Some improvements in vc
[emacs.git] / src / lread.c
bloba42c1f60c9555a2b0c7327ec0d9606af57e9f573
1 /* Lisp parsing and input streams.
3 Copyright (C) 1985-1989, 1993-1995, 1997-2016 Free Software Foundation,
4 Inc.
6 This file is part of GNU Emacs.
8 GNU Emacs is free software: you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation, either version 3 of the License, or (at
11 your option) any later version.
13 GNU Emacs is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with GNU Emacs. If not, see <http://www.gnu.org/licenses/>. */
21 /* Tell globals.h to define tables needed by init_obarray. */
22 #define DEFINE_SYMBOLS
24 #include <config.h>
25 #include "sysstdio.h"
26 #include <sys/types.h>
27 #include <sys/stat.h>
28 #include <sys/file.h>
29 #include <errno.h>
30 #include <limits.h> /* For CHAR_BIT. */
31 #include <math.h>
32 #include <stat-time.h>
33 #include "lisp.h"
34 #include "dispextern.h"
35 #include "intervals.h"
36 #include "character.h"
37 #include "buffer.h"
38 #include "charset.h"
39 #include "coding.h"
40 #include <epaths.h>
41 #include "commands.h"
42 #include "keyboard.h"
43 #include "systime.h"
44 #include "termhooks.h"
45 #include "blockinput.h"
46 #include <c-ctype.h>
48 #ifdef MSDOS
49 #include "msdos.h"
50 #endif
52 #ifdef HAVE_NS
53 #include "nsterm.h"
54 #endif
56 #include <unistd.h>
58 #ifdef HAVE_SETLOCALE
59 #include <locale.h>
60 #endif /* HAVE_SETLOCALE */
62 #include <fcntl.h>
64 #ifdef HAVE_FSEEKO
65 #define file_offset off_t
66 #define file_tell ftello
67 #else
68 #define file_offset long
69 #define file_tell ftell
70 #endif
72 /* The association list of objects read with the #n=object form.
73 Each member of the list has the form (n . object), and is used to
74 look up the object for the corresponding #n# construct.
75 It must be set to nil before all top-level calls to read0. */
76 static Lisp_Object read_objects;
78 /* File for get_file_char to read from. Use by load. */
79 static FILE *instream;
81 /* For use within read-from-string (this reader is non-reentrant!!) */
82 static ptrdiff_t read_from_string_index;
83 static ptrdiff_t read_from_string_index_byte;
84 static ptrdiff_t read_from_string_limit;
86 /* Number of characters read in the current call to Fread or
87 Fread_from_string. */
88 static EMACS_INT readchar_count;
90 /* This contains the last string skipped with #@. */
91 static char *saved_doc_string;
92 /* Length of buffer allocated in saved_doc_string. */
93 static ptrdiff_t saved_doc_string_size;
94 /* Length of actual data in saved_doc_string. */
95 static ptrdiff_t saved_doc_string_length;
96 /* This is the file position that string came from. */
97 static file_offset saved_doc_string_position;
99 /* This contains the previous string skipped with #@.
100 We copy it from saved_doc_string when a new string
101 is put in saved_doc_string. */
102 static char *prev_saved_doc_string;
103 /* Length of buffer allocated in prev_saved_doc_string. */
104 static ptrdiff_t prev_saved_doc_string_size;
105 /* Length of actual data in prev_saved_doc_string. */
106 static ptrdiff_t prev_saved_doc_string_length;
107 /* This is the file position that string came from. */
108 static file_offset prev_saved_doc_string_position;
110 /* True means inside a new-style backquote
111 with no surrounding parentheses.
112 Fread initializes this to false, so we need not specbind it
113 or worry about what happens to it when there is an error. */
114 static bool new_backquote_flag;
116 /* A list of file names for files being loaded in Fload. Used to
117 check for recursive loads. */
119 static Lisp_Object Vloads_in_progress;
121 static int read_emacs_mule_char (int, int (*) (int, Lisp_Object),
122 Lisp_Object);
124 static void readevalloop (Lisp_Object, FILE *, Lisp_Object, bool,
125 Lisp_Object, Lisp_Object,
126 Lisp_Object, Lisp_Object);
128 /* Functions that read one byte from the current source READCHARFUN
129 or unreads one byte. If the integer argument C is -1, it returns
130 one read byte, or -1 when there's no more byte in the source. If C
131 is 0 or positive, it unreads C, and the return value is not
132 interesting. */
134 static int readbyte_for_lambda (int, Lisp_Object);
135 static int readbyte_from_file (int, Lisp_Object);
136 static int readbyte_from_string (int, Lisp_Object);
138 /* Handle unreading and rereading of characters.
139 Write READCHAR to read a character,
140 UNREAD(c) to unread c to be read again.
142 These macros correctly read/unread multibyte characters. */
144 #define READCHAR readchar (readcharfun, NULL)
145 #define UNREAD(c) unreadchar (readcharfun, c)
147 /* Same as READCHAR but set *MULTIBYTE to the multibyteness of the source. */
148 #define READCHAR_REPORT_MULTIBYTE(multibyte) readchar (readcharfun, multibyte)
150 /* When READCHARFUN is Qget_file_char, Qget_emacs_mule_file_char,
151 Qlambda, or a cons, we use this to keep an unread character because
152 a file stream can't handle multibyte-char unreading. The value -1
153 means that there's no unread character. */
154 static int unread_char;
156 static int
157 readchar (Lisp_Object readcharfun, bool *multibyte)
159 Lisp_Object tem;
160 register int c;
161 int (*readbyte) (int, Lisp_Object);
162 unsigned char buf[MAX_MULTIBYTE_LENGTH];
163 int i, len;
164 bool emacs_mule_encoding = 0;
166 if (multibyte)
167 *multibyte = 0;
169 readchar_count++;
171 if (BUFFERP (readcharfun))
173 register struct buffer *inbuffer = XBUFFER (readcharfun);
175 ptrdiff_t pt_byte = BUF_PT_BYTE (inbuffer);
177 if (! BUFFER_LIVE_P (inbuffer))
178 return -1;
180 if (pt_byte >= BUF_ZV_BYTE (inbuffer))
181 return -1;
183 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
185 /* Fetch the character code from the buffer. */
186 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, pt_byte);
187 BUF_INC_POS (inbuffer, pt_byte);
188 c = STRING_CHAR (p);
189 if (multibyte)
190 *multibyte = 1;
192 else
194 c = BUF_FETCH_BYTE (inbuffer, pt_byte);
195 if (! ASCII_CHAR_P (c))
196 c = BYTE8_TO_CHAR (c);
197 pt_byte++;
199 SET_BUF_PT_BOTH (inbuffer, BUF_PT (inbuffer) + 1, pt_byte);
201 return c;
203 if (MARKERP (readcharfun))
205 register struct buffer *inbuffer = XMARKER (readcharfun)->buffer;
207 ptrdiff_t bytepos = marker_byte_position (readcharfun);
209 if (bytepos >= BUF_ZV_BYTE (inbuffer))
210 return -1;
212 if (! NILP (BVAR (inbuffer, enable_multibyte_characters)))
214 /* Fetch the character code from the buffer. */
215 unsigned char *p = BUF_BYTE_ADDRESS (inbuffer, bytepos);
216 BUF_INC_POS (inbuffer, bytepos);
217 c = STRING_CHAR (p);
218 if (multibyte)
219 *multibyte = 1;
221 else
223 c = BUF_FETCH_BYTE (inbuffer, bytepos);
224 if (! ASCII_CHAR_P (c))
225 c = BYTE8_TO_CHAR (c);
226 bytepos++;
229 XMARKER (readcharfun)->bytepos = bytepos;
230 XMARKER (readcharfun)->charpos++;
232 return c;
235 if (EQ (readcharfun, Qlambda))
237 readbyte = readbyte_for_lambda;
238 goto read_multibyte;
241 if (EQ (readcharfun, Qget_file_char))
243 readbyte = readbyte_from_file;
244 goto read_multibyte;
247 if (STRINGP (readcharfun))
249 if (read_from_string_index >= read_from_string_limit)
250 c = -1;
251 else if (STRING_MULTIBYTE (readcharfun))
253 if (multibyte)
254 *multibyte = 1;
255 FETCH_STRING_CHAR_ADVANCE_NO_CHECK (c, readcharfun,
256 read_from_string_index,
257 read_from_string_index_byte);
259 else
261 c = SREF (readcharfun, read_from_string_index_byte);
262 read_from_string_index++;
263 read_from_string_index_byte++;
265 return c;
268 if (CONSP (readcharfun) && STRINGP (XCAR (readcharfun)))
270 /* This is the case that read_vector is reading from a unibyte
271 string that contains a byte sequence previously skipped
272 because of #@NUMBER. The car part of readcharfun is that
273 string, and the cdr part is a value of readcharfun given to
274 read_vector. */
275 readbyte = readbyte_from_string;
276 if (EQ (XCDR (readcharfun), Qget_emacs_mule_file_char))
277 emacs_mule_encoding = 1;
278 goto read_multibyte;
281 if (EQ (readcharfun, Qget_emacs_mule_file_char))
283 readbyte = readbyte_from_file;
284 emacs_mule_encoding = 1;
285 goto read_multibyte;
288 tem = call0 (readcharfun);
290 if (NILP (tem))
291 return -1;
292 return XINT (tem);
294 read_multibyte:
295 if (unread_char >= 0)
297 c = unread_char;
298 unread_char = -1;
299 return c;
301 c = (*readbyte) (-1, readcharfun);
302 if (c < 0)
303 return c;
304 if (multibyte)
305 *multibyte = 1;
306 if (ASCII_CHAR_P (c))
307 return c;
308 if (emacs_mule_encoding)
309 return read_emacs_mule_char (c, readbyte, readcharfun);
310 i = 0;
311 buf[i++] = c;
312 len = BYTES_BY_CHAR_HEAD (c);
313 while (i < len)
315 c = (*readbyte) (-1, readcharfun);
316 if (c < 0 || ! TRAILING_CODE_P (c))
318 while (--i > 1)
319 (*readbyte) (buf[i], readcharfun);
320 return BYTE8_TO_CHAR (buf[0]);
322 buf[i++] = c;
324 return STRING_CHAR (buf);
327 #define FROM_FILE_P(readcharfun) \
328 (EQ (readcharfun, Qget_file_char) \
329 || EQ (readcharfun, Qget_emacs_mule_file_char))
331 static void
332 skip_dyn_bytes (Lisp_Object readcharfun, ptrdiff_t n)
334 if (FROM_FILE_P (readcharfun))
336 block_input (); /* FIXME: Not sure if it's needed. */
337 fseek (instream, n, SEEK_CUR);
338 unblock_input ();
340 else
341 { /* We're not reading directly from a file. In that case, it's difficult
342 to reliably count bytes, since these are usually meant for the file's
343 encoding, whereas we're now typically in the internal encoding.
344 But luckily, skip_dyn_bytes is used to skip over a single
345 dynamic-docstring (or dynamic byte-code) which is always quoted such
346 that \037 is the final char. */
347 int c;
348 do {
349 c = READCHAR;
350 } while (c >= 0 && c != '\037');
354 static void
355 skip_dyn_eof (Lisp_Object readcharfun)
357 if (FROM_FILE_P (readcharfun))
359 block_input (); /* FIXME: Not sure if it's needed. */
360 fseek (instream, 0, SEEK_END);
361 unblock_input ();
363 else
364 while (READCHAR >= 0);
367 /* Unread the character C in the way appropriate for the stream READCHARFUN.
368 If the stream is a user function, call it with the char as argument. */
370 static void
371 unreadchar (Lisp_Object readcharfun, int c)
373 readchar_count--;
374 if (c == -1)
375 /* Don't back up the pointer if we're unreading the end-of-input mark,
376 since readchar didn't advance it when we read it. */
378 else if (BUFFERP (readcharfun))
380 struct buffer *b = XBUFFER (readcharfun);
381 ptrdiff_t charpos = BUF_PT (b);
382 ptrdiff_t bytepos = BUF_PT_BYTE (b);
384 if (! NILP (BVAR (b, enable_multibyte_characters)))
385 BUF_DEC_POS (b, bytepos);
386 else
387 bytepos--;
389 SET_BUF_PT_BOTH (b, charpos - 1, bytepos);
391 else if (MARKERP (readcharfun))
393 struct buffer *b = XMARKER (readcharfun)->buffer;
394 ptrdiff_t bytepos = XMARKER (readcharfun)->bytepos;
396 XMARKER (readcharfun)->charpos--;
397 if (! NILP (BVAR (b, enable_multibyte_characters)))
398 BUF_DEC_POS (b, bytepos);
399 else
400 bytepos--;
402 XMARKER (readcharfun)->bytepos = bytepos;
404 else if (STRINGP (readcharfun))
406 read_from_string_index--;
407 read_from_string_index_byte
408 = string_char_to_byte (readcharfun, read_from_string_index);
410 else if (CONSP (readcharfun) && STRINGP (XCAR (readcharfun)))
412 unread_char = c;
414 else if (EQ (readcharfun, Qlambda))
416 unread_char = c;
418 else if (FROM_FILE_P (readcharfun))
420 unread_char = c;
422 else
423 call1 (readcharfun, make_number (c));
426 static int
427 readbyte_for_lambda (int c, Lisp_Object readcharfun)
429 return read_bytecode_char (c >= 0);
433 static int
434 readbyte_from_file (int c, Lisp_Object readcharfun)
436 if (c >= 0)
438 block_input ();
439 ungetc (c, instream);
440 unblock_input ();
441 return 0;
444 block_input ();
445 c = getc (instream);
447 /* Interrupted reads have been observed while reading over the network. */
448 while (c == EOF && ferror (instream) && errno == EINTR)
450 unblock_input ();
451 QUIT;
452 block_input ();
453 clearerr (instream);
454 c = getc (instream);
457 unblock_input ();
459 return (c == EOF ? -1 : c);
462 static int
463 readbyte_from_string (int c, Lisp_Object readcharfun)
465 Lisp_Object string = XCAR (readcharfun);
467 if (c >= 0)
469 read_from_string_index--;
470 read_from_string_index_byte
471 = string_char_to_byte (string, read_from_string_index);
474 if (read_from_string_index >= read_from_string_limit)
475 c = -1;
476 else
477 FETCH_STRING_CHAR_ADVANCE (c, string,
478 read_from_string_index,
479 read_from_string_index_byte);
480 return c;
484 /* Read one non-ASCII character from INSTREAM. The character is
485 encoded in `emacs-mule' and the first byte is already read in
486 C. */
488 static int
489 read_emacs_mule_char (int c, int (*readbyte) (int, Lisp_Object), Lisp_Object readcharfun)
491 /* Emacs-mule coding uses at most 4-byte for one character. */
492 unsigned char buf[4];
493 int len = emacs_mule_bytes[c];
494 struct charset *charset;
495 int i;
496 unsigned code;
498 if (len == 1)
499 /* C is not a valid leading-code of `emacs-mule'. */
500 return BYTE8_TO_CHAR (c);
502 i = 0;
503 buf[i++] = c;
504 while (i < len)
506 c = (*readbyte) (-1, readcharfun);
507 if (c < 0xA0)
509 while (--i > 1)
510 (*readbyte) (buf[i], readcharfun);
511 return BYTE8_TO_CHAR (buf[0]);
513 buf[i++] = c;
516 if (len == 2)
518 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
519 code = buf[1] & 0x7F;
521 else if (len == 3)
523 if (buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_11
524 || buf[0] == EMACS_MULE_LEADING_CODE_PRIVATE_12)
526 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
527 code = buf[2] & 0x7F;
529 else
531 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[0]]);
532 code = ((buf[1] << 8) | buf[2]) & 0x7F7F;
535 else
537 charset = CHARSET_FROM_ID (emacs_mule_charset[buf[1]]);
538 code = ((buf[2] << 8) | buf[3]) & 0x7F7F;
540 c = DECODE_CHAR (charset, code);
541 if (c < 0)
542 Fsignal (Qinvalid_read_syntax,
543 list1 (build_string ("invalid multibyte form")));
544 return c;
548 static Lisp_Object read_internal_start (Lisp_Object, Lisp_Object,
549 Lisp_Object);
550 static Lisp_Object read0 (Lisp_Object);
551 static Lisp_Object read1 (Lisp_Object, int *, bool);
553 static Lisp_Object read_list (bool, Lisp_Object);
554 static Lisp_Object read_vector (Lisp_Object, bool);
556 static Lisp_Object substitute_object_recurse (Lisp_Object, Lisp_Object,
557 Lisp_Object);
558 static void substitute_object_in_subtree (Lisp_Object,
559 Lisp_Object);
560 static void substitute_in_interval (INTERVAL, Lisp_Object);
563 /* Get a character from the tty. */
565 /* Read input events until we get one that's acceptable for our purposes.
567 If NO_SWITCH_FRAME, switch-frame events are stashed
568 until we get a character we like, and then stuffed into
569 unread_switch_frame.
571 If ASCII_REQUIRED, check function key events to see
572 if the unmodified version of the symbol has a Qascii_character
573 property, and use that character, if present.
575 If ERROR_NONASCII, signal an error if the input we
576 get isn't an ASCII character with modifiers. If it's false but
577 ASCII_REQUIRED is true, just re-read until we get an ASCII
578 character.
580 If INPUT_METHOD, invoke the current input method
581 if the character warrants that.
583 If SECONDS is a number, wait that many seconds for input, and
584 return Qnil if no input arrives within that time. */
586 static Lisp_Object
587 read_filtered_event (bool no_switch_frame, bool ascii_required,
588 bool error_nonascii, bool input_method, Lisp_Object seconds)
590 Lisp_Object val, delayed_switch_frame;
591 struct timespec end_time;
593 #ifdef HAVE_WINDOW_SYSTEM
594 if (display_hourglass_p)
595 cancel_hourglass ();
596 #endif
598 delayed_switch_frame = Qnil;
600 /* Compute timeout. */
601 if (NUMBERP (seconds))
603 double duration = extract_float (seconds);
604 struct timespec wait_time = dtotimespec (duration);
605 end_time = timespec_add (current_timespec (), wait_time);
608 /* Read until we get an acceptable event. */
609 retry:
611 val = read_char (0, Qnil, (input_method ? Qnil : Qt), 0,
612 NUMBERP (seconds) ? &end_time : NULL);
613 while (INTEGERP (val) && XINT (val) == -2); /* wrong_kboard_jmpbuf */
615 if (BUFFERP (val))
616 goto retry;
618 /* `switch-frame' events are put off until after the next ASCII
619 character. This is better than signaling an error just because
620 the last characters were typed to a separate minibuffer frame,
621 for example. Eventually, some code which can deal with
622 switch-frame events will read it and process it. */
623 if (no_switch_frame
624 && EVENT_HAS_PARAMETERS (val)
625 && EQ (EVENT_HEAD_KIND (EVENT_HEAD (val)), Qswitch_frame))
627 delayed_switch_frame = val;
628 goto retry;
631 if (ascii_required && !(NUMBERP (seconds) && NILP (val)))
633 /* Convert certain symbols to their ASCII equivalents. */
634 if (SYMBOLP (val))
636 Lisp_Object tem, tem1;
637 tem = Fget (val, Qevent_symbol_element_mask);
638 if (!NILP (tem))
640 tem1 = Fget (Fcar (tem), Qascii_character);
641 /* Merge this symbol's modifier bits
642 with the ASCII equivalent of its basic code. */
643 if (!NILP (tem1))
644 XSETFASTINT (val, XINT (tem1) | XINT (Fcar (Fcdr (tem))));
648 /* If we don't have a character now, deal with it appropriately. */
649 if (!INTEGERP (val))
651 if (error_nonascii)
653 Vunread_command_events = list1 (val);
654 error ("Non-character input-event");
656 else
657 goto retry;
661 if (! NILP (delayed_switch_frame))
662 unread_switch_frame = delayed_switch_frame;
664 #if 0
666 #ifdef HAVE_WINDOW_SYSTEM
667 if (display_hourglass_p)
668 start_hourglass ();
669 #endif
671 #endif
673 return val;
676 DEFUN ("read-char", Fread_char, Sread_char, 0, 3, 0,
677 doc: /* Read a character from the command input (keyboard or macro).
678 It is returned as a number.
679 If the character has modifiers, they are resolved and reflected to the
680 character code if possible (e.g. C-SPC -> 0).
682 If the user generates an event which is not a character (i.e. a mouse
683 click or function key event), `read-char' signals an error. As an
684 exception, switch-frame events are put off until non-character events
685 can be read.
686 If you want to read non-character events, or ignore them, call
687 `read-event' or `read-char-exclusive' instead.
689 If the optional argument PROMPT is non-nil, display that as a prompt.
690 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
691 input method is turned on in the current buffer, that input method
692 is used for reading a character.
693 If the optional argument SECONDS is non-nil, it should be a number
694 specifying the maximum number of seconds to wait for input. If no
695 input arrives in that time, return nil. SECONDS may be a
696 floating-point value. */)
697 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
699 Lisp_Object val;
701 if (! NILP (prompt))
702 message_with_string ("%s", prompt, 0);
703 val = read_filtered_event (1, 1, 1, ! NILP (inherit_input_method), seconds);
705 return (NILP (val) ? Qnil
706 : make_number (char_resolve_modifier_mask (XINT (val))));
709 DEFUN ("read-event", Fread_event, Sread_event, 0, 3, 0,
710 doc: /* Read an event object from the input stream.
711 If the optional argument PROMPT is non-nil, display that as a prompt.
712 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
713 input method is turned on in the current buffer, that input method
714 is used for reading a character.
715 If the optional argument SECONDS is non-nil, it should be a number
716 specifying the maximum number of seconds to wait for input. If no
717 input arrives in that time, return nil. SECONDS may be a
718 floating-point value. */)
719 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
721 if (! NILP (prompt))
722 message_with_string ("%s", prompt, 0);
723 return read_filtered_event (0, 0, 0, ! NILP (inherit_input_method), seconds);
726 DEFUN ("read-char-exclusive", Fread_char_exclusive, Sread_char_exclusive, 0, 3, 0,
727 doc: /* Read a character from the command input (keyboard or macro).
728 It is returned as a number. Non-character events are ignored.
729 If the character has modifiers, they are resolved and reflected to the
730 character code if possible (e.g. C-SPC -> 0).
732 If the optional argument PROMPT is non-nil, display that as a prompt.
733 If the optional argument INHERIT-INPUT-METHOD is non-nil and some
734 input method is turned on in the current buffer, that input method
735 is used for reading a character.
736 If the optional argument SECONDS is non-nil, it should be a number
737 specifying the maximum number of seconds to wait for input. If no
738 input arrives in that time, return nil. SECONDS may be a
739 floating-point value. */)
740 (Lisp_Object prompt, Lisp_Object inherit_input_method, Lisp_Object seconds)
742 Lisp_Object val;
744 if (! NILP (prompt))
745 message_with_string ("%s", prompt, 0);
747 val = read_filtered_event (1, 1, 0, ! NILP (inherit_input_method), seconds);
749 return (NILP (val) ? Qnil
750 : make_number (char_resolve_modifier_mask (XINT (val))));
753 DEFUN ("get-file-char", Fget_file_char, Sget_file_char, 0, 0, 0,
754 doc: /* Don't use this yourself. */)
755 (void)
757 register Lisp_Object val;
758 block_input ();
759 XSETINT (val, getc (instream));
760 unblock_input ();
761 return val;
767 /* Return true if the lisp code read using READCHARFUN defines a non-nil
768 `lexical-binding' file variable. After returning, the stream is
769 positioned following the first line, if it is a comment or #! line,
770 otherwise nothing is read. */
772 static bool
773 lisp_file_lexically_bound_p (Lisp_Object readcharfun)
775 int ch = READCHAR;
777 if (ch == '#')
779 ch = READCHAR;
780 if (ch != '!')
782 UNREAD (ch);
783 UNREAD ('#');
784 return 0;
786 while (ch != '\n' && ch != EOF)
787 ch = READCHAR;
788 if (ch == '\n') ch = READCHAR;
789 /* It is OK to leave the position after a #! line, since
790 that is what read1 does. */
793 if (ch != ';')
794 /* The first line isn't a comment, just give up. */
796 UNREAD (ch);
797 return 0;
799 else
800 /* Look for an appropriate file-variable in the first line. */
802 bool rv = 0;
803 enum {
804 NOMINAL, AFTER_FIRST_DASH, AFTER_ASTERIX
805 } beg_end_state = NOMINAL;
806 bool in_file_vars = 0;
808 #define UPDATE_BEG_END_STATE(ch) \
809 if (beg_end_state == NOMINAL) \
810 beg_end_state = (ch == '-' ? AFTER_FIRST_DASH : NOMINAL); \
811 else if (beg_end_state == AFTER_FIRST_DASH) \
812 beg_end_state = (ch == '*' ? AFTER_ASTERIX : NOMINAL); \
813 else if (beg_end_state == AFTER_ASTERIX) \
815 if (ch == '-') \
816 in_file_vars = !in_file_vars; \
817 beg_end_state = NOMINAL; \
820 /* Skip until we get to the file vars, if any. */
823 ch = READCHAR;
824 UPDATE_BEG_END_STATE (ch);
826 while (!in_file_vars && ch != '\n' && ch != EOF);
828 while (in_file_vars)
830 char var[100], val[100];
831 unsigned i;
833 ch = READCHAR;
835 /* Read a variable name. */
836 while (ch == ' ' || ch == '\t')
837 ch = READCHAR;
839 i = 0;
840 while (ch != ':' && ch != '\n' && ch != EOF && in_file_vars)
842 if (i < sizeof var - 1)
843 var[i++] = ch;
844 UPDATE_BEG_END_STATE (ch);
845 ch = READCHAR;
848 /* Stop scanning if no colon was found before end marker. */
849 if (!in_file_vars || ch == '\n' || ch == EOF)
850 break;
852 while (i > 0 && (var[i - 1] == ' ' || var[i - 1] == '\t'))
853 i--;
854 var[i] = '\0';
856 if (ch == ':')
858 /* Read a variable value. */
859 ch = READCHAR;
861 while (ch == ' ' || ch == '\t')
862 ch = READCHAR;
864 i = 0;
865 while (ch != ';' && ch != '\n' && ch != EOF && in_file_vars)
867 if (i < sizeof val - 1)
868 val[i++] = ch;
869 UPDATE_BEG_END_STATE (ch);
870 ch = READCHAR;
872 if (! in_file_vars)
873 /* The value was terminated by an end-marker, which remove. */
874 i -= 3;
875 while (i > 0 && (val[i - 1] == ' ' || val[i - 1] == '\t'))
876 i--;
877 val[i] = '\0';
879 if (strcmp (var, "lexical-binding") == 0)
880 /* This is it... */
882 rv = (strcmp (val, "nil") != 0);
883 break;
888 while (ch != '\n' && ch != EOF)
889 ch = READCHAR;
891 return rv;
895 /* Value is a version number of byte compiled code if the file
896 associated with file descriptor FD is a compiled Lisp file that's
897 safe to load. Only files compiled with Emacs are safe to load.
898 Files compiled with XEmacs can lead to a crash in Fbyte_code
899 because of an incompatible change in the byte compiler. */
901 static int
902 safe_to_load_version (int fd)
904 char buf[512];
905 int nbytes, i;
906 int version = 1;
908 /* Read the first few bytes from the file, and look for a line
909 specifying the byte compiler version used. */
910 nbytes = emacs_read (fd, buf, sizeof buf);
911 if (nbytes > 0)
913 /* Skip to the next newline, skipping over the initial `ELC'
914 with NUL bytes following it, but note the version. */
915 for (i = 0; i < nbytes && buf[i] != '\n'; ++i)
916 if (i == 4)
917 version = buf[i];
919 if (i >= nbytes
920 || fast_c_string_match_ignore_case (Vbytecomp_version_regexp,
921 buf + i, nbytes - i) < 0)
922 version = 0;
925 lseek (fd, 0, SEEK_SET);
926 return version;
930 /* Callback for record_unwind_protect. Restore the old load list OLD,
931 after loading a file successfully. */
933 static void
934 record_load_unwind (Lisp_Object old)
936 Vloads_in_progress = old;
939 /* This handler function is used via internal_condition_case_1. */
941 static Lisp_Object
942 load_error_handler (Lisp_Object data)
944 return Qnil;
947 static void
948 load_warn_old_style_backquotes (Lisp_Object file)
950 if (!NILP (Vold_style_backquotes))
952 AUTO_STRING (format, "Loading `%s': old-style backquotes detected!");
953 CALLN (Fmessage, format, file);
957 DEFUN ("get-load-suffixes", Fget_load_suffixes, Sget_load_suffixes, 0, 0, 0,
958 doc: /* Return the suffixes that `load' should try if a suffix is \
959 required.
960 This uses the variables `load-suffixes' and `load-file-rep-suffixes'. */)
961 (void)
963 Lisp_Object lst = Qnil, suffixes = Vload_suffixes, suffix, ext;
964 while (CONSP (suffixes))
966 Lisp_Object exts = Vload_file_rep_suffixes;
967 suffix = XCAR (suffixes);
968 suffixes = XCDR (suffixes);
969 while (CONSP (exts))
971 ext = XCAR (exts);
972 exts = XCDR (exts);
973 lst = Fcons (concat2 (suffix, ext), lst);
976 return Fnreverse (lst);
979 /* Returns true if STRING ends with SUFFIX */
980 static bool
981 suffix_p (Lisp_Object string, const char *suffix)
983 ptrdiff_t suffix_len = strlen (suffix);
984 ptrdiff_t string_len = SBYTES (string);
986 return string_len >= suffix_len && !strcmp (SSDATA (string) + string_len - suffix_len, suffix);
989 DEFUN ("load", Fload, Sload, 1, 5, 0,
990 doc: /* Execute a file of Lisp code named FILE.
991 First try FILE with `.elc' appended, then try with `.el', then try
992 with a system-dependent suffix of dynamic modules (see `load-suffixes'),
993 then try FILE unmodified (the exact suffixes in the exact order are
994 determined by `load-suffixes'). Environment variable references in
995 FILE are replaced with their values by calling `substitute-in-file-name'.
996 This function searches the directories in `load-path'.
998 If optional second arg NOERROR is non-nil,
999 report no error if FILE doesn't exist.
1000 Print messages at start and end of loading unless
1001 optional third arg NOMESSAGE is non-nil (but `force-load-messages'
1002 overrides that).
1003 If optional fourth arg NOSUFFIX is non-nil, don't try adding
1004 suffixes to the specified name FILE.
1005 If optional fifth arg MUST-SUFFIX is non-nil, insist on
1006 the suffix `.elc' or `.el' or the module suffix; don't accept just
1007 FILE unless it ends in one of those suffixes or includes a directory name.
1009 If NOSUFFIX is nil, then if a file could not be found, try looking for
1010 a different representation of the file by adding non-empty suffixes to
1011 its name, before trying another file. Emacs uses this feature to find
1012 compressed versions of files when Auto Compression mode is enabled.
1013 If NOSUFFIX is non-nil, disable this feature.
1015 The suffixes that this function tries out, when NOSUFFIX is nil, are
1016 given by the return value of `get-load-suffixes' and the values listed
1017 in `load-file-rep-suffixes'. If MUST-SUFFIX is non-nil, only the
1018 return value of `get-load-suffixes' is used, i.e. the file name is
1019 required to have a non-empty suffix.
1021 When searching suffixes, this function normally stops at the first
1022 one that exists. If the option `load-prefer-newer' is non-nil,
1023 however, it tries all suffixes, and uses whichever file is the newest.
1025 Loading a file records its definitions, and its `provide' and
1026 `require' calls, in an element of `load-history' whose
1027 car is the file name loaded. See `load-history'.
1029 While the file is in the process of being loaded, the variable
1030 `load-in-progress' is non-nil and the variable `load-file-name'
1031 is bound to the file's name.
1033 Return t if the file exists and loads successfully. */)
1034 (Lisp_Object file, Lisp_Object noerror, Lisp_Object nomessage,
1035 Lisp_Object nosuffix, Lisp_Object must_suffix)
1037 FILE *stream;
1038 int fd;
1039 int fd_index;
1040 ptrdiff_t count = SPECPDL_INDEX ();
1041 Lisp_Object found, efound, hist_file_name;
1042 /* True means we printed the ".el is newer" message. */
1043 bool newer = 0;
1044 /* True means we are loading a compiled file. */
1045 bool compiled = 0;
1046 Lisp_Object handler;
1047 bool safe_p = 1;
1048 const char *fmode = "r" FOPEN_TEXT;
1049 int version;
1051 CHECK_STRING (file);
1053 /* If file name is magic, call the handler. */
1054 /* This shouldn't be necessary any more now that `openp' handles it right.
1055 handler = Ffind_file_name_handler (file, Qload);
1056 if (!NILP (handler))
1057 return call5 (handler, Qload, file, noerror, nomessage, nosuffix); */
1059 /* The presence of this call is the result of a historical accident:
1060 it used to be in every file-operation and when it got removed
1061 everywhere, it accidentally stayed here. Since then, enough people
1062 supposedly have things like (load "$PROJECT/foo.el") in their .emacs
1063 that it seemed risky to remove. */
1064 if (! NILP (noerror))
1066 file = internal_condition_case_1 (Fsubstitute_in_file_name, file,
1067 Qt, load_error_handler);
1068 if (NILP (file))
1069 return Qnil;
1071 else
1072 file = Fsubstitute_in_file_name (file);
1074 /* Avoid weird lossage with null string as arg,
1075 since it would try to load a directory as a Lisp file. */
1076 if (SCHARS (file) == 0)
1078 fd = -1;
1079 errno = ENOENT;
1081 else
1083 Lisp_Object suffixes;
1084 found = Qnil;
1086 if (! NILP (must_suffix))
1088 /* Don't insist on adding a suffix if FILE already ends with one. */
1089 if (suffix_p (file, ".el")
1090 || suffix_p (file, ".elc")
1091 #ifdef HAVE_MODULES
1092 || suffix_p (file, MODULES_SUFFIX)
1093 #endif
1095 must_suffix = Qnil;
1096 /* Don't insist on adding a suffix
1097 if the argument includes a directory name. */
1098 else if (! NILP (Ffile_name_directory (file)))
1099 must_suffix = Qnil;
1102 if (!NILP (nosuffix))
1103 suffixes = Qnil;
1104 else
1106 suffixes = Fget_load_suffixes ();
1107 if (NILP (must_suffix))
1108 suffixes = CALLN (Fappend, suffixes, Vload_file_rep_suffixes);
1111 fd = openp (Vload_path, file, suffixes, &found, Qnil, load_prefer_newer);
1114 if (fd == -1)
1116 if (NILP (noerror))
1117 report_file_error ("Cannot open load file", file);
1118 return Qnil;
1121 /* Tell startup.el whether or not we found the user's init file. */
1122 if (EQ (Qt, Vuser_init_file))
1123 Vuser_init_file = found;
1125 /* If FD is -2, that means openp found a magic file. */
1126 if (fd == -2)
1128 if (NILP (Fequal (found, file)))
1129 /* If FOUND is a different file name from FILE,
1130 find its handler even if we have already inhibited
1131 the `load' operation on FILE. */
1132 handler = Ffind_file_name_handler (found, Qt);
1133 else
1134 handler = Ffind_file_name_handler (found, Qload);
1135 if (! NILP (handler))
1136 return call5 (handler, Qload, found, noerror, nomessage, Qt);
1137 #ifdef DOS_NT
1138 /* Tramp has to deal with semi-broken packages that prepend
1139 drive letters to remote files. For that reason, Tramp
1140 catches file operations that test for file existence, which
1141 makes openp think X:/foo.elc files are remote. However,
1142 Tramp does not catch `load' operations for such files, so we
1143 end up with a nil as the `load' handler above. If we would
1144 continue with fd = -2, we will behave wrongly, and in
1145 particular try reading a .elc file in the "rt" mode instead
1146 of "rb". See bug #9311 for the results. To work around
1147 this, we try to open the file locally, and go with that if it
1148 succeeds. */
1149 fd = emacs_open (SSDATA (ENCODE_FILE (found)), O_RDONLY, 0);
1150 if (fd == -1)
1151 fd = -2;
1152 #endif
1155 if (fd < 0)
1157 /* Pacify older GCC with --enable-gcc-warnings. */
1158 IF_LINT (fd_index = 0);
1160 else
1162 fd_index = SPECPDL_INDEX ();
1163 record_unwind_protect_int (close_file_unwind, fd);
1166 #ifdef HAVE_MODULES
1167 if (suffix_p (found, MODULES_SUFFIX))
1168 return unbind_to (count, Fmodule_load (found));
1169 #endif
1171 /* Check if we're stuck in a recursive load cycle.
1173 2000-09-21: It's not possible to just check for the file loaded
1174 being a member of Vloads_in_progress. This fails because of the
1175 way the byte compiler currently works; `provide's are not
1176 evaluated, see font-lock.el/jit-lock.el as an example. This
1177 leads to a certain amount of ``normal'' recursion.
1179 Also, just loading a file recursively is not always an error in
1180 the general case; the second load may do something different. */
1182 int load_count = 0;
1183 Lisp_Object tem;
1184 for (tem = Vloads_in_progress; CONSP (tem); tem = XCDR (tem))
1185 if (!NILP (Fequal (found, XCAR (tem))) && (++load_count > 3))
1186 signal_error ("Recursive load", Fcons (found, Vloads_in_progress));
1187 record_unwind_protect (record_load_unwind, Vloads_in_progress);
1188 Vloads_in_progress = Fcons (found, Vloads_in_progress);
1191 /* All loads are by default dynamic, unless the file itself specifies
1192 otherwise using a file-variable in the first line. This is bound here
1193 so that it takes effect whether or not we use
1194 Vload_source_file_function. */
1195 specbind (Qlexical_binding, Qnil);
1197 /* Get the name for load-history. */
1198 hist_file_name = (! NILP (Vpurify_flag)
1199 ? concat2 (Ffile_name_directory (file),
1200 Ffile_name_nondirectory (found))
1201 : found) ;
1203 version = -1;
1205 /* Check for the presence of old-style quotes and warn about them. */
1206 specbind (Qold_style_backquotes, Qnil);
1207 record_unwind_protect (load_warn_old_style_backquotes, file);
1209 if (suffix_p (found, ".elc") || (fd >= 0 && (version = safe_to_load_version (fd)) > 0))
1210 /* Load .elc files directly, but not when they are
1211 remote and have no handler! */
1213 if (fd != -2)
1215 struct stat s1, s2;
1216 int result;
1218 if (version < 0
1219 && ! (version = safe_to_load_version (fd)))
1221 safe_p = 0;
1222 if (!load_dangerous_libraries)
1223 error ("File `%s' was not compiled in Emacs", SDATA (found));
1224 else if (!NILP (nomessage) && !force_load_messages)
1225 message_with_string ("File `%s' not compiled in Emacs", found, 1);
1228 compiled = 1;
1230 efound = ENCODE_FILE (found);
1231 fmode = "r" FOPEN_BINARY;
1233 /* openp already checked for newness, no point doing it again.
1234 FIXME would be nice to get a message when openp
1235 ignores suffix order due to load_prefer_newer. */
1236 if (!load_prefer_newer)
1238 result = stat (SSDATA (efound), &s1);
1239 if (result == 0)
1241 SSET (efound, SBYTES (efound) - 1, 0);
1242 result = stat (SSDATA (efound), &s2);
1243 SSET (efound, SBYTES (efound) - 1, 'c');
1246 if (result == 0
1247 && timespec_cmp (get_stat_mtime (&s1), get_stat_mtime (&s2)) < 0)
1249 /* Make the progress messages mention that source is newer. */
1250 newer = 1;
1252 /* If we won't print another message, mention this anyway. */
1253 if (!NILP (nomessage) && !force_load_messages)
1255 Lisp_Object msg_file;
1256 msg_file = Fsubstring (found, make_number (0), make_number (-1));
1257 message_with_string ("Source file `%s' newer than byte-compiled file",
1258 msg_file, 1);
1261 } /* !load_prefer_newer */
1264 else
1266 /* We are loading a source file (*.el). */
1267 if (!NILP (Vload_source_file_function))
1269 Lisp_Object val;
1271 if (fd >= 0)
1273 emacs_close (fd);
1274 clear_unwind_protect (fd_index);
1276 val = call4 (Vload_source_file_function, found, hist_file_name,
1277 NILP (noerror) ? Qnil : Qt,
1278 (NILP (nomessage) || force_load_messages) ? Qnil : Qt);
1279 return unbind_to (count, val);
1283 if (fd < 0)
1285 /* We somehow got here with fd == -2, meaning the file is deemed
1286 to be remote. Don't even try to reopen the file locally;
1287 just force a failure. */
1288 stream = NULL;
1289 errno = EINVAL;
1291 else
1293 #ifdef WINDOWSNT
1294 emacs_close (fd);
1295 clear_unwind_protect (fd_index);
1296 efound = ENCODE_FILE (found);
1297 stream = emacs_fopen (SSDATA (efound), fmode);
1298 #else
1299 stream = fdopen (fd, fmode);
1300 #endif
1302 if (! stream)
1303 report_file_error ("Opening stdio stream", file);
1304 set_unwind_protect_ptr (fd_index, fclose_unwind, stream);
1306 if (! NILP (Vpurify_flag))
1307 Vpreloaded_file_list = Fcons (Fpurecopy (file), Vpreloaded_file_list);
1309 if (NILP (nomessage) || force_load_messages)
1311 if (!safe_p)
1312 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...",
1313 file, 1);
1314 else if (!compiled)
1315 message_with_string ("Loading %s (source)...", file, 1);
1316 else if (newer)
1317 message_with_string ("Loading %s (compiled; note, source file is newer)...",
1318 file, 1);
1319 else /* The typical case; compiled file newer than source file. */
1320 message_with_string ("Loading %s...", file, 1);
1323 specbind (Qload_file_name, found);
1324 specbind (Qinhibit_file_name_operation, Qnil);
1325 specbind (Qload_in_progress, Qt);
1327 instream = stream;
1328 if (lisp_file_lexically_bound_p (Qget_file_char))
1329 Fset (Qlexical_binding, Qt);
1331 if (! version || version >= 22)
1332 readevalloop (Qget_file_char, stream, hist_file_name,
1333 0, Qnil, Qnil, Qnil, Qnil);
1334 else
1336 /* We can't handle a file which was compiled with
1337 byte-compile-dynamic by older version of Emacs. */
1338 specbind (Qload_force_doc_strings, Qt);
1339 readevalloop (Qget_emacs_mule_file_char, stream, hist_file_name,
1340 0, Qnil, Qnil, Qnil, Qnil);
1342 unbind_to (count, Qnil);
1344 /* Run any eval-after-load forms for this file. */
1345 if (!NILP (Ffboundp (Qdo_after_load_evaluation)))
1346 call1 (Qdo_after_load_evaluation, hist_file_name) ;
1348 xfree (saved_doc_string);
1349 saved_doc_string = 0;
1350 saved_doc_string_size = 0;
1352 xfree (prev_saved_doc_string);
1353 prev_saved_doc_string = 0;
1354 prev_saved_doc_string_size = 0;
1356 if (!noninteractive && (NILP (nomessage) || force_load_messages))
1358 if (!safe_p)
1359 message_with_string ("Loading %s (compiled; note unsafe, not compiled in Emacs)...done",
1360 file, 1);
1361 else if (!compiled)
1362 message_with_string ("Loading %s (source)...done", file, 1);
1363 else if (newer)
1364 message_with_string ("Loading %s (compiled; note, source file is newer)...done",
1365 file, 1);
1366 else /* The typical case; compiled file newer than source file. */
1367 message_with_string ("Loading %s...done", file, 1);
1370 return Qt;
1373 static bool
1374 complete_filename_p (Lisp_Object pathname)
1376 const unsigned char *s = SDATA (pathname);
1377 return (IS_DIRECTORY_SEP (s[0])
1378 || (SCHARS (pathname) > 2
1379 && IS_DEVICE_SEP (s[1]) && IS_DIRECTORY_SEP (s[2])));
1382 DEFUN ("locate-file-internal", Flocate_file_internal, Slocate_file_internal, 2, 4, 0,
1383 doc: /* Search for FILENAME through PATH.
1384 Returns the file's name in absolute form, or nil if not found.
1385 If SUFFIXES is non-nil, it should be a list of suffixes to append to
1386 file name when searching.
1387 If non-nil, PREDICATE is used instead of `file-readable-p'.
1388 PREDICATE can also be an integer to pass to the faccessat(2) function,
1389 in which case file-name-handlers are ignored.
1390 This function will normally skip directories, so if you want it to find
1391 directories, make sure the PREDICATE function returns `dir-ok' for them. */)
1392 (Lisp_Object filename, Lisp_Object path, Lisp_Object suffixes, Lisp_Object predicate)
1394 Lisp_Object file;
1395 int fd = openp (path, filename, suffixes, &file, predicate, false);
1396 if (NILP (predicate) && fd >= 0)
1397 emacs_close (fd);
1398 return file;
1401 /* Search for a file whose name is STR, looking in directories
1402 in the Lisp list PATH, and trying suffixes from SUFFIX.
1403 On success, return a file descriptor (or 1 or -2 as described below).
1404 On failure, return -1 and set errno.
1406 SUFFIXES is a list of strings containing possible suffixes.
1407 The empty suffix is automatically added if the list is empty.
1409 PREDICATE t means the files are binary.
1410 PREDICATE non-nil and non-t means don't open the files,
1411 just look for one that satisfies the predicate. In this case,
1412 return 1 on success. The predicate can be a lisp function or
1413 an integer to pass to `access' (in which case file-name-handlers
1414 are ignored).
1416 If STOREPTR is nonzero, it points to a slot where the name of
1417 the file actually found should be stored as a Lisp string.
1418 nil is stored there on failure.
1420 If the file we find is remote, return -2
1421 but store the found remote file name in *STOREPTR.
1423 If NEWER is true, try all SUFFIXes and return the result for the
1424 newest file that exists. Does not apply to remote files,
1425 or if a non-nil and non-t PREDICATE is specified. */
1428 openp (Lisp_Object path, Lisp_Object str, Lisp_Object suffixes,
1429 Lisp_Object *storeptr, Lisp_Object predicate, bool newer)
1431 ptrdiff_t fn_size = 100;
1432 char buf[100];
1433 char *fn = buf;
1434 bool absolute;
1435 ptrdiff_t want_length;
1436 Lisp_Object filename;
1437 Lisp_Object string, tail, encoded_fn, save_string;
1438 ptrdiff_t max_suffix_len = 0;
1439 int last_errno = ENOENT;
1440 int save_fd = -1;
1441 USE_SAFE_ALLOCA;
1443 /* The last-modified time of the newest matching file found.
1444 Initialize it to something less than all valid timestamps. */
1445 struct timespec save_mtime = make_timespec (TYPE_MINIMUM (time_t), -1);
1447 CHECK_STRING (str);
1449 for (tail = suffixes; CONSP (tail); tail = XCDR (tail))
1451 CHECK_STRING_CAR (tail);
1452 max_suffix_len = max (max_suffix_len,
1453 SBYTES (XCAR (tail)));
1456 string = filename = encoded_fn = save_string = Qnil;
1458 if (storeptr)
1459 *storeptr = Qnil;
1461 absolute = complete_filename_p (str);
1463 for (; CONSP (path); path = XCDR (path))
1465 filename = Fexpand_file_name (str, XCAR (path));
1466 if (!complete_filename_p (filename))
1467 /* If there are non-absolute elts in PATH (eg "."). */
1468 /* Of course, this could conceivably lose if luser sets
1469 default-directory to be something non-absolute... */
1471 filename = Fexpand_file_name (filename, BVAR (current_buffer, directory));
1472 if (!complete_filename_p (filename))
1473 /* Give up on this path element! */
1474 continue;
1477 /* Calculate maximum length of any filename made from
1478 this path element/specified file name and any possible suffix. */
1479 want_length = max_suffix_len + SBYTES (filename);
1480 if (fn_size <= want_length)
1482 fn_size = 100 + want_length;
1483 fn = SAFE_ALLOCA (fn_size);
1486 /* Loop over suffixes. */
1487 for (tail = NILP (suffixes) ? list1 (empty_unibyte_string) : suffixes;
1488 CONSP (tail); tail = XCDR (tail))
1490 Lisp_Object suffix = XCAR (tail);
1491 ptrdiff_t fnlen, lsuffix = SBYTES (suffix);
1492 Lisp_Object handler;
1494 /* Concatenate path element/specified name with the suffix.
1495 If the directory starts with /:, remove that. */
1496 int prefixlen = ((SCHARS (filename) > 2
1497 && SREF (filename, 0) == '/'
1498 && SREF (filename, 1) == ':')
1499 ? 2 : 0);
1500 fnlen = SBYTES (filename) - prefixlen;
1501 memcpy (fn, SDATA (filename) + prefixlen, fnlen);
1502 memcpy (fn + fnlen, SDATA (suffix), lsuffix + 1);
1503 fnlen += lsuffix;
1504 /* Check that the file exists and is not a directory. */
1505 /* We used to only check for handlers on non-absolute file names:
1506 if (absolute)
1507 handler = Qnil;
1508 else
1509 handler = Ffind_file_name_handler (filename, Qfile_exists_p);
1510 It's not clear why that was the case and it breaks things like
1511 (load "/bar.el") where the file is actually "/bar.el.gz". */
1512 /* make_string has its own ideas on when to return a unibyte
1513 string and when a multibyte string, but we know better.
1514 We must have a unibyte string when dumping, since
1515 file-name encoding is shaky at best at that time, and in
1516 particular default-file-name-coding-system is reset
1517 several times during loadup. We therefore don't want to
1518 encode the file before passing it to file I/O library
1519 functions. */
1520 if (!STRING_MULTIBYTE (filename) && !STRING_MULTIBYTE (suffix))
1521 string = make_unibyte_string (fn, fnlen);
1522 else
1523 string = make_string (fn, fnlen);
1524 handler = Ffind_file_name_handler (string, Qfile_exists_p);
1525 if ((!NILP (handler) || (!NILP (predicate) && !EQ (predicate, Qt)))
1526 && !NATNUMP (predicate))
1528 bool exists;
1529 if (NILP (predicate) || EQ (predicate, Qt))
1530 exists = !NILP (Ffile_readable_p (string));
1531 else
1533 Lisp_Object tmp = call1 (predicate, string);
1534 if (NILP (tmp))
1535 exists = false;
1536 else if (EQ (tmp, Qdir_ok)
1537 || NILP (Ffile_directory_p (string)))
1538 exists = true;
1539 else
1541 exists = false;
1542 last_errno = EISDIR;
1546 if (exists)
1548 /* We succeeded; return this descriptor and filename. */
1549 if (storeptr)
1550 *storeptr = string;
1551 SAFE_FREE ();
1552 return -2;
1555 else
1557 int fd;
1558 const char *pfn;
1559 struct stat st;
1561 encoded_fn = ENCODE_FILE (string);
1562 pfn = SSDATA (encoded_fn);
1564 /* Check that we can access or open it. */
1565 if (NATNUMP (predicate))
1567 fd = -1;
1568 if (INT_MAX < XFASTINT (predicate))
1569 last_errno = EINVAL;
1570 else if (faccessat (AT_FDCWD, pfn, XFASTINT (predicate),
1571 AT_EACCESS)
1572 == 0)
1574 if (file_directory_p (pfn))
1575 last_errno = EISDIR;
1576 else
1577 fd = 1;
1580 else
1582 fd = emacs_open (pfn, O_RDONLY, 0);
1583 if (fd < 0)
1585 if (errno != ENOENT)
1586 last_errno = errno;
1588 else
1590 int err = (fstat (fd, &st) != 0 ? errno
1591 : S_ISDIR (st.st_mode) ? EISDIR : 0);
1592 if (err)
1594 last_errno = err;
1595 emacs_close (fd);
1596 fd = -1;
1601 if (fd >= 0)
1603 if (newer && !NATNUMP (predicate))
1605 struct timespec mtime = get_stat_mtime (&st);
1607 if (timespec_cmp (mtime, save_mtime) <= 0)
1608 emacs_close (fd);
1609 else
1611 if (0 <= save_fd)
1612 emacs_close (save_fd);
1613 save_fd = fd;
1614 save_mtime = mtime;
1615 save_string = string;
1618 else
1620 /* We succeeded; return this descriptor and filename. */
1621 if (storeptr)
1622 *storeptr = string;
1623 SAFE_FREE ();
1624 return fd;
1628 /* No more suffixes. Return the newest. */
1629 if (0 <= save_fd && ! CONSP (XCDR (tail)))
1631 if (storeptr)
1632 *storeptr = save_string;
1633 SAFE_FREE ();
1634 return save_fd;
1638 if (absolute)
1639 break;
1642 SAFE_FREE ();
1643 errno = last_errno;
1644 return -1;
1648 /* Merge the list we've accumulated of globals from the current input source
1649 into the load_history variable. The details depend on whether
1650 the source has an associated file name or not.
1652 FILENAME is the file name that we are loading from.
1654 ENTIRE is true if loading that entire file, false if evaluating
1655 part of it. */
1657 static void
1658 build_load_history (Lisp_Object filename, bool entire)
1660 Lisp_Object tail, prev, newelt;
1661 Lisp_Object tem, tem2;
1662 bool foundit = 0;
1664 tail = Vload_history;
1665 prev = Qnil;
1667 while (CONSP (tail))
1669 tem = XCAR (tail);
1671 /* Find the feature's previous assoc list... */
1672 if (!NILP (Fequal (filename, Fcar (tem))))
1674 foundit = 1;
1676 /* If we're loading the entire file, remove old data. */
1677 if (entire)
1679 if (NILP (prev))
1680 Vload_history = XCDR (tail);
1681 else
1682 Fsetcdr (prev, XCDR (tail));
1685 /* Otherwise, cons on new symbols that are not already members. */
1686 else
1688 tem2 = Vcurrent_load_list;
1690 while (CONSP (tem2))
1692 newelt = XCAR (tem2);
1694 if (NILP (Fmember (newelt, tem)))
1695 Fsetcar (tail, Fcons (XCAR (tem),
1696 Fcons (newelt, XCDR (tem))));
1698 tem2 = XCDR (tem2);
1699 QUIT;
1703 else
1704 prev = tail;
1705 tail = XCDR (tail);
1706 QUIT;
1709 /* If we're loading an entire file, cons the new assoc onto the
1710 front of load-history, the most-recently-loaded position. Also
1711 do this if we didn't find an existing member for the file. */
1712 if (entire || !foundit)
1713 Vload_history = Fcons (Fnreverse (Vcurrent_load_list),
1714 Vload_history);
1717 static void
1718 readevalloop_1 (int old)
1720 load_convert_to_unibyte = old;
1723 /* Signal an `end-of-file' error, if possible with file name
1724 information. */
1726 static _Noreturn void
1727 end_of_file_error (void)
1729 if (STRINGP (Vload_file_name))
1730 xsignal1 (Qend_of_file, Vload_file_name);
1732 xsignal0 (Qend_of_file);
1735 static Lisp_Object
1736 readevalloop_eager_expand_eval (Lisp_Object val, Lisp_Object macroexpand)
1738 /* If we macroexpand the toplevel form non-recursively and it ends
1739 up being a `progn' (or if it was a progn to start), treat each
1740 form in the progn as a top-level form. This way, if one form in
1741 the progn defines a macro, that macro is in effect when we expand
1742 the remaining forms. See similar code in bytecomp.el. */
1743 val = call2 (macroexpand, val, Qnil);
1744 if (EQ (CAR_SAFE (val), Qprogn))
1746 Lisp_Object subforms = XCDR (val);
1748 for (val = Qnil; CONSP (subforms); subforms = XCDR (subforms))
1749 val = readevalloop_eager_expand_eval (XCAR (subforms),
1750 macroexpand);
1752 else
1753 val = eval_sub (call2 (macroexpand, val, Qt));
1754 return val;
1757 /* UNIBYTE specifies how to set load_convert_to_unibyte
1758 for this invocation.
1759 READFUN, if non-nil, is used instead of `read'.
1761 START, END specify region to read in current buffer (from eval-region).
1762 If the input is not from a buffer, they must be nil. */
1764 static void
1765 readevalloop (Lisp_Object readcharfun,
1766 FILE *stream,
1767 Lisp_Object sourcename,
1768 bool printflag,
1769 Lisp_Object unibyte, Lisp_Object readfun,
1770 Lisp_Object start, Lisp_Object end)
1772 int c;
1773 Lisp_Object val;
1774 ptrdiff_t count = SPECPDL_INDEX ();
1775 struct buffer *b = 0;
1776 bool continue_reading_p;
1777 Lisp_Object lex_bound;
1778 /* True if reading an entire buffer. */
1779 bool whole_buffer = 0;
1780 /* True on the first time around. */
1781 bool first_sexp = 1;
1782 Lisp_Object macroexpand = intern ("internal-macroexpand-for-load");
1784 if (NILP (Ffboundp (macroexpand))
1785 /* Don't macroexpand in .elc files, since it should have been done
1786 already. We actually don't know whether we're in a .elc file or not,
1787 so we use circumstantial evidence: .el files normally go through
1788 Vload_source_file_function -> load-with-code-conversion
1789 -> eval-buffer. */
1790 || EQ (readcharfun, Qget_file_char)
1791 || EQ (readcharfun, Qget_emacs_mule_file_char))
1792 macroexpand = Qnil;
1794 if (MARKERP (readcharfun))
1796 if (NILP (start))
1797 start = readcharfun;
1800 if (BUFFERP (readcharfun))
1801 b = XBUFFER (readcharfun);
1802 else if (MARKERP (readcharfun))
1803 b = XMARKER (readcharfun)->buffer;
1805 /* We assume START is nil when input is not from a buffer. */
1806 if (! NILP (start) && !b)
1807 emacs_abort ();
1809 specbind (Qstandard_input, readcharfun);
1810 specbind (Qcurrent_load_list, Qnil);
1811 record_unwind_protect_int (readevalloop_1, load_convert_to_unibyte);
1812 load_convert_to_unibyte = !NILP (unibyte);
1814 /* If lexical binding is active (either because it was specified in
1815 the file's header, or via a buffer-local variable), create an empty
1816 lexical environment, otherwise, turn off lexical binding. */
1817 lex_bound = find_symbol_value (Qlexical_binding);
1818 specbind (Qinternal_interpreter_environment,
1819 (NILP (lex_bound) || EQ (lex_bound, Qunbound)
1820 ? Qnil : list1 (Qt)));
1822 /* Try to ensure sourcename is a truename, except whilst preloading. */
1823 if (NILP (Vpurify_flag)
1824 && !NILP (sourcename) && !NILP (Ffile_name_absolute_p (sourcename))
1825 && !NILP (Ffboundp (Qfile_truename)))
1826 sourcename = call1 (Qfile_truename, sourcename) ;
1828 LOADHIST_ATTACH (sourcename);
1830 continue_reading_p = 1;
1831 while (continue_reading_p)
1833 ptrdiff_t count1 = SPECPDL_INDEX ();
1835 if (b != 0 && !BUFFER_LIVE_P (b))
1836 error ("Reading from killed buffer");
1838 if (!NILP (start))
1840 /* Switch to the buffer we are reading from. */
1841 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1842 set_buffer_internal (b);
1844 /* Save point in it. */
1845 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1846 /* Save ZV in it. */
1847 record_unwind_protect (save_restriction_restore, save_restriction_save ());
1848 /* Those get unbound after we read one expression. */
1850 /* Set point and ZV around stuff to be read. */
1851 Fgoto_char (start);
1852 if (!NILP (end))
1853 Fnarrow_to_region (make_number (BEGV), end);
1855 /* Just for cleanliness, convert END to a marker
1856 if it is an integer. */
1857 if (INTEGERP (end))
1858 end = Fpoint_max_marker ();
1861 /* On the first cycle, we can easily test here
1862 whether we are reading the whole buffer. */
1863 if (b && first_sexp)
1864 whole_buffer = (PT == BEG && ZV == Z);
1866 instream = stream;
1867 read_next:
1868 c = READCHAR;
1869 if (c == ';')
1871 while ((c = READCHAR) != '\n' && c != -1);
1872 goto read_next;
1874 if (c < 0)
1876 unbind_to (count1, Qnil);
1877 break;
1880 /* Ignore whitespace here, so we can detect eof. */
1881 if (c == ' ' || c == '\t' || c == '\n' || c == '\f' || c == '\r'
1882 || c == NO_BREAK_SPACE)
1883 goto read_next;
1885 if (!NILP (Vpurify_flag) && c == '(')
1887 val = read_list (0, readcharfun);
1889 else
1891 UNREAD (c);
1892 read_objects = Qnil;
1893 if (!NILP (readfun))
1895 val = call1 (readfun, readcharfun);
1897 /* If READCHARFUN has set point to ZV, we should
1898 stop reading, even if the form read sets point
1899 to a different value when evaluated. */
1900 if (BUFFERP (readcharfun))
1902 struct buffer *buf = XBUFFER (readcharfun);
1903 if (BUF_PT (buf) == BUF_ZV (buf))
1904 continue_reading_p = 0;
1907 else if (! NILP (Vload_read_function))
1908 val = call1 (Vload_read_function, readcharfun);
1909 else
1910 val = read_internal_start (readcharfun, Qnil, Qnil);
1913 if (!NILP (start) && continue_reading_p)
1914 start = Fpoint_marker ();
1916 /* Restore saved point and BEGV. */
1917 unbind_to (count1, Qnil);
1919 /* Now eval what we just read. */
1920 if (!NILP (macroexpand))
1921 val = readevalloop_eager_expand_eval (val, macroexpand);
1922 else
1923 val = eval_sub (val);
1925 if (printflag)
1927 Vvalues = Fcons (val, Vvalues);
1928 if (EQ (Vstandard_output, Qt))
1929 Fprin1 (val, Qnil);
1930 else
1931 Fprint (val, Qnil);
1934 first_sexp = 0;
1937 build_load_history (sourcename,
1938 stream || whole_buffer);
1940 unbind_to (count, Qnil);
1943 DEFUN ("eval-buffer", Feval_buffer, Seval_buffer, 0, 5, "",
1944 doc: /* Execute the accessible portion of current buffer as Lisp code.
1945 You can use \\[narrow-to-region] to limit the part of buffer to be evaluated.
1946 When called from a Lisp program (i.e., not interactively), this
1947 function accepts up to five optional arguments:
1948 BUFFER is the buffer to evaluate (nil means use current buffer),
1949 or a name of a buffer (a string).
1950 PRINTFLAG controls printing of output by any output functions in the
1951 evaluated code, such as `print', `princ', and `prin1':
1952 a value of nil means discard it; anything else is the stream to print to.
1953 See Info node `(elisp)Output Streams' for details on streams.
1954 FILENAME specifies the file name to use for `load-history'.
1955 UNIBYTE, if non-nil, specifies `load-convert-to-unibyte' for this
1956 invocation.
1957 DO-ALLOW-PRINT, if non-nil, specifies that output functions in the
1958 evaluated code should work normally even if PRINTFLAG is nil, in
1959 which case the output is displayed in the echo area.
1961 This function preserves the position of point. */)
1962 (Lisp_Object buffer, Lisp_Object printflag, Lisp_Object filename, Lisp_Object unibyte, Lisp_Object do_allow_print)
1964 ptrdiff_t count = SPECPDL_INDEX ();
1965 Lisp_Object tem, buf;
1967 if (NILP (buffer))
1968 buf = Fcurrent_buffer ();
1969 else
1970 buf = Fget_buffer (buffer);
1971 if (NILP (buf))
1972 error ("No such buffer");
1974 if (NILP (printflag) && NILP (do_allow_print))
1975 tem = Qsymbolp;
1976 else
1977 tem = printflag;
1979 if (NILP (filename))
1980 filename = BVAR (XBUFFER (buf), filename);
1982 specbind (Qeval_buffer_list, Fcons (buf, Veval_buffer_list));
1983 specbind (Qstandard_output, tem);
1984 record_unwind_protect (save_excursion_restore, save_excursion_save ());
1985 BUF_TEMP_SET_PT (XBUFFER (buf), BUF_BEGV (XBUFFER (buf)));
1986 specbind (Qlexical_binding, lisp_file_lexically_bound_p (buf) ? Qt : Qnil);
1987 readevalloop (buf, 0, filename,
1988 !NILP (printflag), unibyte, Qnil, Qnil, Qnil);
1989 unbind_to (count, Qnil);
1991 return Qnil;
1994 DEFUN ("eval-region", Feval_region, Seval_region, 2, 4, "r",
1995 doc: /* Execute the region as Lisp code.
1996 When called from programs, expects two arguments,
1997 giving starting and ending indices in the current buffer
1998 of the text to be executed.
1999 Programs can pass third argument PRINTFLAG which controls output:
2000 a value of nil means discard it; anything else is stream for printing it.
2001 See Info node `(elisp)Output Streams' for details on streams.
2002 Also the fourth argument READ-FUNCTION, if non-nil, is used
2003 instead of `read' to read each expression. It gets one argument
2004 which is the input stream for reading characters.
2006 This function does not move point. */)
2007 (Lisp_Object start, Lisp_Object end, Lisp_Object printflag, Lisp_Object read_function)
2009 /* FIXME: Do the eval-sexp-add-defvars dance! */
2010 ptrdiff_t count = SPECPDL_INDEX ();
2011 Lisp_Object tem, cbuf;
2013 cbuf = Fcurrent_buffer ();
2015 if (NILP (printflag))
2016 tem = Qsymbolp;
2017 else
2018 tem = printflag;
2019 specbind (Qstandard_output, tem);
2020 specbind (Qeval_buffer_list, Fcons (cbuf, Veval_buffer_list));
2022 /* `readevalloop' calls functions which check the type of start and end. */
2023 readevalloop (cbuf, 0, BVAR (XBUFFER (cbuf), filename),
2024 !NILP (printflag), Qnil, read_function,
2025 start, end);
2027 return unbind_to (count, Qnil);
2031 DEFUN ("read", Fread, Sread, 0, 1, 0,
2032 doc: /* Read one Lisp expression as text from STREAM, return as Lisp object.
2033 If STREAM is nil, use the value of `standard-input' (which see).
2034 STREAM or the value of `standard-input' may be:
2035 a buffer (read from point and advance it)
2036 a marker (read from where it points and advance it)
2037 a function (call it with no arguments for each character,
2038 call it with a char as argument to push a char back)
2039 a string (takes text from string, starting at the beginning)
2040 t (read text line using minibuffer and use it, or read from
2041 standard input in batch mode). */)
2042 (Lisp_Object stream)
2044 if (NILP (stream))
2045 stream = Vstandard_input;
2046 if (EQ (stream, Qt))
2047 stream = Qread_char;
2048 if (EQ (stream, Qread_char))
2049 /* FIXME: ?! When is this used !? */
2050 return call1 (intern ("read-minibuffer"),
2051 build_string ("Lisp expression: "));
2053 return read_internal_start (stream, Qnil, Qnil);
2056 DEFUN ("read-from-string", Fread_from_string, Sread_from_string, 1, 3, 0,
2057 doc: /* Read one Lisp expression which is represented as text by STRING.
2058 Returns a cons: (OBJECT-READ . FINAL-STRING-INDEX).
2059 FINAL-STRING-INDEX is an integer giving the position of the next
2060 remaining character in STRING. START and END optionally delimit
2061 a substring of STRING from which to read; they default to 0 and
2062 \(length STRING) respectively. Negative values are counted from
2063 the end of STRING. */)
2064 (Lisp_Object string, Lisp_Object start, Lisp_Object end)
2066 Lisp_Object ret;
2067 CHECK_STRING (string);
2068 /* `read_internal_start' sets `read_from_string_index'. */
2069 ret = read_internal_start (string, start, end);
2070 return Fcons (ret, make_number (read_from_string_index));
2073 /* Function to set up the global context we need in toplevel read
2074 calls. START and END only used when STREAM is a string. */
2075 static Lisp_Object
2076 read_internal_start (Lisp_Object stream, Lisp_Object start, Lisp_Object end)
2078 Lisp_Object retval;
2080 readchar_count = 0;
2081 new_backquote_flag = 0;
2082 read_objects = Qnil;
2083 if (EQ (Vread_with_symbol_positions, Qt)
2084 || EQ (Vread_with_symbol_positions, stream))
2085 Vread_symbol_positions_list = Qnil;
2087 if (STRINGP (stream)
2088 || ((CONSP (stream) && STRINGP (XCAR (stream)))))
2090 ptrdiff_t startval, endval;
2091 Lisp_Object string;
2093 if (STRINGP (stream))
2094 string = stream;
2095 else
2096 string = XCAR (stream);
2098 validate_subarray (string, start, end, SCHARS (string),
2099 &startval, &endval);
2101 read_from_string_index = startval;
2102 read_from_string_index_byte = string_char_to_byte (string, startval);
2103 read_from_string_limit = endval;
2106 retval = read0 (stream);
2107 if (EQ (Vread_with_symbol_positions, Qt)
2108 || EQ (Vread_with_symbol_positions, stream))
2109 Vread_symbol_positions_list = Fnreverse (Vread_symbol_positions_list);
2110 return retval;
2114 /* Signal Qinvalid_read_syntax error.
2115 S is error string of length N (if > 0) */
2117 static _Noreturn void
2118 invalid_syntax (const char *s)
2120 xsignal1 (Qinvalid_read_syntax, build_string (s));
2124 /* Use this for recursive reads, in contexts where internal tokens
2125 are not allowed. */
2127 static Lisp_Object
2128 read0 (Lisp_Object readcharfun)
2130 register Lisp_Object val;
2131 int c;
2133 val = read1 (readcharfun, &c, 0);
2134 if (!c)
2135 return val;
2137 xsignal1 (Qinvalid_read_syntax,
2138 Fmake_string (make_number (1), make_number (c)));
2141 static ptrdiff_t read_buffer_size;
2142 static char *read_buffer;
2144 /* Grow the read buffer by at least MAX_MULTIBYTE_LENGTH bytes. */
2146 static void
2147 grow_read_buffer (void)
2149 read_buffer = xpalloc (read_buffer, &read_buffer_size,
2150 MAX_MULTIBYTE_LENGTH, -1, 1);
2153 /* Return the scalar value that has the Unicode character name NAME.
2154 Raise 'invalid-read-syntax' if there is no such character. */
2155 static int
2156 character_name_to_code (char const *name, ptrdiff_t name_len)
2158 Lisp_Object code;
2160 /* Code point as U+XXXX.... */
2161 if (name[0] == 'U' && name[1] == '+')
2163 /* Pass the leading '+' to string_to_number, so that it
2164 rejects monstrosities such as negative values. */
2165 code = string_to_number (name + 1, 16, false);
2167 else
2169 /* Look up the name in the table returned by 'ucs-names'. */
2170 AUTO_STRING_WITH_LEN (namestr, name, name_len);
2171 Lisp_Object names = call0 (Qucs_names);
2172 code = CDR (Fassoc (namestr, names));
2175 if (! (INTEGERP (code)
2176 && 0 <= XINT (code) && XINT (code) <= MAX_UNICODE_CHAR
2177 && ! char_surrogate_p (XINT (code))))
2179 AUTO_STRING (format, "\\N{%s}");
2180 AUTO_STRING_WITH_LEN (namestr, name, name_len);
2181 xsignal1 (Qinvalid_read_syntax, CALLN (Fformat, format, namestr));
2184 return XINT (code);
2187 /* Bound on the length of a Unicode character name. As of
2188 Unicode 9.0.0 the maximum is 83, so this should be safe. */
2189 enum { UNICODE_CHARACTER_NAME_LENGTH_BOUND = 200 };
2191 /* Read a \-escape sequence, assuming we already read the `\'.
2192 If the escape sequence forces unibyte, return eight-bit char. */
2194 static int
2195 read_escape (Lisp_Object readcharfun, bool stringp)
2197 int c = READCHAR;
2198 /* \u allows up to four hex digits, \U up to eight. Default to the
2199 behavior for \u, and change this value in the case that \U is seen. */
2200 int unicode_hex_count = 4;
2202 switch (c)
2204 case -1:
2205 end_of_file_error ();
2207 case 'a':
2208 return '\007';
2209 case 'b':
2210 return '\b';
2211 case 'd':
2212 return 0177;
2213 case 'e':
2214 return 033;
2215 case 'f':
2216 return '\f';
2217 case 'n':
2218 return '\n';
2219 case 'r':
2220 return '\r';
2221 case 't':
2222 return '\t';
2223 case 'v':
2224 return '\v';
2225 case '\n':
2226 return -1;
2227 case ' ':
2228 if (stringp)
2229 return -1;
2230 return ' ';
2232 case 'M':
2233 c = READCHAR;
2234 if (c != '-')
2235 error ("Invalid escape character syntax");
2236 c = READCHAR;
2237 if (c == '\\')
2238 c = read_escape (readcharfun, 0);
2239 return c | meta_modifier;
2241 case 'S':
2242 c = READCHAR;
2243 if (c != '-')
2244 error ("Invalid escape character syntax");
2245 c = READCHAR;
2246 if (c == '\\')
2247 c = read_escape (readcharfun, 0);
2248 return c | shift_modifier;
2250 case 'H':
2251 c = READCHAR;
2252 if (c != '-')
2253 error ("Invalid escape character syntax");
2254 c = READCHAR;
2255 if (c == '\\')
2256 c = read_escape (readcharfun, 0);
2257 return c | hyper_modifier;
2259 case 'A':
2260 c = READCHAR;
2261 if (c != '-')
2262 error ("Invalid escape character syntax");
2263 c = READCHAR;
2264 if (c == '\\')
2265 c = read_escape (readcharfun, 0);
2266 return c | alt_modifier;
2268 case 's':
2269 c = READCHAR;
2270 if (stringp || c != '-')
2272 UNREAD (c);
2273 return ' ';
2275 c = READCHAR;
2276 if (c == '\\')
2277 c = read_escape (readcharfun, 0);
2278 return c | super_modifier;
2280 case 'C':
2281 c = READCHAR;
2282 if (c != '-')
2283 error ("Invalid escape character syntax");
2284 case '^':
2285 c = READCHAR;
2286 if (c == '\\')
2287 c = read_escape (readcharfun, 0);
2288 if ((c & ~CHAR_MODIFIER_MASK) == '?')
2289 return 0177 | (c & CHAR_MODIFIER_MASK);
2290 else if (! SINGLE_BYTE_CHAR_P ((c & ~CHAR_MODIFIER_MASK)))
2291 return c | ctrl_modifier;
2292 /* ASCII control chars are made from letters (both cases),
2293 as well as the non-letters within 0100...0137. */
2294 else if ((c & 0137) >= 0101 && (c & 0137) <= 0132)
2295 return (c & (037 | ~0177));
2296 else if ((c & 0177) >= 0100 && (c & 0177) <= 0137)
2297 return (c & (037 | ~0177));
2298 else
2299 return c | ctrl_modifier;
2301 case '0':
2302 case '1':
2303 case '2':
2304 case '3':
2305 case '4':
2306 case '5':
2307 case '6':
2308 case '7':
2309 /* An octal escape, as in ANSI C. */
2311 register int i = c - '0';
2312 register int count = 0;
2313 while (++count < 3)
2315 if ((c = READCHAR) >= '0' && c <= '7')
2317 i *= 8;
2318 i += c - '0';
2320 else
2322 UNREAD (c);
2323 break;
2327 if (i >= 0x80 && i < 0x100)
2328 i = BYTE8_TO_CHAR (i);
2329 return i;
2332 case 'x':
2333 /* A hex escape, as in ANSI C. */
2335 unsigned int i = 0;
2336 int count = 0;
2337 while (1)
2339 c = READCHAR;
2340 if (c >= '0' && c <= '9')
2342 i *= 16;
2343 i += c - '0';
2345 else if ((c >= 'a' && c <= 'f')
2346 || (c >= 'A' && c <= 'F'))
2348 i *= 16;
2349 if (c >= 'a' && c <= 'f')
2350 i += c - 'a' + 10;
2351 else
2352 i += c - 'A' + 10;
2354 else
2356 UNREAD (c);
2357 break;
2359 /* Allow hex escapes as large as ?\xfffffff, because some
2360 packages use them to denote characters with modifiers. */
2361 if ((CHAR_META | (CHAR_META - 1)) < i)
2362 error ("Hex character out of range: \\x%x...", i);
2363 count += count < 3;
2366 if (count < 3 && i >= 0x80)
2367 return BYTE8_TO_CHAR (i);
2368 return i;
2371 case 'U':
2372 /* Post-Unicode-2.0: Up to eight hex chars. */
2373 unicode_hex_count = 8;
2374 case 'u':
2376 /* A Unicode escape. We only permit them in strings and characters,
2377 not arbitrarily in the source code, as in some other languages. */
2379 unsigned int i = 0;
2380 int count = 0;
2382 while (++count <= unicode_hex_count)
2384 c = READCHAR;
2385 /* `isdigit' and `isalpha' may be locale-specific, which we don't
2386 want. */
2387 if (c >= '0' && c <= '9') i = (i << 4) + (c - '0');
2388 else if (c >= 'a' && c <= 'f') i = (i << 4) + (c - 'a') + 10;
2389 else if (c >= 'A' && c <= 'F') i = (i << 4) + (c - 'A') + 10;
2390 else
2391 error ("Non-hex digit used for Unicode escape");
2393 if (i > 0x10FFFF)
2394 error ("Non-Unicode character: 0x%x", i);
2395 return i;
2398 case 'N':
2399 /* Named character. */
2401 c = READCHAR;
2402 if (c != '{')
2403 invalid_syntax ("Expected opening brace after \\N");
2404 char name[UNICODE_CHARACTER_NAME_LENGTH_BOUND + 1];
2405 bool whitespace = false;
2406 ptrdiff_t length = 0;
2407 while (true)
2409 c = READCHAR;
2410 if (c < 0)
2411 end_of_file_error ();
2412 if (c == '}')
2413 break;
2414 if (! (0 < c && c < 0x80))
2416 AUTO_STRING (format,
2417 "Invalid character U+%04X in character name");
2418 xsignal1 (Qinvalid_read_syntax,
2419 CALLN (Fformat, format, make_natnum (c)));
2421 /* Treat multiple adjacent whitespace characters as a
2422 single space character. This makes it easier to use
2423 character names in e.g. multi-line strings. */
2424 if (c_isspace (c))
2426 if (whitespace)
2427 continue;
2428 c = ' ';
2429 whitespace = true;
2431 else
2432 whitespace = false;
2433 name[length++] = c;
2434 if (length >= sizeof name)
2435 invalid_syntax ("Character name too long");
2437 if (length == 0)
2438 invalid_syntax ("Empty character name");
2439 name[length] = '\0';
2440 return character_name_to_code (name, length);
2443 default:
2444 return c;
2448 /* Return the digit that CHARACTER stands for in the given BASE.
2449 Return -1 if CHARACTER is out of range for BASE,
2450 and -2 if CHARACTER is not valid for any supported BASE. */
2451 static int
2452 digit_to_number (int character, int base)
2454 int digit;
2456 if ('0' <= character && character <= '9')
2457 digit = character - '0';
2458 else if ('a' <= character && character <= 'z')
2459 digit = character - 'a' + 10;
2460 else if ('A' <= character && character <= 'Z')
2461 digit = character - 'A' + 10;
2462 else
2463 return -2;
2465 return digit < base ? digit : -1;
2468 /* Read an integer in radix RADIX using READCHARFUN to read
2469 characters. RADIX must be in the interval [2..36]; if it isn't, a
2470 read error is signaled . Value is the integer read. Signals an
2471 error if encountering invalid read syntax or if RADIX is out of
2472 range. */
2474 static Lisp_Object
2475 read_integer (Lisp_Object readcharfun, EMACS_INT radix)
2477 /* Room for sign, leading 0, other digits, trailing null byte.
2478 Also, room for invalid syntax diagnostic. */
2479 char buf[max (1 + 1 + sizeof (uintmax_t) * CHAR_BIT + 1,
2480 sizeof "integer, radix " + INT_STRLEN_BOUND (EMACS_INT))];
2482 int valid = -1; /* 1 if valid, 0 if not, -1 if incomplete. */
2484 if (radix < 2 || radix > 36)
2485 valid = 0;
2486 else
2488 char *p = buf;
2489 int c, digit;
2491 c = READCHAR;
2492 if (c == '-' || c == '+')
2494 *p++ = c;
2495 c = READCHAR;
2498 if (c == '0')
2500 *p++ = c;
2501 valid = 1;
2503 /* Ignore redundant leading zeros, so the buffer doesn't
2504 fill up with them. */
2506 c = READCHAR;
2507 while (c == '0');
2510 while ((digit = digit_to_number (c, radix)) >= -1)
2512 if (digit == -1)
2513 valid = 0;
2514 if (valid < 0)
2515 valid = 1;
2517 if (p < buf + sizeof buf - 1)
2518 *p++ = c;
2519 else
2520 valid = 0;
2522 c = READCHAR;
2525 UNREAD (c);
2526 *p = '\0';
2529 if (! valid)
2531 sprintf (buf, "integer, radix %"pI"d", radix);
2532 invalid_syntax (buf);
2535 return string_to_number (buf, radix, 0);
2539 /* If the next token is ')' or ']' or '.', we store that character
2540 in *PCH and the return value is not interesting. Else, we store
2541 zero in *PCH and we read and return one lisp object.
2543 FIRST_IN_LIST is true if this is the first element of a list. */
2545 static Lisp_Object
2546 read1 (Lisp_Object readcharfun, int *pch, bool first_in_list)
2548 int c;
2549 bool uninterned_symbol = 0;
2550 bool multibyte;
2552 *pch = 0;
2554 retry:
2556 c = READCHAR_REPORT_MULTIBYTE (&multibyte);
2557 if (c < 0)
2558 end_of_file_error ();
2560 switch (c)
2562 case '(':
2563 return read_list (0, readcharfun);
2565 case '[':
2566 return read_vector (readcharfun, 0);
2568 case ')':
2569 case ']':
2571 *pch = c;
2572 return Qnil;
2575 case '#':
2576 c = READCHAR;
2577 if (c == 's')
2579 c = READCHAR;
2580 if (c == '(')
2582 /* Accept extended format for hashtables (extensible to
2583 other types), e.g.
2584 #s(hash-table size 2 test equal data (k1 v1 k2 v2)) */
2585 Lisp_Object tmp = read_list (0, readcharfun);
2586 Lisp_Object head = CAR_SAFE (tmp);
2587 Lisp_Object data = Qnil;
2588 Lisp_Object val = Qnil;
2589 /* The size is 2 * number of allowed keywords to
2590 make-hash-table. */
2591 Lisp_Object params[10];
2592 Lisp_Object ht;
2593 Lisp_Object key = Qnil;
2594 int param_count = 0;
2596 if (!EQ (head, Qhash_table))
2597 error ("Invalid extended read marker at head of #s list "
2598 "(only hash-table allowed)");
2600 tmp = CDR_SAFE (tmp);
2602 /* This is repetitive but fast and simple. */
2603 params[param_count] = QCsize;
2604 params[param_count + 1] = Fplist_get (tmp, Qsize);
2605 if (!NILP (params[param_count + 1]))
2606 param_count += 2;
2608 params[param_count] = QCtest;
2609 params[param_count + 1] = Fplist_get (tmp, Qtest);
2610 if (!NILP (params[param_count + 1]))
2611 param_count += 2;
2613 params[param_count] = QCweakness;
2614 params[param_count + 1] = Fplist_get (tmp, Qweakness);
2615 if (!NILP (params[param_count + 1]))
2616 param_count += 2;
2618 params[param_count] = QCrehash_size;
2619 params[param_count + 1] = Fplist_get (tmp, Qrehash_size);
2620 if (!NILP (params[param_count + 1]))
2621 param_count += 2;
2623 params[param_count] = QCrehash_threshold;
2624 params[param_count + 1] = Fplist_get (tmp, Qrehash_threshold);
2625 if (!NILP (params[param_count + 1]))
2626 param_count += 2;
2628 /* This is the hashtable data. */
2629 data = Fplist_get (tmp, Qdata);
2631 /* Now use params to make a new hashtable and fill it. */
2632 ht = Fmake_hash_table (param_count, params);
2634 while (CONSP (data))
2636 key = XCAR (data);
2637 data = XCDR (data);
2638 if (!CONSP (data))
2639 error ("Odd number of elements in hashtable data");
2640 val = XCAR (data);
2641 data = XCDR (data);
2642 Fputhash (key, val, ht);
2645 return ht;
2647 UNREAD (c);
2648 invalid_syntax ("#");
2650 if (c == '^')
2652 c = READCHAR;
2653 if (c == '[')
2655 Lisp_Object tmp;
2656 tmp = read_vector (readcharfun, 0);
2657 if (ASIZE (tmp) < CHAR_TABLE_STANDARD_SLOTS)
2658 error ("Invalid size char-table");
2659 XSETPVECTYPE (XVECTOR (tmp), PVEC_CHAR_TABLE);
2660 return tmp;
2662 else if (c == '^')
2664 c = READCHAR;
2665 if (c == '[')
2667 /* Sub char-table can't be read as a regular
2668 vector because of a two C integer fields. */
2669 Lisp_Object tbl, tmp = read_list (1, readcharfun);
2670 ptrdiff_t size = XINT (Flength (tmp));
2671 int i, depth, min_char;
2672 struct Lisp_Cons *cell;
2674 if (size == 0)
2675 error ("Zero-sized sub char-table");
2677 if (! RANGED_INTEGERP (1, XCAR (tmp), 3))
2678 error ("Invalid depth in sub char-table");
2679 depth = XINT (XCAR (tmp));
2680 if (chartab_size[depth] != size - 2)
2681 error ("Invalid size in sub char-table");
2682 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2683 free_cons (cell);
2685 if (! RANGED_INTEGERP (0, XCAR (tmp), MAX_CHAR))
2686 error ("Invalid minimum character in sub-char-table");
2687 min_char = XINT (XCAR (tmp));
2688 cell = XCONS (tmp), tmp = XCDR (tmp), size--;
2689 free_cons (cell);
2691 tbl = make_uninit_sub_char_table (depth, min_char);
2692 for (i = 0; i < size; i++)
2694 XSUB_CHAR_TABLE (tbl)->contents[i] = XCAR (tmp);
2695 cell = XCONS (tmp), tmp = XCDR (tmp);
2696 free_cons (cell);
2698 return tbl;
2700 invalid_syntax ("#^^");
2702 invalid_syntax ("#^");
2704 if (c == '&')
2706 Lisp_Object length;
2707 length = read1 (readcharfun, pch, first_in_list);
2708 c = READCHAR;
2709 if (c == '"')
2711 Lisp_Object tmp, val;
2712 EMACS_INT size_in_chars = bool_vector_bytes (XFASTINT (length));
2713 unsigned char *data;
2715 UNREAD (c);
2716 tmp = read1 (readcharfun, pch, first_in_list);
2717 if (STRING_MULTIBYTE (tmp)
2718 || (size_in_chars != SCHARS (tmp)
2719 /* We used to print 1 char too many
2720 when the number of bits was a multiple of 8.
2721 Accept such input in case it came from an old
2722 version. */
2723 && ! (XFASTINT (length)
2724 == (SCHARS (tmp) - 1) * BOOL_VECTOR_BITS_PER_CHAR)))
2725 invalid_syntax ("#&...");
2727 val = make_uninit_bool_vector (XFASTINT (length));
2728 data = bool_vector_uchar_data (val);
2729 memcpy (data, SDATA (tmp), size_in_chars);
2730 /* Clear the extraneous bits in the last byte. */
2731 if (XINT (length) != size_in_chars * BOOL_VECTOR_BITS_PER_CHAR)
2732 data[size_in_chars - 1]
2733 &= (1 << (XINT (length) % BOOL_VECTOR_BITS_PER_CHAR)) - 1;
2734 return val;
2736 invalid_syntax ("#&...");
2738 if (c == '[')
2740 /* Accept compiled functions at read-time so that we don't have to
2741 build them using function calls. */
2742 Lisp_Object tmp;
2743 struct Lisp_Vector *vec;
2744 tmp = read_vector (readcharfun, 1);
2745 vec = XVECTOR (tmp);
2746 if (vec->header.size == 0)
2747 invalid_syntax ("Empty byte-code object");
2748 make_byte_code (vec);
2749 return tmp;
2751 if (c == '(')
2753 Lisp_Object tmp;
2754 int ch;
2756 /* Read the string itself. */
2757 tmp = read1 (readcharfun, &ch, 0);
2758 if (ch != 0 || !STRINGP (tmp))
2759 invalid_syntax ("#");
2760 /* Read the intervals and their properties. */
2761 while (1)
2763 Lisp_Object beg, end, plist;
2765 beg = read1 (readcharfun, &ch, 0);
2766 end = plist = Qnil;
2767 if (ch == ')')
2768 break;
2769 if (ch == 0)
2770 end = read1 (readcharfun, &ch, 0);
2771 if (ch == 0)
2772 plist = read1 (readcharfun, &ch, 0);
2773 if (ch)
2774 invalid_syntax ("Invalid string property list");
2775 Fset_text_properties (beg, end, plist, tmp);
2778 return tmp;
2781 /* #@NUMBER is used to skip NUMBER following bytes.
2782 That's used in .elc files to skip over doc strings
2783 and function definitions. */
2784 if (c == '@')
2786 enum { extra = 100 };
2787 ptrdiff_t i, nskip = 0, digits = 0;
2789 /* Read a decimal integer. */
2790 while ((c = READCHAR) >= 0
2791 && c >= '0' && c <= '9')
2793 if ((STRING_BYTES_BOUND - extra) / 10 <= nskip)
2794 string_overflow ();
2795 digits++;
2796 nskip *= 10;
2797 nskip += c - '0';
2798 if (digits == 2 && nskip == 0)
2799 { /* We've just seen #@00, which means "skip to end". */
2800 skip_dyn_eof (readcharfun);
2801 return Qnil;
2804 if (nskip > 0)
2805 /* We can't use UNREAD here, because in the code below we side-step
2806 READCHAR. Instead, assume the first char after #@NNN occupies
2807 a single byte, which is the case normally since it's just
2808 a space. */
2809 nskip--;
2810 else
2811 UNREAD (c);
2813 if (load_force_doc_strings
2814 && (FROM_FILE_P (readcharfun)))
2816 /* If we are supposed to force doc strings into core right now,
2817 record the last string that we skipped,
2818 and record where in the file it comes from. */
2820 /* But first exchange saved_doc_string
2821 with prev_saved_doc_string, so we save two strings. */
2823 char *temp = saved_doc_string;
2824 ptrdiff_t temp_size = saved_doc_string_size;
2825 file_offset temp_pos = saved_doc_string_position;
2826 ptrdiff_t temp_len = saved_doc_string_length;
2828 saved_doc_string = prev_saved_doc_string;
2829 saved_doc_string_size = prev_saved_doc_string_size;
2830 saved_doc_string_position = prev_saved_doc_string_position;
2831 saved_doc_string_length = prev_saved_doc_string_length;
2833 prev_saved_doc_string = temp;
2834 prev_saved_doc_string_size = temp_size;
2835 prev_saved_doc_string_position = temp_pos;
2836 prev_saved_doc_string_length = temp_len;
2839 if (saved_doc_string_size == 0)
2841 saved_doc_string = xmalloc (nskip + extra);
2842 saved_doc_string_size = nskip + extra;
2844 if (nskip > saved_doc_string_size)
2846 saved_doc_string = xrealloc (saved_doc_string, nskip + extra);
2847 saved_doc_string_size = nskip + extra;
2850 saved_doc_string_position = file_tell (instream);
2852 /* Copy that many characters into saved_doc_string. */
2853 block_input ();
2854 for (i = 0; i < nskip && c >= 0; i++)
2855 saved_doc_string[i] = c = getc (instream);
2856 unblock_input ();
2858 saved_doc_string_length = i;
2860 else
2861 /* Skip that many bytes. */
2862 skip_dyn_bytes (readcharfun, nskip);
2864 goto retry;
2866 if (c == '!')
2868 /* #! appears at the beginning of an executable file.
2869 Skip the first line. */
2870 while (c != '\n' && c >= 0)
2871 c = READCHAR;
2872 goto retry;
2874 if (c == '$')
2875 return Vload_file_name;
2876 if (c == '\'')
2877 return list2 (Qfunction, read0 (readcharfun));
2878 /* #:foo is the uninterned symbol named foo. */
2879 if (c == ':')
2881 uninterned_symbol = 1;
2882 c = READCHAR;
2883 if (!(c > 040
2884 && c != NO_BREAK_SPACE
2885 && (c >= 0200
2886 || strchr ("\"';()[]#`,", c) == NULL)))
2888 /* No symbol character follows, this is the empty
2889 symbol. */
2890 UNREAD (c);
2891 return Fmake_symbol (empty_unibyte_string);
2893 goto read_symbol;
2895 /* ## is the empty symbol. */
2896 if (c == '#')
2897 return Fintern (empty_unibyte_string, Qnil);
2898 /* Reader forms that can reuse previously read objects. */
2899 if (c >= '0' && c <= '9')
2901 EMACS_INT n = 0;
2902 Lisp_Object tem;
2904 /* Read a non-negative integer. */
2905 while (c >= '0' && c <= '9')
2907 if (MOST_POSITIVE_FIXNUM / 10 < n
2908 || MOST_POSITIVE_FIXNUM < n * 10 + c - '0')
2909 n = MOST_POSITIVE_FIXNUM + 1;
2910 else
2911 n = n * 10 + c - '0';
2912 c = READCHAR;
2915 if (n <= MOST_POSITIVE_FIXNUM)
2917 if (c == 'r' || c == 'R')
2918 return read_integer (readcharfun, n);
2920 if (! NILP (Vread_circle))
2922 /* #n=object returns object, but associates it with
2923 n for #n#. */
2924 if (c == '=')
2926 /* Make a placeholder for #n# to use temporarily. */
2927 AUTO_CONS (placeholder, Qnil, Qnil);
2928 Lisp_Object cell = Fcons (make_number (n), placeholder);
2929 read_objects = Fcons (cell, read_objects);
2931 /* Read the object itself. */
2932 tem = read0 (readcharfun);
2934 /* Now put it everywhere the placeholder was... */
2935 substitute_object_in_subtree (tem, placeholder);
2937 /* ...and #n# will use the real value from now on. */
2938 Fsetcdr (cell, tem);
2940 return tem;
2943 /* #n# returns a previously read object. */
2944 if (c == '#')
2946 tem = Fassq (make_number (n), read_objects);
2947 if (CONSP (tem))
2948 return XCDR (tem);
2952 /* Fall through to error message. */
2954 else if (c == 'x' || c == 'X')
2955 return read_integer (readcharfun, 16);
2956 else if (c == 'o' || c == 'O')
2957 return read_integer (readcharfun, 8);
2958 else if (c == 'b' || c == 'B')
2959 return read_integer (readcharfun, 2);
2961 UNREAD (c);
2962 invalid_syntax ("#");
2964 case ';':
2965 while ((c = READCHAR) >= 0 && c != '\n');
2966 goto retry;
2968 case '\'':
2969 return list2 (Qquote, read0 (readcharfun));
2971 case '`':
2973 int next_char = READCHAR;
2974 UNREAD (next_char);
2975 /* Transition from old-style to new-style:
2976 If we see "(`" it used to mean old-style, which usually works
2977 fine because ` should almost never appear in such a position
2978 for new-style. But occasionally we need "(`" to mean new
2979 style, so we try to distinguish the two by the fact that we
2980 can either write "( `foo" or "(` foo", where the first
2981 intends to use new-style whereas the second intends to use
2982 old-style. For Emacs-25, we should completely remove this
2983 first_in_list exception (old-style can still be obtained via
2984 "(\`" anyway). */
2985 if (!new_backquote_flag && first_in_list && next_char == ' ')
2987 Vold_style_backquotes = Qt;
2988 goto default_label;
2990 else
2992 Lisp_Object value;
2993 bool saved_new_backquote_flag = new_backquote_flag;
2995 new_backquote_flag = 1;
2996 value = read0 (readcharfun);
2997 new_backquote_flag = saved_new_backquote_flag;
2999 return list2 (Qbackquote, value);
3002 case ',':
3004 int next_char = READCHAR;
3005 UNREAD (next_char);
3006 /* Transition from old-style to new-style:
3007 It used to be impossible to have a new-style , other than within
3008 a new-style `. This is sufficient when ` and , are used in the
3009 normal way, but ` and , can also appear in args to macros that
3010 will not interpret them in the usual way, in which case , may be
3011 used without any ` anywhere near.
3012 So we now use the same heuristic as for backquote: old-style
3013 unquotes are only recognized when first on a list, and when
3014 followed by a space.
3015 Because it's more difficult to peek 2 chars ahead, a new-style
3016 ,@ can still not be used outside of a `, unless it's in the middle
3017 of a list. */
3018 if (new_backquote_flag
3019 || !first_in_list
3020 || (next_char != ' ' && next_char != '@'))
3022 Lisp_Object comma_type = Qnil;
3023 Lisp_Object value;
3024 int ch = READCHAR;
3026 if (ch == '@')
3027 comma_type = Qcomma_at;
3028 else if (ch == '.')
3029 comma_type = Qcomma_dot;
3030 else
3032 if (ch >= 0) UNREAD (ch);
3033 comma_type = Qcomma;
3036 value = read0 (readcharfun);
3037 return list2 (comma_type, value);
3039 else
3041 Vold_style_backquotes = Qt;
3042 goto default_label;
3045 case '?':
3047 int modifiers;
3048 int next_char;
3049 bool ok;
3051 c = READCHAR;
3052 if (c < 0)
3053 end_of_file_error ();
3055 /* Accept `single space' syntax like (list ? x) where the
3056 whitespace character is SPC or TAB.
3057 Other literal whitespace like NL, CR, and FF are not accepted,
3058 as there are well-established escape sequences for these. */
3059 if (c == ' ' || c == '\t')
3060 return make_number (c);
3062 if (c == '\\')
3063 c = read_escape (readcharfun, 0);
3064 modifiers = c & CHAR_MODIFIER_MASK;
3065 c &= ~CHAR_MODIFIER_MASK;
3066 if (CHAR_BYTE8_P (c))
3067 c = CHAR_TO_BYTE8 (c);
3068 c |= modifiers;
3070 next_char = READCHAR;
3071 ok = (next_char <= 040
3072 || (next_char < 0200
3073 && strchr ("\"';()[]#?`,.", next_char) != NULL));
3074 UNREAD (next_char);
3075 if (ok)
3076 return make_number (c);
3078 invalid_syntax ("?");
3081 case '"':
3083 char *p = read_buffer;
3084 char *end = read_buffer + read_buffer_size;
3085 int ch;
3086 /* True if we saw an escape sequence specifying
3087 a multibyte character. */
3088 bool force_multibyte = 0;
3089 /* True if we saw an escape sequence specifying
3090 a single-byte character. */
3091 bool force_singlebyte = 0;
3092 bool cancel = 0;
3093 ptrdiff_t nchars = 0;
3095 while ((ch = READCHAR) >= 0
3096 && ch != '\"')
3098 if (end - p < MAX_MULTIBYTE_LENGTH)
3100 ptrdiff_t offset = p - read_buffer;
3101 grow_read_buffer ();
3102 p = read_buffer + offset;
3103 end = read_buffer + read_buffer_size;
3106 if (ch == '\\')
3108 int modifiers;
3110 ch = read_escape (readcharfun, 1);
3112 /* CH is -1 if \ newline or \ space has just been seen. */
3113 if (ch == -1)
3115 if (p == read_buffer)
3116 cancel = 1;
3117 continue;
3120 modifiers = ch & CHAR_MODIFIER_MASK;
3121 ch = ch & ~CHAR_MODIFIER_MASK;
3123 if (CHAR_BYTE8_P (ch))
3124 force_singlebyte = 1;
3125 else if (! ASCII_CHAR_P (ch))
3126 force_multibyte = 1;
3127 else /* I.e. ASCII_CHAR_P (ch). */
3129 /* Allow `\C- ' and `\C-?'. */
3130 if (modifiers == CHAR_CTL)
3132 if (ch == ' ')
3133 ch = 0, modifiers = 0;
3134 else if (ch == '?')
3135 ch = 127, modifiers = 0;
3137 if (modifiers & CHAR_SHIFT)
3139 /* Shift modifier is valid only with [A-Za-z]. */
3140 if (ch >= 'A' && ch <= 'Z')
3141 modifiers &= ~CHAR_SHIFT;
3142 else if (ch >= 'a' && ch <= 'z')
3143 ch -= ('a' - 'A'), modifiers &= ~CHAR_SHIFT;
3146 if (modifiers & CHAR_META)
3148 /* Move the meta bit to the right place for a
3149 string. */
3150 modifiers &= ~CHAR_META;
3151 ch = BYTE8_TO_CHAR (ch | 0x80);
3152 force_singlebyte = 1;
3156 /* Any modifiers remaining are invalid. */
3157 if (modifiers)
3158 error ("Invalid modifier in string");
3159 p += CHAR_STRING (ch, (unsigned char *) p);
3161 else
3163 p += CHAR_STRING (ch, (unsigned char *) p);
3164 if (CHAR_BYTE8_P (ch))
3165 force_singlebyte = 1;
3166 else if (! ASCII_CHAR_P (ch))
3167 force_multibyte = 1;
3169 nchars++;
3172 if (ch < 0)
3173 end_of_file_error ();
3175 /* If purifying, and string starts with \ newline,
3176 return zero instead. This is for doc strings
3177 that we are really going to find in etc/DOC.nn.nn. */
3178 if (!NILP (Vpurify_flag) && NILP (Vdoc_file_name) && cancel)
3179 return make_number (0);
3181 if (! force_multibyte && force_singlebyte)
3183 /* READ_BUFFER contains raw 8-bit bytes and no multibyte
3184 forms. Convert it to unibyte. */
3185 nchars = str_as_unibyte ((unsigned char *) read_buffer,
3186 p - read_buffer);
3187 p = read_buffer + nchars;
3190 return make_specified_string (read_buffer, nchars, p - read_buffer,
3191 (force_multibyte
3192 || (p - read_buffer != nchars)));
3195 case '.':
3197 int next_char = READCHAR;
3198 UNREAD (next_char);
3200 if (next_char <= 040
3201 || (next_char < 0200
3202 && strchr ("\"';([#?`,", next_char) != NULL))
3204 *pch = c;
3205 return Qnil;
3208 /* Otherwise, we fall through! Note that the atom-reading loop
3209 below will now loop at least once, assuring that we will not
3210 try to UNREAD two characters in a row. */
3212 default:
3213 default_label:
3214 if (c <= 040) goto retry;
3215 if (c == NO_BREAK_SPACE)
3216 goto retry;
3218 read_symbol:
3220 char *p = read_buffer;
3221 bool quoted = 0;
3222 EMACS_INT start_position = readchar_count - 1;
3225 char *end = read_buffer + read_buffer_size;
3229 if (end - p < MAX_MULTIBYTE_LENGTH)
3231 ptrdiff_t offset = p - read_buffer;
3232 grow_read_buffer ();
3233 p = read_buffer + offset;
3234 end = read_buffer + read_buffer_size;
3237 if (c == '\\')
3239 c = READCHAR;
3240 if (c == -1)
3241 end_of_file_error ();
3242 quoted = 1;
3245 if (multibyte)
3246 p += CHAR_STRING (c, (unsigned char *) p);
3247 else
3248 *p++ = c;
3249 c = READCHAR;
3251 while (c > 040
3252 && c != NO_BREAK_SPACE
3253 && (c >= 0200
3254 || strchr ("\"';()[]#`,", c) == NULL));
3256 if (p == end)
3258 ptrdiff_t offset = p - read_buffer;
3259 grow_read_buffer ();
3260 p = read_buffer + offset;
3261 end = read_buffer + read_buffer_size;
3263 *p = 0;
3264 UNREAD (c);
3267 if (!quoted && !uninterned_symbol)
3269 Lisp_Object result = string_to_number (read_buffer, 10, 0);
3270 if (! NILP (result))
3271 return result;
3274 Lisp_Object name, result;
3275 ptrdiff_t nbytes = p - read_buffer;
3276 ptrdiff_t nchars
3277 = (multibyte
3278 ? multibyte_chars_in_text ((unsigned char *) read_buffer,
3279 nbytes)
3280 : nbytes);
3282 name = ((uninterned_symbol && ! NILP (Vpurify_flag)
3283 ? make_pure_string : make_specified_string)
3284 (read_buffer, nchars, nbytes, multibyte));
3285 result = (uninterned_symbol ? Fmake_symbol (name)
3286 : Fintern (name, Qnil));
3288 if (EQ (Vread_with_symbol_positions, Qt)
3289 || EQ (Vread_with_symbol_positions, readcharfun))
3290 Vread_symbol_positions_list
3291 = Fcons (Fcons (result, make_number (start_position)),
3292 Vread_symbol_positions_list);
3293 return result;
3300 /* List of nodes we've seen during substitute_object_in_subtree. */
3301 static Lisp_Object seen_list;
3303 static void
3304 substitute_object_in_subtree (Lisp_Object object, Lisp_Object placeholder)
3306 Lisp_Object check_object;
3308 /* We haven't seen any objects when we start. */
3309 seen_list = Qnil;
3311 /* Make all the substitutions. */
3312 check_object
3313 = substitute_object_recurse (object, placeholder, object);
3315 /* Clear seen_list because we're done with it. */
3316 seen_list = Qnil;
3318 /* The returned object here is expected to always eq the
3319 original. */
3320 if (!EQ (check_object, object))
3321 error ("Unexpected mutation error in reader");
3324 /* Feval doesn't get called from here, so no gc protection is needed. */
3325 #define SUBSTITUTE(get_val, set_val) \
3326 do { \
3327 Lisp_Object old_value = get_val; \
3328 Lisp_Object true_value \
3329 = substitute_object_recurse (object, placeholder, \
3330 old_value); \
3332 if (!EQ (old_value, true_value)) \
3334 set_val; \
3336 } while (0)
3338 static Lisp_Object
3339 substitute_object_recurse (Lisp_Object object, Lisp_Object placeholder, Lisp_Object subtree)
3341 /* If we find the placeholder, return the target object. */
3342 if (EQ (placeholder, subtree))
3343 return object;
3345 /* If we've been to this node before, don't explore it again. */
3346 if (!EQ (Qnil, Fmemq (subtree, seen_list)))
3347 return subtree;
3349 /* If this node can be the entry point to a cycle, remember that
3350 we've seen it. It can only be such an entry point if it was made
3351 by #n=, which means that we can find it as a value in
3352 read_objects. */
3353 if (!EQ (Qnil, Frassq (subtree, read_objects)))
3354 seen_list = Fcons (subtree, seen_list);
3356 /* Recurse according to subtree's type.
3357 Every branch must return a Lisp_Object. */
3358 switch (XTYPE (subtree))
3360 case Lisp_Vectorlike:
3362 ptrdiff_t i = 0, length = 0;
3363 if (BOOL_VECTOR_P (subtree))
3364 return subtree; /* No sub-objects anyway. */
3365 else if (CHAR_TABLE_P (subtree) || SUB_CHAR_TABLE_P (subtree)
3366 || COMPILEDP (subtree) || HASH_TABLE_P (subtree))
3367 length = ASIZE (subtree) & PSEUDOVECTOR_SIZE_MASK;
3368 else if (VECTORP (subtree))
3369 length = ASIZE (subtree);
3370 else
3371 /* An unknown pseudovector may contain non-Lisp fields, so we
3372 can't just blindly traverse all its fields. We used to call
3373 `Flength' which signaled `sequencep', so I just preserved this
3374 behavior. */
3375 wrong_type_argument (Qsequencep, subtree);
3377 if (SUB_CHAR_TABLE_P (subtree))
3378 i = 2;
3379 for ( ; i < length; i++)
3380 SUBSTITUTE (AREF (subtree, i),
3381 ASET (subtree, i, true_value));
3382 return subtree;
3385 case Lisp_Cons:
3387 SUBSTITUTE (XCAR (subtree),
3388 XSETCAR (subtree, true_value));
3389 SUBSTITUTE (XCDR (subtree),
3390 XSETCDR (subtree, true_value));
3391 return subtree;
3394 case Lisp_String:
3396 /* Check for text properties in each interval.
3397 substitute_in_interval contains part of the logic. */
3399 INTERVAL root_interval = string_intervals (subtree);
3400 AUTO_CONS (arg, object, placeholder);
3402 traverse_intervals_noorder (root_interval,
3403 &substitute_in_interval, arg);
3405 return subtree;
3408 /* Other types don't recurse any further. */
3409 default:
3410 return subtree;
3414 /* Helper function for substitute_object_recurse. */
3415 static void
3416 substitute_in_interval (INTERVAL interval, Lisp_Object arg)
3418 Lisp_Object object = Fcar (arg);
3419 Lisp_Object placeholder = Fcdr (arg);
3421 SUBSTITUTE (interval->plist, set_interval_plist (interval, true_value));
3425 #define LEAD_INT 1
3426 #define DOT_CHAR 2
3427 #define TRAIL_INT 4
3428 #define E_EXP 16
3431 /* Convert STRING to a number, assuming base BASE. Return a fixnum if CP has
3432 integer syntax and fits in a fixnum, else return the nearest float if CP has
3433 either floating point or integer syntax and BASE is 10, else return nil. If
3434 IGNORE_TRAILING, consider just the longest prefix of CP that has
3435 valid floating point syntax. Signal an overflow if BASE is not 10 and the
3436 number has integer syntax but does not fit. */
3438 Lisp_Object
3439 string_to_number (char const *string, int base, bool ignore_trailing)
3441 int state;
3442 char const *cp = string;
3443 int leading_digit;
3444 bool float_syntax = 0;
3445 double value = 0;
3447 /* Negate the value ourselves. This treats 0, NaNs, and infinity properly on
3448 IEEE floating point hosts, and works around a formerly-common bug where
3449 atof ("-0.0") drops the sign. */
3450 bool negative = *cp == '-';
3452 bool signedp = negative || *cp == '+';
3453 cp += signedp;
3455 state = 0;
3457 leading_digit = digit_to_number (*cp, base);
3458 if (leading_digit >= 0)
3460 state |= LEAD_INT;
3462 ++cp;
3463 while (digit_to_number (*cp, base) >= 0);
3465 if (*cp == '.')
3467 state |= DOT_CHAR;
3468 cp++;
3471 if (base == 10)
3473 if ('0' <= *cp && *cp <= '9')
3475 state |= TRAIL_INT;
3477 cp++;
3478 while ('0' <= *cp && *cp <= '9');
3480 if (*cp == 'e' || *cp == 'E')
3482 char const *ecp = cp;
3483 cp++;
3484 if (*cp == '+' || *cp == '-')
3485 cp++;
3486 if ('0' <= *cp && *cp <= '9')
3488 state |= E_EXP;
3490 cp++;
3491 while ('0' <= *cp && *cp <= '9');
3493 else if (cp[-1] == '+'
3494 && cp[0] == 'I' && cp[1] == 'N' && cp[2] == 'F')
3496 state |= E_EXP;
3497 cp += 3;
3498 value = INFINITY;
3500 else if (cp[-1] == '+'
3501 && cp[0] == 'N' && cp[1] == 'a' && cp[2] == 'N')
3503 state |= E_EXP;
3504 cp += 3;
3505 /* NAN is a "positive" NaN on all known Emacs hosts. */
3506 value = NAN;
3508 else
3509 cp = ecp;
3512 float_syntax = ((state & (DOT_CHAR|TRAIL_INT)) == (DOT_CHAR|TRAIL_INT)
3513 || state == (LEAD_INT|E_EXP));
3516 /* Return nil if the number uses invalid syntax. If IGNORE_TRAILING, accept
3517 any prefix that matches. Otherwise, the entire string must match. */
3518 if (! (ignore_trailing
3519 ? ((state & LEAD_INT) != 0 || float_syntax)
3520 : (!*cp && ((state & ~DOT_CHAR) == LEAD_INT || float_syntax))))
3521 return Qnil;
3523 /* If the number uses integer and not float syntax, and is in C-language
3524 range, use its value, preferably as a fixnum. */
3525 if (leading_digit >= 0 && ! float_syntax)
3527 uintmax_t n;
3529 /* Fast special case for single-digit integers. This also avoids a
3530 glitch when BASE is 16 and IGNORE_TRAILING, because in that
3531 case some versions of strtoumax accept numbers like "0x1" that Emacs
3532 does not allow. */
3533 if (digit_to_number (string[signedp + 1], base) < 0)
3534 return make_number (negative ? -leading_digit : leading_digit);
3536 errno = 0;
3537 n = strtoumax (string + signedp, NULL, base);
3538 if (errno == ERANGE)
3540 /* Unfortunately there's no simple and accurate way to convert
3541 non-base-10 numbers that are out of C-language range. */
3542 if (base != 10)
3543 xsignal1 (Qoverflow_error, build_string (string));
3545 else if (n <= (negative ? -MOST_NEGATIVE_FIXNUM : MOST_POSITIVE_FIXNUM))
3547 EMACS_INT signed_n = n;
3548 return make_number (negative ? -signed_n : signed_n);
3550 else
3551 value = n;
3554 /* Either the number uses float syntax, or it does not fit into a fixnum.
3555 Convert it from string to floating point, unless the value is already
3556 known because it is an infinity, a NAN, or its absolute value fits in
3557 uintmax_t. */
3558 if (! value)
3559 value = atof (string + signedp);
3561 return make_float (negative ? -value : value);
3565 static Lisp_Object
3566 read_vector (Lisp_Object readcharfun, bool bytecodeflag)
3568 ptrdiff_t i, size;
3569 Lisp_Object *ptr;
3570 Lisp_Object tem, item, vector;
3571 struct Lisp_Cons *otem;
3572 Lisp_Object len;
3574 tem = read_list (1, readcharfun);
3575 len = Flength (tem);
3576 vector = Fmake_vector (len, Qnil);
3578 size = ASIZE (vector);
3579 ptr = XVECTOR (vector)->contents;
3580 for (i = 0; i < size; i++)
3582 item = Fcar (tem);
3583 /* If `load-force-doc-strings' is t when reading a lazily-loaded
3584 bytecode object, the docstring containing the bytecode and
3585 constants values must be treated as unibyte and passed to
3586 Fread, to get the actual bytecode string and constants vector. */
3587 if (bytecodeflag && load_force_doc_strings)
3589 if (i == COMPILED_BYTECODE)
3591 if (!STRINGP (item))
3592 error ("Invalid byte code");
3594 /* Delay handling the bytecode slot until we know whether
3595 it is lazily-loaded (we can tell by whether the
3596 constants slot is nil). */
3597 ASET (vector, COMPILED_CONSTANTS, item);
3598 item = Qnil;
3600 else if (i == COMPILED_CONSTANTS)
3602 Lisp_Object bytestr = ptr[COMPILED_CONSTANTS];
3604 if (NILP (item))
3606 /* Coerce string to unibyte (like string-as-unibyte,
3607 but without generating extra garbage and
3608 guaranteeing no change in the contents). */
3609 STRING_SET_CHARS (bytestr, SBYTES (bytestr));
3610 STRING_SET_UNIBYTE (bytestr);
3612 item = Fread (Fcons (bytestr, readcharfun));
3613 if (!CONSP (item))
3614 error ("Invalid byte code");
3616 otem = XCONS (item);
3617 bytestr = XCAR (item);
3618 item = XCDR (item);
3619 free_cons (otem);
3622 /* Now handle the bytecode slot. */
3623 ASET (vector, COMPILED_BYTECODE, bytestr);
3625 else if (i == COMPILED_DOC_STRING
3626 && STRINGP (item)
3627 && ! STRING_MULTIBYTE (item))
3629 if (EQ (readcharfun, Qget_emacs_mule_file_char))
3630 item = Fdecode_coding_string (item, Qemacs_mule, Qnil, Qnil);
3631 else
3632 item = Fstring_as_multibyte (item);
3635 ASET (vector, i, item);
3636 otem = XCONS (tem);
3637 tem = Fcdr (tem);
3638 free_cons (otem);
3640 return vector;
3643 /* FLAG means check for ']' to terminate rather than ')' and '.'. */
3645 static Lisp_Object
3646 read_list (bool flag, Lisp_Object readcharfun)
3648 Lisp_Object val, tail;
3649 Lisp_Object elt, tem;
3650 /* 0 is the normal case.
3651 1 means this list is a doc reference; replace it with the number 0.
3652 2 means this list is a doc reference; replace it with the doc string. */
3653 int doc_reference = 0;
3655 /* Initialize this to 1 if we are reading a list. */
3656 bool first_in_list = flag <= 0;
3658 val = Qnil;
3659 tail = Qnil;
3661 while (1)
3663 int ch;
3664 elt = read1 (readcharfun, &ch, first_in_list);
3666 first_in_list = 0;
3668 /* While building, if the list starts with #$, treat it specially. */
3669 if (EQ (elt, Vload_file_name)
3670 && ! NILP (elt)
3671 && !NILP (Vpurify_flag))
3673 if (NILP (Vdoc_file_name))
3674 /* We have not yet called Snarf-documentation, so assume
3675 this file is described in the DOC file
3676 and Snarf-documentation will fill in the right value later.
3677 For now, replace the whole list with 0. */
3678 doc_reference = 1;
3679 else
3680 /* We have already called Snarf-documentation, so make a relative
3681 file name for this file, so it can be found properly
3682 in the installed Lisp directory.
3683 We don't use Fexpand_file_name because that would make
3684 the directory absolute now. */
3686 AUTO_STRING (dot_dot_lisp, "../lisp/");
3687 elt = concat2 (dot_dot_lisp, Ffile_name_nondirectory (elt));
3690 else if (EQ (elt, Vload_file_name)
3691 && ! NILP (elt)
3692 && load_force_doc_strings)
3693 doc_reference = 2;
3695 if (ch)
3697 if (flag > 0)
3699 if (ch == ']')
3700 return val;
3701 invalid_syntax (") or . in a vector");
3703 if (ch == ')')
3704 return val;
3705 if (ch == '.')
3707 if (!NILP (tail))
3708 XSETCDR (tail, read0 (readcharfun));
3709 else
3710 val = read0 (readcharfun);
3711 read1 (readcharfun, &ch, 0);
3713 if (ch == ')')
3715 if (doc_reference == 1)
3716 return make_number (0);
3717 if (doc_reference == 2 && INTEGERP (XCDR (val)))
3719 char *saved = NULL;
3720 file_offset saved_position;
3721 /* Get a doc string from the file we are loading.
3722 If it's in saved_doc_string, get it from there.
3724 Here, we don't know if the string is a
3725 bytecode string or a doc string. As a
3726 bytecode string must be unibyte, we always
3727 return a unibyte string. If it is actually a
3728 doc string, caller must make it
3729 multibyte. */
3731 /* Position is negative for user variables. */
3732 EMACS_INT pos = eabs (XINT (XCDR (val)));
3733 if (pos >= saved_doc_string_position
3734 && pos < (saved_doc_string_position
3735 + saved_doc_string_length))
3737 saved = saved_doc_string;
3738 saved_position = saved_doc_string_position;
3740 /* Look in prev_saved_doc_string the same way. */
3741 else if (pos >= prev_saved_doc_string_position
3742 && pos < (prev_saved_doc_string_position
3743 + prev_saved_doc_string_length))
3745 saved = prev_saved_doc_string;
3746 saved_position = prev_saved_doc_string_position;
3748 if (saved)
3750 ptrdiff_t start = pos - saved_position;
3751 ptrdiff_t from, to;
3753 /* Process quoting with ^A,
3754 and find the end of the string,
3755 which is marked with ^_ (037). */
3756 for (from = start, to = start;
3757 saved[from] != 037;)
3759 int c = saved[from++];
3760 if (c == 1)
3762 c = saved[from++];
3763 saved[to++] = (c == 1 ? c
3764 : c == '0' ? 0
3765 : c == '_' ? 037
3766 : c);
3768 else
3769 saved[to++] = c;
3772 return make_unibyte_string (saved + start,
3773 to - start);
3775 else
3776 return get_doc_string (val, 1, 0);
3779 return val;
3781 invalid_syntax (". in wrong context");
3783 invalid_syntax ("] in a list");
3785 tem = list1 (elt);
3786 if (!NILP (tail))
3787 XSETCDR (tail, tem);
3788 else
3789 val = tem;
3790 tail = tem;
3794 static Lisp_Object initial_obarray;
3796 /* `oblookup' stores the bucket number here, for the sake of Funintern. */
3798 static size_t oblookup_last_bucket_number;
3800 /* Get an error if OBARRAY is not an obarray.
3801 If it is one, return it. */
3803 Lisp_Object
3804 check_obarray (Lisp_Object obarray)
3806 /* We don't want to signal a wrong-type-argument error when we are
3807 shutting down due to a fatal error, and we don't want to hit
3808 assertions in VECTORP and ASIZE if the fatal error was during GC. */
3809 if (!fatal_error_in_progress
3810 && (!VECTORP (obarray) || ASIZE (obarray) == 0))
3812 /* If Vobarray is now invalid, force it to be valid. */
3813 if (EQ (Vobarray, obarray)) Vobarray = initial_obarray;
3814 wrong_type_argument (Qvectorp, obarray);
3816 return obarray;
3819 /* Intern symbol SYM in OBARRAY using bucket INDEX. */
3821 static Lisp_Object
3822 intern_sym (Lisp_Object sym, Lisp_Object obarray, Lisp_Object index)
3824 Lisp_Object *ptr;
3826 XSYMBOL (sym)->interned = (EQ (obarray, initial_obarray)
3827 ? SYMBOL_INTERNED_IN_INITIAL_OBARRAY
3828 : SYMBOL_INTERNED);
3830 if (SREF (SYMBOL_NAME (sym), 0) == ':' && EQ (obarray, initial_obarray))
3832 XSYMBOL (sym)->constant = 1;
3833 XSYMBOL (sym)->redirect = SYMBOL_PLAINVAL;
3834 SET_SYMBOL_VAL (XSYMBOL (sym), sym);
3837 ptr = aref_addr (obarray, XINT (index));
3838 set_symbol_next (sym, SYMBOLP (*ptr) ? XSYMBOL (*ptr) : NULL);
3839 *ptr = sym;
3840 return sym;
3843 /* Intern a symbol with name STRING in OBARRAY using bucket INDEX. */
3845 Lisp_Object
3846 intern_driver (Lisp_Object string, Lisp_Object obarray, Lisp_Object index)
3848 return intern_sym (Fmake_symbol (string), obarray, index);
3851 /* Intern the C string STR: return a symbol with that name,
3852 interned in the current obarray. */
3854 Lisp_Object
3855 intern_1 (const char *str, ptrdiff_t len)
3857 Lisp_Object obarray = check_obarray (Vobarray);
3858 Lisp_Object tem = oblookup (obarray, str, len, len);
3860 return (SYMBOLP (tem) ? tem
3861 /* The above `oblookup' was done on the basis of nchars==nbytes, so
3862 the string has to be unibyte. */
3863 : intern_driver (make_unibyte_string (str, len),
3864 obarray, tem));
3867 Lisp_Object
3868 intern_c_string_1 (const char *str, ptrdiff_t len)
3870 Lisp_Object obarray = check_obarray (Vobarray);
3871 Lisp_Object tem = oblookup (obarray, str, len, len);
3873 if (!SYMBOLP (tem))
3875 /* Creating a non-pure string from a string literal not implemented yet.
3876 We could just use make_string here and live with the extra copy. */
3877 eassert (!NILP (Vpurify_flag));
3878 tem = intern_driver (make_pure_c_string (str, len), obarray, tem);
3880 return tem;
3883 static void
3884 define_symbol (Lisp_Object sym, char const *str)
3886 ptrdiff_t len = strlen (str);
3887 Lisp_Object string = make_pure_c_string (str, len);
3888 init_symbol (sym, string);
3890 /* Qunbound is uninterned, so that it's not confused with any symbol
3891 'unbound' created by a Lisp program. */
3892 if (! EQ (sym, Qunbound))
3894 Lisp_Object bucket = oblookup (initial_obarray, str, len, len);
3895 eassert (INTEGERP (bucket));
3896 intern_sym (sym, initial_obarray, bucket);
3900 DEFUN ("intern", Fintern, Sintern, 1, 2, 0,
3901 doc: /* Return the canonical symbol whose name is STRING.
3902 If there is none, one is created by this function and returned.
3903 A second optional argument specifies the obarray to use;
3904 it defaults to the value of `obarray'. */)
3905 (Lisp_Object string, Lisp_Object obarray)
3907 Lisp_Object tem;
3909 obarray = check_obarray (NILP (obarray) ? Vobarray : obarray);
3910 CHECK_STRING (string);
3912 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3913 if (!SYMBOLP (tem))
3914 tem = intern_driver (NILP (Vpurify_flag) ? string : Fpurecopy (string),
3915 obarray, tem);
3916 return tem;
3919 DEFUN ("intern-soft", Fintern_soft, Sintern_soft, 1, 2, 0,
3920 doc: /* Return the canonical symbol named NAME, or nil if none exists.
3921 NAME may be a string or a symbol. If it is a symbol, that exact
3922 symbol is searched for.
3923 A second optional argument specifies the obarray to use;
3924 it defaults to the value of `obarray'. */)
3925 (Lisp_Object name, Lisp_Object obarray)
3927 register Lisp_Object tem, string;
3929 if (NILP (obarray)) obarray = Vobarray;
3930 obarray = check_obarray (obarray);
3932 if (!SYMBOLP (name))
3934 CHECK_STRING (name);
3935 string = name;
3937 else
3938 string = SYMBOL_NAME (name);
3940 tem = oblookup (obarray, SSDATA (string), SCHARS (string), SBYTES (string));
3941 if (INTEGERP (tem) || (SYMBOLP (name) && !EQ (name, tem)))
3942 return Qnil;
3943 else
3944 return tem;
3947 DEFUN ("unintern", Funintern, Sunintern, 1, 2, 0,
3948 doc: /* Delete the symbol named NAME, if any, from OBARRAY.
3949 The value is t if a symbol was found and deleted, nil otherwise.
3950 NAME may be a string or a symbol. If it is a symbol, that symbol
3951 is deleted, if it belongs to OBARRAY--no other symbol is deleted.
3952 OBARRAY, if nil, defaults to the value of the variable `obarray'.
3953 usage: (unintern NAME OBARRAY) */)
3954 (Lisp_Object name, Lisp_Object obarray)
3956 register Lisp_Object string, tem;
3957 size_t hash;
3959 if (NILP (obarray)) obarray = Vobarray;
3960 obarray = check_obarray (obarray);
3962 if (SYMBOLP (name))
3963 string = SYMBOL_NAME (name);
3964 else
3966 CHECK_STRING (name);
3967 string = name;
3970 tem = oblookup (obarray, SSDATA (string),
3971 SCHARS (string),
3972 SBYTES (string));
3973 if (INTEGERP (tem))
3974 return Qnil;
3975 /* If arg was a symbol, don't delete anything but that symbol itself. */
3976 if (SYMBOLP (name) && !EQ (name, tem))
3977 return Qnil;
3979 /* There are plenty of other symbols which will screw up the Emacs
3980 session if we unintern them, as well as even more ways to use
3981 `setq' or `fset' or whatnot to make the Emacs session
3982 unusable. Let's not go down this silly road. --Stef */
3983 /* if (EQ (tem, Qnil) || EQ (tem, Qt))
3984 error ("Attempt to unintern t or nil"); */
3986 XSYMBOL (tem)->interned = SYMBOL_UNINTERNED;
3988 hash = oblookup_last_bucket_number;
3990 if (EQ (AREF (obarray, hash), tem))
3992 if (XSYMBOL (tem)->next)
3994 Lisp_Object sym;
3995 XSETSYMBOL (sym, XSYMBOL (tem)->next);
3996 ASET (obarray, hash, sym);
3998 else
3999 ASET (obarray, hash, make_number (0));
4001 else
4003 Lisp_Object tail, following;
4005 for (tail = AREF (obarray, hash);
4006 XSYMBOL (tail)->next;
4007 tail = following)
4009 XSETSYMBOL (following, XSYMBOL (tail)->next);
4010 if (EQ (following, tem))
4012 set_symbol_next (tail, XSYMBOL (following)->next);
4013 break;
4018 return Qt;
4021 /* Return the symbol in OBARRAY whose names matches the string
4022 of SIZE characters (SIZE_BYTE bytes) at PTR.
4023 If there is no such symbol, return the integer bucket number of
4024 where the symbol would be if it were present.
4026 Also store the bucket number in oblookup_last_bucket_number. */
4028 Lisp_Object
4029 oblookup (Lisp_Object obarray, register const char *ptr, ptrdiff_t size, ptrdiff_t size_byte)
4031 size_t hash;
4032 size_t obsize;
4033 register Lisp_Object tail;
4034 Lisp_Object bucket, tem;
4036 obarray = check_obarray (obarray);
4037 /* This is sometimes needed in the middle of GC. */
4038 obsize = gc_asize (obarray);
4039 hash = hash_string (ptr, size_byte) % obsize;
4040 bucket = AREF (obarray, hash);
4041 oblookup_last_bucket_number = hash;
4042 if (EQ (bucket, make_number (0)))
4044 else if (!SYMBOLP (bucket))
4045 error ("Bad data in guts of obarray"); /* Like CADR error message. */
4046 else
4047 for (tail = bucket; ; XSETSYMBOL (tail, XSYMBOL (tail)->next))
4049 if (SBYTES (SYMBOL_NAME (tail)) == size_byte
4050 && SCHARS (SYMBOL_NAME (tail)) == size
4051 && !memcmp (SDATA (SYMBOL_NAME (tail)), ptr, size_byte))
4052 return tail;
4053 else if (XSYMBOL (tail)->next == 0)
4054 break;
4056 XSETINT (tem, hash);
4057 return tem;
4060 void
4061 map_obarray (Lisp_Object obarray, void (*fn) (Lisp_Object, Lisp_Object), Lisp_Object arg)
4063 ptrdiff_t i;
4064 register Lisp_Object tail;
4065 CHECK_VECTOR (obarray);
4066 for (i = ASIZE (obarray) - 1; i >= 0; i--)
4068 tail = AREF (obarray, i);
4069 if (SYMBOLP (tail))
4070 while (1)
4072 (*fn) (tail, arg);
4073 if (XSYMBOL (tail)->next == 0)
4074 break;
4075 XSETSYMBOL (tail, XSYMBOL (tail)->next);
4080 static void
4081 mapatoms_1 (Lisp_Object sym, Lisp_Object function)
4083 call1 (function, sym);
4086 DEFUN ("mapatoms", Fmapatoms, Smapatoms, 1, 2, 0,
4087 doc: /* Call FUNCTION on every symbol in OBARRAY.
4088 OBARRAY defaults to the value of `obarray'. */)
4089 (Lisp_Object function, Lisp_Object obarray)
4091 if (NILP (obarray)) obarray = Vobarray;
4092 obarray = check_obarray (obarray);
4094 map_obarray (obarray, mapatoms_1, function);
4095 return Qnil;
4098 #define OBARRAY_SIZE 1511
4100 void
4101 init_obarray (void)
4103 Lisp_Object oblength;
4104 ptrdiff_t size = 100 + MAX_MULTIBYTE_LENGTH;
4106 XSETFASTINT (oblength, OBARRAY_SIZE);
4108 Vobarray = Fmake_vector (oblength, make_number (0));
4109 initial_obarray = Vobarray;
4110 staticpro (&initial_obarray);
4112 for (int i = 0; i < ARRAYELTS (lispsym); i++)
4113 define_symbol (builtin_lisp_symbol (i), defsym_name[i]);
4115 DEFSYM (Qunbound, "unbound");
4117 DEFSYM (Qnil, "nil");
4118 SET_SYMBOL_VAL (XSYMBOL (Qnil), Qnil);
4119 XSYMBOL (Qnil)->constant = 1;
4120 XSYMBOL (Qnil)->declared_special = true;
4122 DEFSYM (Qt, "t");
4123 SET_SYMBOL_VAL (XSYMBOL (Qt), Qt);
4124 XSYMBOL (Qt)->constant = 1;
4125 XSYMBOL (Qt)->declared_special = true;
4127 /* Qt is correct even if CANNOT_DUMP. loadup.el will set to nil at end. */
4128 Vpurify_flag = Qt;
4130 DEFSYM (Qvariable_documentation, "variable-documentation");
4132 read_buffer = xmalloc (size);
4133 read_buffer_size = size;
4136 void
4137 defsubr (struct Lisp_Subr *sname)
4139 Lisp_Object sym, tem;
4140 sym = intern_c_string (sname->symbol_name);
4141 XSETPVECTYPE (sname, PVEC_SUBR);
4142 XSETSUBR (tem, sname);
4143 set_symbol_function (sym, tem);
4146 #ifdef NOTDEF /* Use fset in subr.el now! */
4147 void
4148 defalias (struct Lisp_Subr *sname, char *string)
4150 Lisp_Object sym;
4151 sym = intern (string);
4152 XSETSUBR (XSYMBOL (sym)->function, sname);
4154 #endif /* NOTDEF */
4156 /* Define an "integer variable"; a symbol whose value is forwarded to a
4157 C variable of type EMACS_INT. Sample call (with "xx" to fool make-docfile):
4158 DEFxxVAR_INT ("emacs-priority", &emacs_priority, "Documentation"); */
4159 void
4160 defvar_int (struct Lisp_Intfwd *i_fwd,
4161 const char *namestring, EMACS_INT *address)
4163 Lisp_Object sym;
4164 sym = intern_c_string (namestring);
4165 i_fwd->type = Lisp_Fwd_Int;
4166 i_fwd->intvar = address;
4167 XSYMBOL (sym)->declared_special = 1;
4168 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4169 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)i_fwd);
4172 /* Similar but define a variable whose value is t if address contains 1,
4173 nil if address contains 0. */
4174 void
4175 defvar_bool (struct Lisp_Boolfwd *b_fwd,
4176 const char *namestring, bool *address)
4178 Lisp_Object sym;
4179 sym = intern_c_string (namestring);
4180 b_fwd->type = Lisp_Fwd_Bool;
4181 b_fwd->boolvar = address;
4182 XSYMBOL (sym)->declared_special = 1;
4183 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4184 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)b_fwd);
4185 Vbyte_boolean_vars = Fcons (sym, Vbyte_boolean_vars);
4188 /* Similar but define a variable whose value is the Lisp Object stored
4189 at address. Two versions: with and without gc-marking of the C
4190 variable. The nopro version is used when that variable will be
4191 gc-marked for some other reason, since marking the same slot twice
4192 can cause trouble with strings. */
4193 void
4194 defvar_lisp_nopro (struct Lisp_Objfwd *o_fwd,
4195 const char *namestring, Lisp_Object *address)
4197 Lisp_Object sym;
4198 sym = intern_c_string (namestring);
4199 o_fwd->type = Lisp_Fwd_Obj;
4200 o_fwd->objvar = address;
4201 XSYMBOL (sym)->declared_special = 1;
4202 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4203 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)o_fwd);
4206 void
4207 defvar_lisp (struct Lisp_Objfwd *o_fwd,
4208 const char *namestring, Lisp_Object *address)
4210 defvar_lisp_nopro (o_fwd, namestring, address);
4211 staticpro (address);
4214 /* Similar but define a variable whose value is the Lisp Object stored
4215 at a particular offset in the current kboard object. */
4217 void
4218 defvar_kboard (struct Lisp_Kboard_Objfwd *ko_fwd,
4219 const char *namestring, int offset)
4221 Lisp_Object sym;
4222 sym = intern_c_string (namestring);
4223 ko_fwd->type = Lisp_Fwd_Kboard_Obj;
4224 ko_fwd->offset = offset;
4225 XSYMBOL (sym)->declared_special = 1;
4226 XSYMBOL (sym)->redirect = SYMBOL_FORWARDED;
4227 SET_SYMBOL_FWD (XSYMBOL (sym), (union Lisp_Fwd *)ko_fwd);
4230 /* Check that the elements of lpath exist. */
4232 static void
4233 load_path_check (Lisp_Object lpath)
4235 Lisp_Object path_tail;
4237 /* The only elements that might not exist are those from
4238 PATH_LOADSEARCH, EMACSLOADPATH. Anything else is only added if
4239 it exists. */
4240 for (path_tail = lpath; !NILP (path_tail); path_tail = XCDR (path_tail))
4242 Lisp_Object dirfile;
4243 dirfile = Fcar (path_tail);
4244 if (STRINGP (dirfile))
4246 dirfile = Fdirectory_file_name (dirfile);
4247 if (! file_accessible_directory_p (dirfile))
4248 dir_warning ("Lisp directory", XCAR (path_tail));
4253 /* Return the default load-path, to be used if EMACSLOADPATH is unset.
4254 This does not include the standard site-lisp directories
4255 under the installation prefix (i.e., PATH_SITELOADSEARCH),
4256 but it does (unless no_site_lisp is set) include site-lisp
4257 directories in the source/build directories if those exist and we
4258 are running uninstalled.
4260 Uses the following logic:
4261 If CANNOT_DUMP: Use PATH_LOADSEARCH.
4262 The remainder is what happens when dumping works:
4263 If purify-flag (ie dumping) just use PATH_DUMPLOADSEARCH.
4264 Otherwise use PATH_LOADSEARCH.
4266 If !initialized, then just return PATH_DUMPLOADSEARCH.
4267 If initialized:
4268 If Vinstallation_directory is not nil (ie, running uninstalled):
4269 If installation-dir/lisp exists and not already a member,
4270 we must be running uninstalled. Reset the load-path
4271 to just installation-dir/lisp. (The default PATH_LOADSEARCH
4272 refers to the eventual installation directories. Since we
4273 are not yet installed, we should not use them, even if they exist.)
4274 If installation-dir/lisp does not exist, just add
4275 PATH_DUMPLOADSEARCH at the end instead.
4276 Add installation-dir/site-lisp (if !no_site_lisp, and exists
4277 and not already a member) at the front.
4278 If installation-dir != source-dir (ie running an uninstalled,
4279 out-of-tree build) AND install-dir/src/Makefile exists BUT
4280 install-dir/src/Makefile.in does NOT exist (this is a sanity
4281 check), then repeat the above steps for source-dir/lisp, site-lisp. */
4283 static Lisp_Object
4284 load_path_default (void)
4286 Lisp_Object lpath = Qnil;
4287 const char *normal;
4289 #ifdef CANNOT_DUMP
4290 #ifdef HAVE_NS
4291 const char *loadpath = ns_load_path ();
4292 #endif
4294 normal = PATH_LOADSEARCH;
4295 #ifdef HAVE_NS
4296 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4297 #else
4298 lpath = decode_env_path (0, normal, 0);
4299 #endif
4301 #else /* !CANNOT_DUMP */
4303 normal = NILP (Vpurify_flag) ? PATH_LOADSEARCH : PATH_DUMPLOADSEARCH;
4305 if (initialized)
4307 #ifdef HAVE_NS
4308 const char *loadpath = ns_load_path ();
4309 lpath = decode_env_path (0, loadpath ? loadpath : normal, 0);
4310 #else
4311 lpath = decode_env_path (0, normal, 0);
4312 #endif
4313 if (!NILP (Vinstallation_directory))
4315 Lisp_Object tem, tem1;
4317 /* Add to the path the lisp subdir of the installation
4318 dir, if it is accessible. Note: in out-of-tree builds,
4319 this directory is empty save for Makefile. */
4320 tem = Fexpand_file_name (build_string ("lisp"),
4321 Vinstallation_directory);
4322 tem1 = Ffile_accessible_directory_p (tem);
4323 if (!NILP (tem1))
4325 if (NILP (Fmember (tem, lpath)))
4327 /* We are running uninstalled. The default load-path
4328 points to the eventual installed lisp directories.
4329 We should not use those now, even if they exist,
4330 so start over from a clean slate. */
4331 lpath = list1 (tem);
4334 else
4335 /* That dir doesn't exist, so add the build-time
4336 Lisp dirs instead. */
4338 Lisp_Object dump_path =
4339 decode_env_path (0, PATH_DUMPLOADSEARCH, 0);
4340 lpath = nconc2 (lpath, dump_path);
4343 /* Add site-lisp under the installation dir, if it exists. */
4344 if (!no_site_lisp)
4346 tem = Fexpand_file_name (build_string ("site-lisp"),
4347 Vinstallation_directory);
4348 tem1 = Ffile_accessible_directory_p (tem);
4349 if (!NILP (tem1))
4351 if (NILP (Fmember (tem, lpath)))
4352 lpath = Fcons (tem, lpath);
4356 /* If Emacs was not built in the source directory,
4357 and it is run from where it was built, add to load-path
4358 the lisp and site-lisp dirs under that directory. */
4360 if (NILP (Fequal (Vinstallation_directory, Vsource_directory)))
4362 Lisp_Object tem2;
4364 tem = Fexpand_file_name (build_string ("src/Makefile"),
4365 Vinstallation_directory);
4366 tem1 = Ffile_exists_p (tem);
4368 /* Don't be fooled if they moved the entire source tree
4369 AFTER dumping Emacs. If the build directory is indeed
4370 different from the source dir, src/Makefile.in and
4371 src/Makefile will not be found together. */
4372 tem = Fexpand_file_name (build_string ("src/Makefile.in"),
4373 Vinstallation_directory);
4374 tem2 = Ffile_exists_p (tem);
4375 if (!NILP (tem1) && NILP (tem2))
4377 tem = Fexpand_file_name (build_string ("lisp"),
4378 Vsource_directory);
4380 if (NILP (Fmember (tem, lpath)))
4381 lpath = Fcons (tem, lpath);
4383 if (!no_site_lisp)
4385 tem = Fexpand_file_name (build_string ("site-lisp"),
4386 Vsource_directory);
4387 tem1 = Ffile_accessible_directory_p (tem);
4388 if (!NILP (tem1))
4390 if (NILP (Fmember (tem, lpath)))
4391 lpath = Fcons (tem, lpath);
4395 } /* Vinstallation_directory != Vsource_directory */
4397 } /* if Vinstallation_directory */
4399 else /* !initialized */
4401 /* NORMAL refers to PATH_DUMPLOADSEARCH, ie the lisp dir in the
4402 source directory. We used to add ../lisp (ie the lisp dir in
4403 the build directory) at the front here, but that should not
4404 be necessary, since in out of tree builds lisp/ is empty, save
4405 for Makefile. */
4406 lpath = decode_env_path (0, normal, 0);
4408 #endif /* !CANNOT_DUMP */
4410 return lpath;
4413 void
4414 init_lread (void)
4416 /* First, set Vload_path. */
4418 /* Ignore EMACSLOADPATH when dumping. */
4419 #ifdef CANNOT_DUMP
4420 bool use_loadpath = true;
4421 #else
4422 bool use_loadpath = NILP (Vpurify_flag);
4423 #endif
4425 if (use_loadpath && egetenv ("EMACSLOADPATH"))
4427 Vload_path = decode_env_path ("EMACSLOADPATH", 0, 1);
4429 /* Check (non-nil) user-supplied elements. */
4430 load_path_check (Vload_path);
4432 /* If no nils in the environment variable, use as-is.
4433 Otherwise, replace any nils with the default. */
4434 if (! NILP (Fmemq (Qnil, Vload_path)))
4436 Lisp_Object elem, elpath = Vload_path;
4437 Lisp_Object default_lpath = load_path_default ();
4439 /* Check defaults, before adding site-lisp. */
4440 load_path_check (default_lpath);
4442 /* Add the site-lisp directories to the front of the default. */
4443 if (!no_site_lisp && PATH_SITELOADSEARCH[0] != '\0')
4445 Lisp_Object sitelisp;
4446 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4447 if (! NILP (sitelisp))
4448 default_lpath = nconc2 (sitelisp, default_lpath);
4451 Vload_path = Qnil;
4453 /* Replace nils from EMACSLOADPATH by default. */
4454 while (CONSP (elpath))
4456 elem = XCAR (elpath);
4457 elpath = XCDR (elpath);
4458 Vload_path = CALLN (Fappend, Vload_path,
4459 NILP (elem) ? default_lpath : list1 (elem));
4461 } /* Fmemq (Qnil, Vload_path) */
4463 else
4465 Vload_path = load_path_default ();
4467 /* Check before adding site-lisp directories.
4468 The install should have created them, but they are not
4469 required, so no need to warn if they are absent.
4470 Or we might be running before installation. */
4471 load_path_check (Vload_path);
4473 /* Add the site-lisp directories at the front. */
4474 if (initialized && !no_site_lisp && PATH_SITELOADSEARCH[0] != '\0')
4476 Lisp_Object sitelisp;
4477 sitelisp = decode_env_path (0, PATH_SITELOADSEARCH, 0);
4478 if (! NILP (sitelisp)) Vload_path = nconc2 (sitelisp, Vload_path);
4482 Vvalues = Qnil;
4484 load_in_progress = 0;
4485 Vload_file_name = Qnil;
4486 Vstandard_input = Qt;
4487 Vloads_in_progress = Qnil;
4490 /* Print a warning that directory intended for use USE and with name
4491 DIRNAME cannot be accessed. On entry, errno should correspond to
4492 the access failure. Print the warning on stderr and put it in
4493 *Messages*. */
4495 void
4496 dir_warning (char const *use, Lisp_Object dirname)
4498 static char const format[] = "Warning: %s '%s': %s\n";
4499 int access_errno = errno;
4500 fprintf (stderr, format, use, SSDATA (ENCODE_SYSTEM (dirname)),
4501 strerror (access_errno));
4503 /* Don't log the warning before we've initialized!! */
4504 if (initialized)
4506 char const *diagnostic = emacs_strerror (access_errno);
4507 USE_SAFE_ALLOCA;
4508 char *buffer = SAFE_ALLOCA (sizeof format - 3 * (sizeof "%s" - 1)
4509 + strlen (use) + SBYTES (dirname)
4510 + strlen (diagnostic));
4511 ptrdiff_t message_len = esprintf (buffer, format, use, SSDATA (dirname),
4512 diagnostic);
4513 message_dolog (buffer, message_len, 0, STRING_MULTIBYTE (dirname));
4514 SAFE_FREE ();
4518 void
4519 syms_of_lread (void)
4521 defsubr (&Sread);
4522 defsubr (&Sread_from_string);
4523 defsubr (&Sintern);
4524 defsubr (&Sintern_soft);
4525 defsubr (&Sunintern);
4526 defsubr (&Sget_load_suffixes);
4527 defsubr (&Sload);
4528 defsubr (&Seval_buffer);
4529 defsubr (&Seval_region);
4530 defsubr (&Sread_char);
4531 defsubr (&Sread_char_exclusive);
4532 defsubr (&Sread_event);
4533 defsubr (&Sget_file_char);
4534 defsubr (&Smapatoms);
4535 defsubr (&Slocate_file_internal);
4537 DEFVAR_LISP ("obarray", Vobarray,
4538 doc: /* Symbol table for use by `intern' and `read'.
4539 It is a vector whose length ought to be prime for best results.
4540 The vector's contents don't make sense if examined from Lisp programs;
4541 to find all the symbols in an obarray, use `mapatoms'. */);
4543 DEFVAR_LISP ("values", Vvalues,
4544 doc: /* List of values of all expressions which were read, evaluated and printed.
4545 Order is reverse chronological. */);
4546 XSYMBOL (intern ("values"))->declared_special = 0;
4548 DEFVAR_LISP ("standard-input", Vstandard_input,
4549 doc: /* Stream for read to get input from.
4550 See documentation of `read' for possible values. */);
4551 Vstandard_input = Qt;
4553 DEFVAR_LISP ("read-with-symbol-positions", Vread_with_symbol_positions,
4554 doc: /* If non-nil, add position of read symbols to `read-symbol-positions-list'.
4556 If this variable is a buffer, then only forms read from that buffer
4557 will be added to `read-symbol-positions-list'.
4558 If this variable is t, then all read forms will be added.
4559 The effect of all other values other than nil are not currently
4560 defined, although they may be in the future.
4562 The positions are relative to the last call to `read' or
4563 `read-from-string'. It is probably a bad idea to set this variable at
4564 the toplevel; bind it instead. */);
4565 Vread_with_symbol_positions = Qnil;
4567 DEFVAR_LISP ("read-symbol-positions-list", Vread_symbol_positions_list,
4568 doc: /* A list mapping read symbols to their positions.
4569 This variable is modified during calls to `read' or
4570 `read-from-string', but only when `read-with-symbol-positions' is
4571 non-nil.
4573 Each element of the list looks like (SYMBOL . CHAR-POSITION), where
4574 CHAR-POSITION is an integer giving the offset of that occurrence of the
4575 symbol from the position where `read' or `read-from-string' started.
4577 Note that a symbol will appear multiple times in this list, if it was
4578 read multiple times. The list is in the same order as the symbols
4579 were read in. */);
4580 Vread_symbol_positions_list = Qnil;
4582 DEFVAR_LISP ("read-circle", Vread_circle,
4583 doc: /* Non-nil means read recursive structures using #N= and #N# syntax. */);
4584 Vread_circle = Qt;
4586 DEFVAR_LISP ("load-path", Vload_path,
4587 doc: /* List of directories to search for files to load.
4588 Each element is a string (directory file name) or nil (meaning
4589 `default-directory').
4590 Initialized during startup as described in Info node `(elisp)Library Search'.
4591 Use `directory-file-name' when adding items to this path. However, Lisp
4592 programs that process this list should tolerate directories both with
4593 and without trailing slashes. */);
4595 DEFVAR_LISP ("load-suffixes", Vload_suffixes,
4596 doc: /* List of suffixes for Emacs Lisp files and dynamic modules.
4597 This list includes suffixes for both compiled and source Emacs Lisp files.
4598 This list should not include the empty string.
4599 `load' and related functions try to append these suffixes, in order,
4600 to the specified file name if a suffix is allowed or required. */);
4601 #ifdef HAVE_MODULES
4602 Vload_suffixes = list3 (build_pure_c_string (".elc"),
4603 build_pure_c_string (".el"),
4604 build_pure_c_string (MODULES_SUFFIX));
4605 #else
4606 Vload_suffixes = list2 (build_pure_c_string (".elc"),
4607 build_pure_c_string (".el"));
4608 #endif
4609 DEFVAR_LISP ("module-file-suffix", Vmodule_file_suffix,
4610 doc: /* Suffix of loadable module file, or nil of modules are not supported. */);
4611 #ifdef HAVE_MODULES
4612 Vmodule_file_suffix = build_pure_c_string (MODULES_SUFFIX);
4613 #else
4614 Vmodule_file_suffix = Qnil;
4615 #endif
4616 DEFVAR_LISP ("load-file-rep-suffixes", Vload_file_rep_suffixes,
4617 doc: /* List of suffixes that indicate representations of \
4618 the same file.
4619 This list should normally start with the empty string.
4621 Enabling Auto Compression mode appends the suffixes in
4622 `jka-compr-load-suffixes' to this list and disabling Auto Compression
4623 mode removes them again. `load' and related functions use this list to
4624 determine whether they should look for compressed versions of a file
4625 and, if so, which suffixes they should try to append to the file name
4626 in order to do so. However, if you want to customize which suffixes
4627 the loading functions recognize as compression suffixes, you should
4628 customize `jka-compr-load-suffixes' rather than the present variable. */);
4629 Vload_file_rep_suffixes = list1 (empty_unibyte_string);
4631 DEFVAR_BOOL ("load-in-progress", load_in_progress,
4632 doc: /* Non-nil if inside of `load'. */);
4633 DEFSYM (Qload_in_progress, "load-in-progress");
4635 DEFVAR_LISP ("after-load-alist", Vafter_load_alist,
4636 doc: /* An alist of functions to be evalled when particular files are loaded.
4637 Each element looks like (REGEXP-OR-FEATURE FUNCS...).
4639 REGEXP-OR-FEATURE is either a regular expression to match file names, or
4640 a symbol (a feature name).
4642 When `load' is run and the file-name argument matches an element's
4643 REGEXP-OR-FEATURE, or when `provide' is run and provides the symbol
4644 REGEXP-OR-FEATURE, the FUNCS in the element are called.
4646 An error in FORMS does not undo the load, but does prevent execution of
4647 the rest of the FORMS. */);
4648 Vafter_load_alist = Qnil;
4650 DEFVAR_LISP ("load-history", Vload_history,
4651 doc: /* Alist mapping loaded file names to symbols and features.
4652 Each alist element should be a list (FILE-NAME ENTRIES...), where
4653 FILE-NAME is the name of a file that has been loaded into Emacs.
4654 The file name is absolute and true (i.e. it doesn't contain symlinks).
4655 As an exception, one of the alist elements may have FILE-NAME nil,
4656 for symbols and features not associated with any file.
4658 The remaining ENTRIES in the alist element describe the functions and
4659 variables defined in that file, the features provided, and the
4660 features required. Each entry has the form `(provide . FEATURE)',
4661 `(require . FEATURE)', `(defun . FUNCTION)', `(autoload . SYMBOL)',
4662 `(defface . SYMBOL)', or `(t . SYMBOL)'. Entries like `(t . SYMBOL)'
4663 may precede a `(defun . FUNCTION)' entry, and means that SYMBOL was an
4664 autoload before this file redefined it as a function. In addition,
4665 entries may also be single symbols, which means that SYMBOL was
4666 defined by `defvar' or `defconst'.
4668 During preloading, the file name recorded is relative to the main Lisp
4669 directory. These file names are converted to absolute at startup. */);
4670 Vload_history = Qnil;
4672 DEFVAR_LISP ("load-file-name", Vload_file_name,
4673 doc: /* Full name of file being loaded by `load'. */);
4674 Vload_file_name = Qnil;
4676 DEFVAR_LISP ("user-init-file", Vuser_init_file,
4677 doc: /* File name, including directory, of user's initialization file.
4678 If the file loaded had extension `.elc', and the corresponding source file
4679 exists, this variable contains the name of source file, suitable for use
4680 by functions like `custom-save-all' which edit the init file.
4681 While Emacs loads and evaluates the init file, value is the real name
4682 of the file, regardless of whether or not it has the `.elc' extension. */);
4683 Vuser_init_file = Qnil;
4685 DEFVAR_LISP ("current-load-list", Vcurrent_load_list,
4686 doc: /* Used for internal purposes by `load'. */);
4687 Vcurrent_load_list = Qnil;
4689 DEFVAR_LISP ("load-read-function", Vload_read_function,
4690 doc: /* Function used by `load' and `eval-region' for reading expressions.
4691 Called with a single argument (the stream from which to read).
4692 The default is to use the function `read'. */);
4693 DEFSYM (Qread, "read");
4694 Vload_read_function = Qread;
4696 DEFVAR_LISP ("load-source-file-function", Vload_source_file_function,
4697 doc: /* Function called in `load' to load an Emacs Lisp source file.
4698 The value should be a function for doing code conversion before
4699 reading a source file. It can also be nil, in which case loading is
4700 done without any code conversion.
4702 If the value is a function, it is called with four arguments,
4703 FULLNAME, FILE, NOERROR, NOMESSAGE. FULLNAME is the absolute name of
4704 the file to load, FILE is the non-absolute name (for messages etc.),
4705 and NOERROR and NOMESSAGE are the corresponding arguments passed to
4706 `load'. The function should return t if the file was loaded. */);
4707 Vload_source_file_function = Qnil;
4709 DEFVAR_BOOL ("load-force-doc-strings", load_force_doc_strings,
4710 doc: /* Non-nil means `load' should force-load all dynamic doc strings.
4711 This is useful when the file being loaded is a temporary copy. */);
4712 load_force_doc_strings = 0;
4714 DEFVAR_BOOL ("load-convert-to-unibyte", load_convert_to_unibyte,
4715 doc: /* Non-nil means `read' converts strings to unibyte whenever possible.
4716 This is normally bound by `load' and `eval-buffer' to control `read',
4717 and is not meant for users to change. */);
4718 load_convert_to_unibyte = 0;
4720 DEFVAR_LISP ("source-directory", Vsource_directory,
4721 doc: /* Directory in which Emacs sources were found when Emacs was built.
4722 You cannot count on them to still be there! */);
4723 Vsource_directory
4724 = Fexpand_file_name (build_string ("../"),
4725 Fcar (decode_env_path (0, PATH_DUMPLOADSEARCH, 0)));
4727 DEFVAR_LISP ("preloaded-file-list", Vpreloaded_file_list,
4728 doc: /* List of files that were preloaded (when dumping Emacs). */);
4729 Vpreloaded_file_list = Qnil;
4731 DEFVAR_LISP ("byte-boolean-vars", Vbyte_boolean_vars,
4732 doc: /* List of all DEFVAR_BOOL variables, used by the byte code optimizer. */);
4733 Vbyte_boolean_vars = Qnil;
4735 DEFVAR_BOOL ("load-dangerous-libraries", load_dangerous_libraries,
4736 doc: /* Non-nil means load dangerous compiled Lisp files.
4737 Some versions of XEmacs use different byte codes than Emacs. These
4738 incompatible byte codes can make Emacs crash when it tries to execute
4739 them. */);
4740 load_dangerous_libraries = 0;
4742 DEFVAR_BOOL ("force-load-messages", force_load_messages,
4743 doc: /* Non-nil means force printing messages when loading Lisp files.
4744 This overrides the value of the NOMESSAGE argument to `load'. */);
4745 force_load_messages = 0;
4747 DEFVAR_LISP ("bytecomp-version-regexp", Vbytecomp_version_regexp,
4748 doc: /* Regular expression matching safe to load compiled Lisp files.
4749 When Emacs loads a compiled Lisp file, it reads the first 512 bytes
4750 from the file, and matches them against this regular expression.
4751 When the regular expression matches, the file is considered to be safe
4752 to load. See also `load-dangerous-libraries'. */);
4753 Vbytecomp_version_regexp
4754 = build_pure_c_string ("^;;;.\\(in Emacs version\\|bytecomp version FSF\\)");
4756 DEFSYM (Qlexical_binding, "lexical-binding");
4757 DEFVAR_LISP ("lexical-binding", Vlexical_binding,
4758 doc: /* Whether to use lexical binding when evaluating code.
4759 Non-nil means that the code in the current buffer should be evaluated
4760 with lexical binding.
4761 This variable is automatically set from the file variables of an
4762 interpreted Lisp file read using `load'. Unlike other file local
4763 variables, this must be set in the first line of a file. */);
4764 Vlexical_binding = Qnil;
4765 Fmake_variable_buffer_local (Qlexical_binding);
4767 DEFVAR_LISP ("eval-buffer-list", Veval_buffer_list,
4768 doc: /* List of buffers being read from by calls to `eval-buffer' and `eval-region'. */);
4769 Veval_buffer_list = Qnil;
4771 DEFVAR_LISP ("old-style-backquotes", Vold_style_backquotes,
4772 doc: /* Set to non-nil when `read' encounters an old-style backquote. */);
4773 Vold_style_backquotes = Qnil;
4774 DEFSYM (Qold_style_backquotes, "old-style-backquotes");
4776 DEFVAR_BOOL ("load-prefer-newer", load_prefer_newer,
4777 doc: /* Non-nil means `load' prefers the newest version of a file.
4778 This applies when a filename suffix is not explicitly specified and
4779 `load' is trying various possible suffixes (see `load-suffixes' and
4780 `load-file-rep-suffixes'). Normally, it stops at the first file
4781 that exists unless you explicitly specify one or the other. If this
4782 option is non-nil, it checks all suffixes and uses whichever file is
4783 newest.
4784 Note that if you customize this, obviously it will not affect files
4785 that are loaded before your customizations are read! */);
4786 load_prefer_newer = 0;
4788 /* Vsource_directory was initialized in init_lread. */
4790 DEFSYM (Qcurrent_load_list, "current-load-list");
4791 DEFSYM (Qstandard_input, "standard-input");
4792 DEFSYM (Qread_char, "read-char");
4793 DEFSYM (Qget_file_char, "get-file-char");
4795 /* Used instead of Qget_file_char while loading *.elc files compiled
4796 by Emacs 21 or older. */
4797 DEFSYM (Qget_emacs_mule_file_char, "get-emacs-mule-file-char");
4799 DEFSYM (Qload_force_doc_strings, "load-force-doc-strings");
4801 DEFSYM (Qbackquote, "`");
4802 DEFSYM (Qcomma, ",");
4803 DEFSYM (Qcomma_at, ",@");
4804 DEFSYM (Qcomma_dot, ",.");
4806 DEFSYM (Qinhibit_file_name_operation, "inhibit-file-name-operation");
4807 DEFSYM (Qascii_character, "ascii-character");
4808 DEFSYM (Qfunction, "function");
4809 DEFSYM (Qload, "load");
4810 DEFSYM (Qload_file_name, "load-file-name");
4811 DEFSYM (Qeval_buffer_list, "eval-buffer-list");
4812 DEFSYM (Qfile_truename, "file-truename");
4813 DEFSYM (Qdir_ok, "dir-ok");
4814 DEFSYM (Qdo_after_load_evaluation, "do-after-load-evaluation");
4816 staticpro (&read_objects);
4817 read_objects = Qnil;
4818 staticpro (&seen_list);
4819 seen_list = Qnil;
4821 Vloads_in_progress = Qnil;
4822 staticpro (&Vloads_in_progress);
4824 DEFSYM (Qhash_table, "hash-table");
4825 DEFSYM (Qdata, "data");
4826 DEFSYM (Qtest, "test");
4827 DEFSYM (Qsize, "size");
4828 DEFSYM (Qweakness, "weakness");
4829 DEFSYM (Qrehash_size, "rehash-size");
4830 DEFSYM (Qrehash_threshold, "rehash-threshold");
4832 DEFSYM (Qucs_names, "ucs-names");