE(Scm_MakeHashTableSimple(SCM_HASH_EQ, 0)); (void)SCM_INTERNAL_MUTEX_INIT(readCtorData.mutex); Scm_DefineReaderCtor(SCM_SYM_DEFINE_READER_CTOR, Scm_MakeSubr(reader_ctor, NULL, 2, 1, SCM_SYM_DEFINE_READER_CTOR), SCM_FALSE, SCM_FALSE); hashBangData.table = SCM_HASH_TABLE(Scm_MakeHashTableSimple(SCM_HASH_EQ, 0)); (void)SCM_INTERNAL_MUTEX_INIT(hashBangData.mutex); defaultReadContext = Scm_MakePrimitiveParameter(SCM_CLASS_PRIMITIVE_PARAMETER, SCM_FALSE, SCM_OBJ(make_read_context(NULL)), 0); }