1 /* Buffer insertion/deletion and gap motion for XEmacs.
2 Copyright (C) 1985, 1986, 1991, 1992, 1993, 1994, 1995
3 Free Software Foundation, Inc.
4 Copyright (C) 1995 Sun Microsystems, Inc.
6 This file is part of XEmacs.
8 XEmacs is free software; you can redistribute it and/or modify it
9 under the terms of the GNU General Public License as published by the
10 Free Software Foundation; either version 2, or (at your option) any
13 XEmacs is distributed in the hope that it will be useful, but WITHOUT
14 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
15 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
18 You should have received a copy of the GNU General Public License
19 along with XEmacs; see the file COPYING. If not, write to
20 the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
21 Boston, MA 02111-1307, USA. */
23 /* Synched up with: Mule 2.0, FSF 19.30. Diverges significantly. */
25 /* This file has been Mule-ized. */
27 /* Overhauled by Ben Wing, December 1994, for Mule implementation. */
30 There are three possible ways to specify positions in a buffer. All
31 of these are one-based: the beginning of the buffer is position or
32 index 1, and 0 is not a valid position.
34 As a "buffer position" (typedef Bufpos):
36 This is an index specifying an offset in characters from the
37 beginning of the buffer. Note that buffer positions are
38 logically *between* characters, not on a character. The
39 difference between two buffer positions specifies the number of
40 characters between those positions. Buffer positions are the
41 only kind of position externally visible to the user.
43 As a "byte index" (typedef Bytind):
45 This is an index over the bytes used to represent the characters
46 in the buffer. If there is no Mule support, this is identical
47 to a buffer position, because each character is represented
48 using one byte. However, with Mule support, many characters
49 require two or more bytes for their representation, and so a
50 byte index may be greater than the corresponding buffer
53 As a "memory index" (typedef Memind):
55 This is the byte index adjusted for the gap. For positions
56 before the gap, this is identical to the byte index. For
57 positions after the gap, this is the byte index plus the gap
58 size. There are two possible memory indices for the gap
59 position; the memory index at the beginning of the gap should
60 always be used, except in code that deals with manipulating the
61 gap, where both indices may be seen. The address of the
62 character "at" (i.e. following) a particular position can be
63 obtained from the formula
65 buffer_start_address + memory_index(position) - 1
67 except in the case of characters at the gap position.
74 This typedef represents a single Emacs character, which can be
75 ASCII, ISO-8859, or some extended character, as would typically
76 be used for Kanji. Note that the representation of a character
77 as an Emchar is *not* the same as the representation of that
78 same character in a string; thus, you cannot do the standard
79 C trick of passing a pointer to a character to a function that
82 An Emchar takes up 19 bits of representation and (for code
83 compatibility and such) is compatible with an int. This
84 representation is visible on the Lisp level. The important
85 characteristics of the Emchar representation are
87 -- values 0x00 - 0x7f represent ASCII.
88 -- values 0x80 - 0xff represent the right half of ISO-8859-1.
89 -- values 0x100 and up represent all other characters.
91 This means that Emchar values are upwardly compatible with
92 the standard 8-bit representation of ASCII/ISO-8859-1.
96 The data in a buffer or string is logically made up of Bufbyte
97 objects, where a Bufbyte takes up the same amount of space as a
98 char. (It is declared differently, though, to catch invalid
99 usages.) Strings stored using Bufbytes are said to be in
100 "internal format". The important characteristics of internal
103 -- ASCII characters are represented as a single Bufbyte,
104 in the range 0 - 0x7f.
105 -- All other characters are represented as a Bufbyte in
106 the range 0x80 - 0x9f followed by one or more Bufbytes
107 in the range 0xa0 to 0xff.
109 This leads to a number of desirable properties:
111 -- Given the position of the beginning of a character,
112 you can find the beginning of the next or previous
113 character in constant time.
114 -- When searching for a substring or an ASCII character
115 within the string, you need merely use standard
120 Strings that go in or out of Emacs are in "external format",
121 typedef'ed as an array of char or a char *. There is more
122 than one external format (JIS, EUC, etc.) but they all
123 have similar properties. They are modal encodings,
124 which is to say that the meaning of particular bytes is
125 not fixed but depends on what "mode" the string is currently
126 in (e.g. bytes in the range 0 - 0x7f might be
127 interpreted as ASCII, or as Hiragana, or as 2-byte Kanji,
128 depending on the current mode). The mode starts out in
129 ASCII/ISO-8859-1 and is switched using escape sequences --
130 for example, in the JIS encoding, 'ESC $ B' switches to a
131 mode where pairs of bytes in the range 0 - 0x7f
132 are interpreted as Kanji characters.
134 External-formatted data is generally desirable for passing
135 data between programs because it is upwardly compatible
136 with standard ASCII/ISO-8859-1 strings and may require
137 less space than internal encodings such as the one
138 described above. In addition, some encodings (e.g. JIS)
139 keep all characters (except the ESC used to switch modes)
140 in the printing ASCII range 0x20 - 0x7e, which results in
141 a much higher probability that the data will avoid being
142 garbled in transmission. Externally-formatted data is
143 generally not very convenient to work with, however, and
144 for this reason is usually converted to internal format
145 before any work is done on the string.
147 NOTE: filenames need to be in external format so that
148 ISO-8859-1 characters come out correctly.
152 This typedef represents a count of characters, such as
153 a character offset into a string or the number of
154 characters between two positions in a buffer. The
155 difference between two Bufpos's is a Charcount, and
156 character positions in a string are represented using
161 Similar to a Charcount but represents a count of bytes.
162 The difference between two Bytind's is a Bytecount.
165 Usage of the various representations:
166 =====================================
168 Memory indices are used in low-level functions in insdel.c and for
169 extent endpoints and marker positions. The reason for this is that
170 this way, the extents and markers don't need to be updated for most
171 insertions, which merely shrink the gap and don't move any
172 characters around in memory.
174 (The beginning-of-gap memory index simplifies insertions w.r.t.
175 markers, because text usually gets inserted after markers. For
176 extents, it is merely for consistency, because text can get
177 inserted either before or after an extent's endpoint depending on
178 the open/closedness of the endpoint.)
180 Byte indices are used in other code that needs to be fast,
181 such as the searching, redisplay, and extent-manipulation code.
183 Buffer positions are used in all other code. This is because this
184 representation is easiest to work with (especially since Lisp
185 code always uses buffer positions), necessitates the fewest
186 changes to existing code, and is the safest (e.g. if the text gets
187 shifted underneath a buffer position, it will still point to a
188 character; if text is shifted under a byte index, it might point
189 to the middle of a character, which would be bad).
191 Similarly, Charcounts are used in all code that deals with strings
192 except for code that needs to be fast, which used Bytecounts.
194 Strings are always passed around internally using internal format.
195 Conversions between external format are performed at the time
196 that the data goes in or out of Emacs.
198 Working with the various representations:
199 ========================================= */
210 #include "redisplay.h"
211 #include "line-number.h"
213 /* We write things this way because it's very important the
214 MAX_BYTIND_GAP_SIZE_3 is a multiple of 3. (As it happens,
215 65535 is a multiple of 3, but this may not always be the
218 #define MAX_BUFPOS_GAP_SIZE_3 (65535/3)
219 #define MAX_BYTIND_GAP_SIZE_3 (3 * MAX_BUFPOS_GAP_SIZE_3)
221 short three_to_one_table[1 + MAX_BYTIND_GAP_SIZE_3];
223 /* Various macros modelled along the lines of those in buffer.h.
224 Purposefully omitted from buffer.h because files other than this
225 one should not be using them. */
227 /* Address of beginning of buffer. This is an lvalue because
228 BUFFER_ALLOC needs it to be. */
229 #define BUF_BEG_ADDR(buf) ((buf)->text->beg)
231 /* Set the address of beginning of buffer. */
232 #define SET_BUF_BEG_ADDR(buf, addr) do { (buf)->text->beg = (addr); } while (0)
235 #define BUF_GAP_SIZE(buf) ((buf)->text->gap_size + 0)
236 #define BUF_END_GAP_SIZE(buf) ((buf)->text->end_gap_size + 0)
238 #define SET_BUF_GAP_SIZE(buf, value) \
239 do { (buf)->text->gap_size = (value); } while (0)
240 #define SET_BUF_END_GAP_SIZE(buf, value) \
241 do { (buf)->text->end_gap_size = (value); } while (0)
244 #define BI_BUF_GPT(buf) ((buf)->text->gpt + 0)
245 #define BUF_GPT_ADDR(buf) (BUF_BEG_ADDR (buf) + BI_BUF_GPT (buf) - 1)
247 /* Set gap location. */
248 #define SET_BI_BUF_GPT(buf, value) do { (buf)->text->gpt = (value); } while (0)
250 /* Set end of buffer. */
251 #define SET_BOTH_BUF_Z(buf, val, bival) \
254 (buf)->text->z = (bival); \
255 (buf)->text->bufz = (val); \
258 /* Under Mule, we maintain two sentinels in the buffer: one at the
259 beginning of the gap, and one at the end of the buffer. This
260 allows us to move forward, examining bytes looking for the
261 end of a character, and not worry about running off the end.
262 We do not need corresponding sentinels when moving backwards
263 because we do not have to look past the beginning of a character
264 to find the beginning of the character.
266 Every time we change the beginning of the gap, we have to
267 call SET_GAP_SENTINEL().
269 Every time we change the total size (characters plus gap)
270 of the buffer, we have to call SET_END_SENTINEL().
275 # define GAP_CAN_HOLD_SIZE_P(buf, len) (BUF_GAP_SIZE (buf) >= (len) + 1)
276 # define SET_GAP_SENTINEL(buf) (*BUF_GPT_ADDR (buf) = 0)
277 # define BUF_END_SENTINEL_SIZE 1
278 # define SET_END_SENTINEL(buf) \
279 (*(BUF_BEG_ADDR (buf) + BUF_GAP_SIZE (buf) + BI_BUF_Z (buf) - 1) = 0)
281 # define GAP_CAN_HOLD_SIZE_P(buf, len) (BUF_GAP_SIZE (buf) >= (len))
282 # define SET_GAP_SENTINEL(buf)
283 # define BUF_END_SENTINEL_SIZE 0
284 # define SET_END_SENTINEL(buf)
288 /************************************************************************/
289 /* Charcount/Bytecount conversion */
290 /************************************************************************/
292 /* Optimization. Do it. Live it. Love it. */
296 /* We include the basic functions here that require no specific
297 knowledge of how data is Mule-encoded into a buffer other
298 than the basic (00 - 7F), (80 - 9F), (A0 - FF) scheme.
299 Anything that requires more specific knowledge goes into
302 /* Given a pointer to a text string and a length in bytes, return
303 the equivalent length in characters. */
306 bytecount_to_charcount (const Bufbyte *ptr, Bytecount len)
309 const Bufbyte *end = ptr + len;
312 # define STRIDE_TYPE long
313 # define HIGH_BIT_MASK 0x8080808080808080UL
314 #elif SIZEOF_LONG_LONG == 8 && !(defined (i386) || defined (__i386__))
315 # define STRIDE_TYPE long long
316 # define HIGH_BIT_MASK 0x8080808080808080ULL
317 #elif SIZEOF_LONG == 4
318 # define STRIDE_TYPE long
319 # define HIGH_BIT_MASK 0x80808080UL
321 # error Add support for 128-bit systems here
324 #define ALIGN_BITS ((EMACS_UINT) (ALIGNOF (STRIDE_TYPE) - 1))
325 #define ALIGN_MASK (~ ALIGN_BITS)
326 #define ALIGNED(ptr) ((((EMACS_UINT) ptr) & ALIGN_BITS) == 0)
327 #define STRIDE sizeof (STRIDE_TYPE)
331 if (BYTE_ASCII_P (*ptr))
333 /* optimize for long stretches of ASCII */
338 const unsigned STRIDE_TYPE *ascii_end =
339 (const unsigned STRIDE_TYPE *) ptr;
340 /* This loop screams, because we can typically
341 detect ASCII characters 8 at a time. */
342 while ((const Bufbyte *) ascii_end + STRIDE <= end
343 && !(*ascii_end & HIGH_BIT_MASK))
345 if ((Bufbyte *) ascii_end == ptr)
349 count += (Bufbyte *) ascii_end - ptr;
350 ptr = (Bufbyte *) ascii_end;
356 /* optimize for successive characters from the same charset */
357 Bufbyte leading_byte = *ptr;
358 size_t bytes = REP_BYTES_BY_FIRST_BYTE (leading_byte);
359 while ((ptr < end) && (*ptr == leading_byte))
360 ptr += bytes, count++;
364 #ifdef ERROR_CHECK_BUFPOS
365 /* Bomb out if the specified substring ends in the middle
366 of a character. Note that we might have already gotten
367 a core dump above from an invalid reference, but at least
368 we will get no farther than here. */
375 /* Given a pointer to a text string and a length in characters, return
376 the equivalent length in bytes. */
379 charcount_to_bytecount (const Bufbyte *ptr, Charcount len)
381 const Bufbyte *newptr = ptr;
385 INC_CHARPTR (newptr);
391 /* The next two functions are the actual meat behind the
392 bufpos-to-bytind and bytind-to-bufpos conversions. Currently
393 the method they use is fairly unsophisticated; see buffer.h.
395 Note that bufpos_to_bytind_func() is probably the most-called
396 function in all of XEmacs. Therefore, it must be FAST FAST FAST.
397 This is the reason why so much of the code is duplicated.
399 Similar considerations apply to bytind_to_bufpos_func(), although
400 less so because the function is not called so often.
402 #### At some point this should use a more sophisticated method;
405 static int not_very_random_number;
408 bufpos_to_bytind_func (struct buffer *buf, Bufpos x)
418 int add_to_cache = 0;
420 /* Check for some cached positions, for speed. */
421 if (x == BUF_PT (buf))
422 return BI_BUF_PT (buf);
423 if (x == BUF_ZV (buf))
424 return BI_BUF_ZV (buf);
425 if (x == BUF_BEGV (buf))
426 return BI_BUF_BEGV (buf);
428 bufmin = buf->text->mule_bufmin;
429 bufmax = buf->text->mule_bufmax;
430 bytmin = buf->text->mule_bytmin;
431 bytmax = buf->text->mule_bytmax;
432 size = (1 << buf->text->mule_shifter) + !!buf->text->mule_three_p;
434 /* The basic idea here is that we shift the "known region" up or down
435 until it overlaps the specified position. We do this by moving
436 the upper bound of the known region up one character at a time,
437 and moving the lower bound of the known region up as necessary
438 when the size of the character just seen changes.
440 We optimize this, however, by first shifting the known region to
441 one of the cached points if it's close by. (We don't check BEG or
442 Z, even though they're cached; most of the time these will be the
443 same as BEGV and ZV, and when they're not, they're not likely
448 Bufpos diffmax = x - bufmax;
449 Bufpos diffpt = x - BUF_PT (buf);
450 Bufpos diffzv = BUF_ZV (buf) - x;
451 /* #### This value could stand some more exploration. */
452 Charcount heuristic_hack = (bufmax - bufmin) >> 2;
454 /* Check if the position is closer to PT or ZV than to the
455 end of the known region. */
462 /* But also implement a heuristic that favors the known region
463 over PT or ZV. The reason for this is that switching to
464 PT or ZV will wipe out the knowledge in the known region,
465 which might be annoying if the known region is large and
466 PT or ZV is not that much closer than the end of the known
469 diffzv += heuristic_hack;
470 diffpt += heuristic_hack;
471 if (diffpt < diffmax && diffpt <= diffzv)
473 bufmax = bufmin = BUF_PT (buf);
474 bytmax = bytmin = BI_BUF_PT (buf);
475 /* We set the size to 1 even though it doesn't really
476 matter because the new known region contains no
477 characters. We do this because this is the most
478 likely size of the characters around the new known
479 region, and we avoid potential yuckiness that is
480 done when size == 3. */
483 if (diffzv < diffmax)
485 bufmax = bufmin = BUF_ZV (buf);
486 bytmax = bytmin = BI_BUF_ZV (buf);
490 #ifdef ERROR_CHECK_BUFPOS
491 else if (x >= bufmin)
496 Bufpos diffmin = bufmin - x;
497 Bufpos diffpt = BUF_PT (buf) - x;
498 Bufpos diffbegv = x - BUF_BEGV (buf);
499 /* #### This value could stand some more exploration. */
500 Charcount heuristic_hack = (bufmax - bufmin) >> 2;
505 diffbegv = -diffbegv;
507 /* But also implement a heuristic that favors the known region --
510 diffbegv += heuristic_hack;
511 diffpt += heuristic_hack;
513 if (diffpt < diffmin && diffpt <= diffbegv)
515 bufmax = bufmin = BUF_PT (buf);
516 bytmax = bytmin = BI_BUF_PT (buf);
517 /* We set the size to 1 even though it doesn't really
518 matter because the new known region contains no
519 characters. We do this because this is the most
520 likely size of the characters around the new known
521 region, and we avoid potential yuckiness that is
522 done when size == 3. */
525 if (diffbegv < diffmin)
527 bufmax = bufmin = BUF_BEGV (buf);
528 bytmax = bytmin = BI_BUF_BEGV (buf);
533 diff_so_far = x > bufmax ? x - bufmax : bufmin - x;
534 if (diff_so_far > 50)
536 /* If we have to move more than a certain amount, then look
538 int minval = INT_MAX;
543 /* I considered keeping the positions ordered. This would speed
544 up this loop, but updating the cache would take longer, so
545 it doesn't seem like it would really matter. */
546 for (i = 0; i < 16; i++)
548 int diff = buf->text->mule_bufpos_cache[i] - x;
559 if (minval < diff_so_far)
561 bufmax = bufmin = buf->text->mule_bufpos_cache[found];
562 bytmax = bytmin = buf->text->mule_bytind_cache[found];
567 /* It's conceivable that the caching above could lead to X being
568 the same as one of the range edges. */
579 INC_BYTIND (buf, newmax);
580 newsize = newmax - bytmax;
592 /* #### Should go past the found location to reduce the number
593 of times that this function is called */
595 else /* x < bufmin */
605 DEC_BYTIND (buf, newmin);
606 newsize = bytmin - newmin;
618 /* #### Should go past the found location to reduce the number
619 of times that this function is called
623 /* If size is three, than we have to max sure that the range we
624 discovered isn't too large, because we use a fixed-length
625 table to divide by 3. */
629 int gap = bytmax - bytmin;
630 buf->text->mule_three_p = 1;
631 buf->text->mule_shifter = 1;
633 if (gap > MAX_BYTIND_GAP_SIZE_3)
637 bytmin = bytmax - MAX_BYTIND_GAP_SIZE_3;
638 bufmin = bufmax - MAX_BUFPOS_GAP_SIZE_3;
642 bytmax = bytmin + MAX_BYTIND_GAP_SIZE_3;
643 bufmax = bufmin + MAX_BUFPOS_GAP_SIZE_3;
649 buf->text->mule_three_p = 0;
651 buf->text->mule_shifter = 2;
653 buf->text->mule_shifter = size - 1;
656 buf->text->mule_bufmin = bufmin;
657 buf->text->mule_bufmax = bufmax;
658 buf->text->mule_bytmin = bytmin;
659 buf->text->mule_bytmax = bytmax;
665 /* We throw away a "random" cached value and replace it with
666 the new value. It doesn't actually have to be very random
667 at all, just evenly distributed.
669 #### It would be better to use a least-recently-used algorithm
670 or something that tries to space things out, but I'm not sure
671 it's worth it to go to the trouble of maintaining that. */
672 not_very_random_number += 621;
673 replace_loc = not_very_random_number & 15;
674 buf->text->mule_bufpos_cache[replace_loc] = x;
675 buf->text->mule_bytind_cache[replace_loc] = retval;
681 /* The logic in this function is almost identical to the logic in
682 the previous function. */
685 bytind_to_bufpos_func (struct buffer *buf, Bytind x)
695 int add_to_cache = 0;
697 /* Check for some cached positions, for speed. */
698 if (x == BI_BUF_PT (buf))
700 if (x == BI_BUF_ZV (buf))
702 if (x == BI_BUF_BEGV (buf))
703 return BUF_BEGV (buf);
705 bufmin = buf->text->mule_bufmin;
706 bufmax = buf->text->mule_bufmax;
707 bytmin = buf->text->mule_bytmin;
708 bytmax = buf->text->mule_bytmax;
709 size = (1 << buf->text->mule_shifter) + !!buf->text->mule_three_p;
711 /* The basic idea here is that we shift the "known region" up or down
712 until it overlaps the specified position. We do this by moving
713 the upper bound of the known region up one character at a time,
714 and moving the lower bound of the known region up as necessary
715 when the size of the character just seen changes.
717 We optimize this, however, by first shifting the known region to
718 one of the cached points if it's close by. (We don't check BI_BEG or
719 BI_Z, even though they're cached; most of the time these will be the
720 same as BI_BEGV and BI_ZV, and when they're not, they're not likely
725 Bytind diffmax = x - bytmax;
726 Bytind diffpt = x - BI_BUF_PT (buf);
727 Bytind diffzv = BI_BUF_ZV (buf) - x;
728 /* #### This value could stand some more exploration. */
729 Bytecount heuristic_hack = (bytmax - bytmin) >> 2;
731 /* Check if the position is closer to PT or ZV than to the
732 end of the known region. */
739 /* But also implement a heuristic that favors the known region
740 over BI_PT or BI_ZV. The reason for this is that switching to
741 BI_PT or BI_ZV will wipe out the knowledge in the known region,
742 which might be annoying if the known region is large and
743 BI_PT or BI_ZV is not that much closer than the end of the known
746 diffzv += heuristic_hack;
747 diffpt += heuristic_hack;
748 if (diffpt < diffmax && diffpt <= diffzv)
750 bufmax = bufmin = BUF_PT (buf);
751 bytmax = bytmin = BI_BUF_PT (buf);
752 /* We set the size to 1 even though it doesn't really
753 matter because the new known region contains no
754 characters. We do this because this is the most
755 likely size of the characters around the new known
756 region, and we avoid potential yuckiness that is
757 done when size == 3. */
760 if (diffzv < diffmax)
762 bufmax = bufmin = BUF_ZV (buf);
763 bytmax = bytmin = BI_BUF_ZV (buf);
767 #ifdef ERROR_CHECK_BUFPOS
768 else if (x >= bytmin)
773 Bytind diffmin = bytmin - x;
774 Bytind diffpt = BI_BUF_PT (buf) - x;
775 Bytind diffbegv = x - BI_BUF_BEGV (buf);
776 /* #### This value could stand some more exploration. */
777 Bytecount heuristic_hack = (bytmax - bytmin) >> 2;
782 diffbegv = -diffbegv;
784 /* But also implement a heuristic that favors the known region --
787 diffbegv += heuristic_hack;
788 diffpt += heuristic_hack;
790 if (diffpt < diffmin && diffpt <= diffbegv)
792 bufmax = bufmin = BUF_PT (buf);
793 bytmax = bytmin = BI_BUF_PT (buf);
794 /* We set the size to 1 even though it doesn't really
795 matter because the new known region contains no
796 characters. We do this because this is the most
797 likely size of the characters around the new known
798 region, and we avoid potential yuckiness that is
799 done when size == 3. */
802 if (diffbegv < diffmin)
804 bufmax = bufmin = BUF_BEGV (buf);
805 bytmax = bytmin = BI_BUF_BEGV (buf);
810 diff_so_far = x > bytmax ? x - bytmax : bytmin - x;
811 if (diff_so_far > 50)
813 /* If we have to move more than a certain amount, then look
815 int minval = INT_MAX;
820 /* I considered keeping the positions ordered. This would speed
821 up this loop, but updating the cache would take longer, so
822 it doesn't seem like it would really matter. */
823 for (i = 0; i < 16; i++)
825 int diff = buf->text->mule_bytind_cache[i] - x;
836 if (minval < diff_so_far)
838 bufmax = bufmin = buf->text->mule_bufpos_cache[found];
839 bytmax = bytmin = buf->text->mule_bytind_cache[found];
844 /* It's conceivable that the caching above could lead to X being
845 the same as one of the range edges. */
856 INC_BYTIND (buf, newmax);
857 newsize = newmax - bytmax;
869 /* #### Should go past the found location to reduce the number
870 of times that this function is called */
872 else /* x <= bytmin */
882 DEC_BYTIND (buf, newmin);
883 newsize = bytmin - newmin;
895 /* #### Should go past the found location to reduce the number
896 of times that this function is called
900 /* If size is three, than we have to max sure that the range we
901 discovered isn't too large, because we use a fixed-length
902 table to divide by 3. */
906 int gap = bytmax - bytmin;
907 buf->text->mule_three_p = 1;
908 buf->text->mule_shifter = 1;
910 if (gap > MAX_BYTIND_GAP_SIZE_3)
914 bytmin = bytmax - MAX_BYTIND_GAP_SIZE_3;
915 bufmin = bufmax - MAX_BUFPOS_GAP_SIZE_3;
919 bytmax = bytmin + MAX_BYTIND_GAP_SIZE_3;
920 bufmax = bufmin + MAX_BUFPOS_GAP_SIZE_3;
926 buf->text->mule_three_p = 0;
928 buf->text->mule_shifter = 2;
930 buf->text->mule_shifter = size - 1;
933 buf->text->mule_bufmin = bufmin;
934 buf->text->mule_bufmax = bufmax;
935 buf->text->mule_bytmin = bytmin;
936 buf->text->mule_bytmax = bytmax;
942 /* We throw away a "random" cached value and replace it with
943 the new value. It doesn't actually have to be very random
944 at all, just evenly distributed.
946 #### It would be better to use a least-recently-used algorithm
947 or something that tries to space things out, but I'm not sure
948 it's worth it to go to the trouble of maintaining that. */
949 not_very_random_number += 621;
950 replace_loc = not_very_random_number & 15;
951 buf->text->mule_bufpos_cache[replace_loc] = retval;
952 buf->text->mule_bytind_cache[replace_loc] = x;
958 /* Text of length BYTELENGTH and CHARLENGTH (in different units)
959 was inserted at bufpos START. */
962 buffer_mule_signal_inserted_region (struct buffer *buf, Bufpos start,
963 Bytecount bytelength,
964 Charcount charlength)
966 int size = (1 << buf->text->mule_shifter) + !!buf->text->mule_three_p;
969 /* Adjust the cache of known positions. */
970 for (i = 0; i < 16; i++)
973 if (buf->text->mule_bufpos_cache[i] > start)
975 buf->text->mule_bufpos_cache[i] += charlength;
976 buf->text->mule_bytind_cache[i] += bytelength;
980 if (start >= buf->text->mule_bufmax)
983 /* The insertion is either before the known region, in which case
984 it shoves it forward; or within the known region, in which case
985 it shoves the end forward. (But it may make the known region
986 inconsistent, so we may have to shorten it.) */
988 if (start <= buf->text->mule_bufmin)
990 buf->text->mule_bufmin += charlength;
991 buf->text->mule_bufmax += charlength;
992 buf->text->mule_bytmin += bytelength;
993 buf->text->mule_bytmax += bytelength;
997 Bufpos end = start + charlength;
998 /* the insertion point divides the known region in two.
999 Keep the longer half, at least, and expand into the
1000 inserted chunk as much as possible. */
1002 if (start - buf->text->mule_bufmin > buf->text->mule_bufmax - start)
1004 Bytind bytestart = (buf->text->mule_bytmin
1005 + size * (start - buf->text->mule_bufmin));
1010 bytenew = bytestart;
1011 INC_BYTIND (buf, bytenew);
1012 if (bytenew - bytestart != size)
1015 bytestart = bytenew;
1019 buf->text->mule_bufmax = start;
1020 buf->text->mule_bytmax = bytestart;
1024 buf->text->mule_bufmax += charlength;
1025 buf->text->mule_bytmax += bytelength;
1030 Bytind byteend = (buf->text->mule_bytmin
1031 + size * (start - buf->text->mule_bufmin)
1035 buf->text->mule_bufmax += charlength;
1036 buf->text->mule_bytmax += bytelength;
1041 DEC_BYTIND (buf, bytenew);
1042 if (byteend - bytenew != size)
1049 buf->text->mule_bufmin = end;
1050 buf->text->mule_bytmin = byteend;
1056 /* Text from START to END (equivalent in Bytinds: from BI_START to
1057 BI_END) was deleted. */
1060 buffer_mule_signal_deleted_region (struct buffer *buf, Bufpos start,
1061 Bufpos end, Bytind bi_start,
1066 /* Adjust the cache of known positions. */
1067 for (i = 0; i < 16; i++)
1069 /* After the end; gets shoved backward */
1070 if (buf->text->mule_bufpos_cache[i] > end)
1072 buf->text->mule_bufpos_cache[i] -= end - start;
1073 buf->text->mule_bytind_cache[i] -= bi_end - bi_start;
1075 /* In the range; moves to start of range */
1076 else if (buf->text->mule_bufpos_cache[i] > start)
1078 buf->text->mule_bufpos_cache[i] = start;
1079 buf->text->mule_bytind_cache[i] = bi_start;
1083 /* We don't care about any text after the end of the known region. */
1085 end = min (end, buf->text->mule_bufmax);
1086 bi_end = min (bi_end, buf->text->mule_bytmax);
1090 /* The end of the known region offsets by the total amount of deletion,
1091 since it's all before it. */
1093 buf->text->mule_bufmax -= end - start;
1094 buf->text->mule_bytmax -= bi_end - bi_start;
1096 /* Now we don't care about any text after the start of the known region. */
1098 end = min (end, buf->text->mule_bufmin);
1099 bi_end = min (bi_end, buf->text->mule_bytmin);
1103 buf->text->mule_bufmin -= end - start;
1104 buf->text->mule_bytmin -= bi_end - bi_start;
1109 #ifdef ERROR_CHECK_BUFPOS
1112 bufpos_to_bytind (struct buffer *buf, Bufpos x)
1114 Bytind retval = real_bufpos_to_bytind (buf, x);
1115 ASSERT_VALID_BYTIND_UNSAFE (buf, retval);
1120 bytind_to_bufpos (struct buffer *buf, Bytind x)
1122 ASSERT_VALID_BYTIND_UNSAFE (buf, x);
1123 return real_bytind_to_bufpos (buf, x);
1126 #endif /* ERROR_CHECK_BUFPOS */
1129 /************************************************************************/
1130 /* verifying buffer and string positions */
1131 /************************************************************************/
1133 /* Functions below are tagged with either _byte or _char indicating
1134 whether they return byte or character positions. For a buffer,
1135 a character position is a "Bufpos" and a byte position is a "Bytind".
1136 For strings, these are sometimes typed using "Charcount" and
1139 /* Flags for the functions below are:
1141 GB_ALLOW_PAST_ACCESSIBLE
1143 Allow positions to range over the entire buffer (BUF_BEG to BUF_Z),
1144 rather than just the accessible portion (BUF_BEGV to BUF_ZV).
1145 For strings, this flag has no effect.
1149 If the position is outside the allowable range, return the lower
1150 or upper bound of the range, whichever is closer to the specified
1155 If the position is outside the allowable range, return -1.
1157 GB_NEGATIVE_FROM_END
1159 If a value is negative, treat it as an offset from the end.
1160 Only applies to strings.
1162 The following additional flags apply only to the functions
1167 Either or both positions can be nil. If FROM is nil,
1168 FROM_OUT will contain the lower bound of the allowed range.
1169 If TO is nil, TO_OUT will contain the upper bound of the
1174 FROM must contain the lower bound and TO the upper bound
1175 of the range. If the positions are reversed, an error is
1178 The following is a combination flag:
1180 GB_HISTORICAL_STRING_BEHAVIOR
1182 Equivalent to (GB_NEGATIVE_FROM_END | GB_ALLOW_NIL).
1185 /* Return a buffer position stored in a Lisp_Object. Full
1186 error-checking is done on the position. Flags can be specified to
1187 control the behavior of out-of-range values. The default behavior
1188 is to require that the position is within the accessible part of
1189 the buffer (BEGV and ZV), and to signal an error if the position is
1195 get_buffer_pos_char (struct buffer *b, Lisp_Object pos, unsigned int flags)
1199 Bufpos min_allowed, max_allowed;
1201 CHECK_INT_COERCE_MARKER (pos);
1203 min_allowed = flags & GB_ALLOW_PAST_ACCESSIBLE ? BUF_BEG (b) : BUF_BEGV (b);
1204 max_allowed = flags & GB_ALLOW_PAST_ACCESSIBLE ? BUF_Z (b) : BUF_ZV (b);
1206 if (ind < min_allowed || ind > max_allowed)
1208 if (flags & GB_COERCE_RANGE)
1209 ind = ind < min_allowed ? min_allowed : max_allowed;
1210 else if (flags & GB_NO_ERROR_IF_BAD)
1215 XSETBUFFER (buffer, b);
1216 args_out_of_range (buffer, pos);
1224 get_buffer_pos_byte (struct buffer *b, Lisp_Object pos, unsigned int flags)
1226 Bufpos bpos = get_buffer_pos_char (b, pos, flags);
1227 if (bpos < 0) /* could happen with GB_NO_ERROR_IF_BAD */
1229 return bufpos_to_bytind (b, bpos);
1232 /* Return a pair of buffer positions representing a range of text,
1233 taken from a pair of Lisp_Objects. Full error-checking is
1234 done on the positions. Flags can be specified to control the
1235 behavior of out-of-range values. The default behavior is to
1236 allow the range bounds to be specified in either order
1237 (however, FROM_OUT will always be the lower bound of the range
1238 and TO_OUT the upper bound),to require that the positions
1239 are within the accessible part of the buffer (BEGV and ZV),
1240 and to signal an error if the positions are out of range.
1244 get_buffer_range_char (struct buffer *b, Lisp_Object from, Lisp_Object to,
1245 Bufpos *from_out, Bufpos *to_out, unsigned int flags)
1248 Bufpos min_allowed, max_allowed;
1250 min_allowed = (flags & GB_ALLOW_PAST_ACCESSIBLE) ?
1251 BUF_BEG (b) : BUF_BEGV (b);
1252 max_allowed = (flags & GB_ALLOW_PAST_ACCESSIBLE) ?
1253 BUF_Z (b) : BUF_ZV (b);
1255 if (NILP (from) && (flags & GB_ALLOW_NIL))
1256 *from_out = min_allowed;
1258 *from_out = get_buffer_pos_char (b, from, flags | GB_NO_ERROR_IF_BAD);
1260 if (NILP (to) && (flags & GB_ALLOW_NIL))
1261 *to_out = max_allowed;
1263 *to_out = get_buffer_pos_char (b, to, flags | GB_NO_ERROR_IF_BAD);
1265 if ((*from_out < 0 || *to_out < 0) && !(flags & GB_NO_ERROR_IF_BAD))
1268 XSETBUFFER (buffer, b);
1269 args_out_of_range_3 (buffer, from, to);
1272 if (*from_out >= 0 && *to_out >= 0 && *from_out > *to_out)
1274 if (flags & GB_CHECK_ORDER)
1275 signal_simple_error_2 ("start greater than end", from, to);
1278 Bufpos temp = *from_out;
1279 *from_out = *to_out;
1286 get_buffer_range_byte (struct buffer *b, Lisp_Object from, Lisp_Object to,
1287 Bytind *from_out, Bytind *to_out, unsigned int flags)
1291 get_buffer_range_char (b, from, to, &s, &e, flags);
1293 *from_out = bufpos_to_bytind (b, s);
1294 else /* could happen with GB_NO_ERROR_IF_BAD */
1297 *to_out = bufpos_to_bytind (b, e);
1303 get_string_pos_char_1 (Lisp_Object string, Lisp_Object pos, unsigned int flags,
1304 Charcount known_length)
1307 Charcount min_allowed = 0;
1308 Charcount max_allowed = known_length;
1310 /* Computation of KNOWN_LENGTH is potentially expensive so we pass
1314 if (ccpos < 0 && flags & GB_NEGATIVE_FROM_END)
1315 ccpos += max_allowed;
1317 if (ccpos < min_allowed || ccpos > max_allowed)
1319 if (flags & GB_COERCE_RANGE)
1320 ccpos = ccpos < min_allowed ? min_allowed : max_allowed;
1321 else if (flags & GB_NO_ERROR_IF_BAD)
1324 args_out_of_range (string, pos);
1331 get_string_pos_char (Lisp_Object string, Lisp_Object pos, unsigned int flags)
1333 return get_string_pos_char_1 (string, pos, flags,
1334 XSTRING_CHAR_LENGTH (string));
1338 get_string_pos_byte (Lisp_Object string, Lisp_Object pos, unsigned int flags)
1340 Charcount ccpos = get_string_pos_char (string, pos, flags);
1341 if (ccpos < 0) /* could happen with GB_NO_ERROR_IF_BAD */
1343 return charcount_to_bytecount (XSTRING_DATA (string), ccpos);
1347 get_string_range_char (Lisp_Object string, Lisp_Object from, Lisp_Object to,
1348 Charcount *from_out, Charcount *to_out,
1351 Charcount min_allowed = 0;
1352 Charcount max_allowed = XSTRING_CHAR_LENGTH (string);
1354 if (NILP (from) && (flags & GB_ALLOW_NIL))
1355 *from_out = min_allowed;
1357 *from_out = get_string_pos_char_1 (string, from,
1358 flags | GB_NO_ERROR_IF_BAD,
1361 if (NILP (to) && (flags & GB_ALLOW_NIL))
1362 *to_out = max_allowed;
1364 *to_out = get_string_pos_char_1 (string, to,
1365 flags | GB_NO_ERROR_IF_BAD,
1368 if ((*from_out < 0 || *to_out < 0) && !(flags & GB_NO_ERROR_IF_BAD))
1369 args_out_of_range_3 (string, from, to);
1371 if (*from_out >= 0 && *to_out >= 0 && *from_out > *to_out)
1373 if (flags & GB_CHECK_ORDER)
1374 signal_simple_error_2 ("start greater than end", from, to);
1377 Bufpos temp = *from_out;
1378 *from_out = *to_out;
1385 get_string_range_byte (Lisp_Object string, Lisp_Object from, Lisp_Object to,
1386 Bytecount *from_out, Bytecount *to_out,
1391 get_string_range_char (string, from, to, &s, &e, flags);
1393 *from_out = charcount_to_bytecount (XSTRING_DATA (string), s);
1394 else /* could happen with GB_NO_ERROR_IF_BAD */
1397 *to_out = charcount_to_bytecount (XSTRING_DATA (string), e);
1404 get_buffer_or_string_pos_char (Lisp_Object object, Lisp_Object pos,
1407 return STRINGP (object) ?
1408 get_string_pos_char (object, pos, flags) :
1409 get_buffer_pos_char (XBUFFER (object), pos, flags);
1413 get_buffer_or_string_pos_byte (Lisp_Object object, Lisp_Object pos,
1416 return STRINGP (object) ?
1417 get_string_pos_byte (object, pos, flags) :
1418 get_buffer_pos_byte (XBUFFER (object), pos, flags);
1422 get_buffer_or_string_range_char (Lisp_Object object, Lisp_Object from,
1423 Lisp_Object to, Bufpos *from_out,
1424 Bufpos *to_out, unsigned int flags)
1426 if (STRINGP (object))
1427 get_string_range_char (object, from, to, from_out, to_out, flags);
1429 get_buffer_range_char (XBUFFER (object), from, to, from_out, to_out, flags);
1433 get_buffer_or_string_range_byte (Lisp_Object object, Lisp_Object from,
1434 Lisp_Object to, Bytind *from_out,
1435 Bytind *to_out, unsigned int flags)
1437 if (STRINGP (object))
1438 get_string_range_byte (object, from, to, from_out, to_out, flags);
1440 get_buffer_range_byte (XBUFFER (object), from, to, from_out, to_out, flags);
1444 buffer_or_string_accessible_begin_char (Lisp_Object object)
1446 return STRINGP (object) ? 0 : BUF_BEGV (XBUFFER (object));
1450 buffer_or_string_accessible_end_char (Lisp_Object object)
1452 return STRINGP (object) ?
1453 XSTRING_CHAR_LENGTH (object) : BUF_ZV (XBUFFER (object));
1457 buffer_or_string_accessible_begin_byte (Lisp_Object object)
1459 return STRINGP (object) ? 0 : BI_BUF_BEGV (XBUFFER (object));
1463 buffer_or_string_accessible_end_byte (Lisp_Object object)
1465 return STRINGP (object) ?
1466 XSTRING_LENGTH (object) : BI_BUF_ZV (XBUFFER (object));
1470 buffer_or_string_absolute_begin_char (Lisp_Object object)
1472 return STRINGP (object) ? 0 : BUF_BEG (XBUFFER (object));
1476 buffer_or_string_absolute_end_char (Lisp_Object object)
1478 return STRINGP (object) ?
1479 XSTRING_CHAR_LENGTH (object) : BUF_Z (XBUFFER (object));
1483 buffer_or_string_absolute_begin_byte (Lisp_Object object)
1485 return STRINGP (object) ? 0 : BI_BUF_BEG (XBUFFER (object));
1489 buffer_or_string_absolute_end_byte (Lisp_Object object)
1491 return STRINGP (object) ?
1492 XSTRING_LENGTH (object) : BI_BUF_Z (XBUFFER (object));
1496 /************************************************************************/
1497 /* point and marker adjustment */
1498 /************************************************************************/
1500 /* just_set_point() is the only place `PT' is an lvalue in all of emacs.
1501 This function is called from set_buffer_point(), which is the function
1502 that the SET_PT and BUF_SET_PT macros expand into, and from the
1503 routines below that insert and delete text. (This is in cases where
1504 the point marker logically doesn't move but PT (being a byte index)
1505 needs to get adjusted.) */
1507 /* Set point to a specified value. This is used only when the value
1508 of point changes due to an insert or delete; it does not represent
1509 a conceptual change in point as a marker. In particular, point is
1510 not crossing any interval boundaries, so there's no need to use the
1511 usual SET_PT macro. In fact it would be incorrect to do so, because
1512 either the old or the new value of point is out of synch with the
1513 current set of intervals. */
1515 /* This gets called more than enough to make the function call
1516 overhead a significant factor so we've turned it into a macro. */
1517 #define JUST_SET_POINT(buf, bufpos, ind) \
1520 buf->bufpt = (bufpos); \
1524 /* Set a buffer's point. */
1527 set_buffer_point (struct buffer *buf, Bufpos bufpos, Bytind bytpos)
1529 assert (bytpos >= BI_BUF_BEGV (buf) && bytpos <= BI_BUF_ZV (buf));
1530 if (bytpos == BI_BUF_PT (buf))
1532 JUST_SET_POINT (buf, bufpos, bytpos);
1534 assert (MARKERP (buf->point_marker));
1535 XMARKER (buf->point_marker)->memind =
1536 bytind_to_memind (buf, bytpos);
1538 /* FSF makes sure that PT is not being set within invisible text.
1539 However, this is the wrong place for that check. The check
1540 should happen only at the next redisplay. */
1542 /* Some old coder said:
1544 "If there were to be hooks which were run when point entered/left an
1545 extent, this would be the place to put them.
1547 However, it's probably the case that such hooks should be implemented
1548 using a post-command-hook instead, to avoid running the hooks as a
1549 result of intermediate motion inside of save-excursions, for example."
1551 I definitely agree with this. PT gets moved all over the place
1552 and it would be a Bad Thing for any hooks to get called, both for
1553 the reason above and because many callers are not prepared for
1554 a GC within this function. --ben
1558 /* Do the correct marker-like adjustment on MPOS (see below). FROM, TO,
1559 and AMOUNT are as in adjust_markers(). If MPOS doesn't need to be
1560 adjusted, nothing will happen. */
1562 do_marker_adjustment (Memind mpos, Memind from,
1563 Memind to, Bytecount amount)
1567 if (mpos > to && mpos < to + amount)
1572 if (mpos > from + amount && mpos <= from)
1573 mpos = from + amount;
1575 if (mpos > from && mpos <= to)
1580 /* Do the following:
1582 (1) Add `amount' to the position of every marker in the current buffer
1583 whose current position is between `from' (exclusive) and `to' (inclusive).
1585 (2) Also, any markers past the outside of that interval, in the direction
1586 of adjustment, are first moved back to the near end of the interval
1587 and then adjusted by `amount'.
1589 This function is called in two different cases: when a region of
1590 characters adjacent to the gap is moved, causing the gap to shift
1591 to the other side of the region (in this case, `from' and `to'
1592 point to the old position of the region and there should be no
1593 markers affected by (2) because they would be inside the gap),
1594 or when a region of characters adjacent to the gap is wiped out,
1595 causing the gap to increase to include the region (in this case,
1596 `from' and `to' are the same, both pointing to the boundary
1597 between the gap and the deleted region, and there are no markers
1600 The reason for the use of exclusive and inclusive is that markers at
1601 the gap always sit at the beginning, not at the end.
1605 adjust_markers (struct buffer *buf, Memind from, Memind to,
1610 for (m = BUF_MARKERS (buf); m; m = marker_next (m))
1611 m->memind = do_marker_adjustment (m->memind, from, to, amount);
1614 /* Adjust markers whose insertion-type is t
1615 for an insertion of AMOUNT characters at POS. */
1618 adjust_markers_for_insert (struct buffer *buf, Memind ind, Bytecount amount)
1622 for (m = BUF_MARKERS (buf); m; m = marker_next (m))
1624 if (m->insertion_type && m->memind == ind)
1625 m->memind += amount;
1630 /************************************************************************/
1631 /* Routines for dealing with the gap */
1632 /************************************************************************/
1634 /* maximum amount of memory moved in a single chunk. Increasing this
1635 value improves gap-motion efficiency but decreases QUIT responsiveness
1636 time. Was 32000 but today's processors are faster and files are
1638 #define GAP_MOVE_CHUNK 300000
1640 /* Move the gap to POS, which is less than the current GPT. */
1643 gap_left (struct buffer *buf, Bytind pos)
1648 struct buffer *mbuf;
1649 Lisp_Object bufcons;
1651 from = BUF_GPT_ADDR (buf);
1652 to = from + BUF_GAP_SIZE (buf);
1653 new_s1 = BI_BUF_GPT (buf);
1655 /* Now copy the characters. To move the gap down,
1656 copy characters up. */
1660 /* I gets number of characters left to copy. */
1664 /* If a quit is requested, stop copying now.
1665 Change POS to be where we have actually moved the gap to. */
1671 /* Move at most GAP_MOVE_CHUNK chars before checking again for a quit. */
1672 if (i > GAP_MOVE_CHUNK)
1680 memmove (to, from, i);
1690 /* Adjust markers, and buffer data structure, to put the gap at POS.
1691 POS is where the loop above stopped, which may be what was specified
1692 or may be where a quit was detected. */
1693 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
1695 adjust_markers (mbuf, pos, BI_BUF_GPT (mbuf), BUF_GAP_SIZE (mbuf));
1697 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
1699 adjust_extents (make_buffer (mbuf), pos, BI_BUF_GPT (mbuf),
1700 BUF_GAP_SIZE (mbuf));
1702 SET_BI_BUF_GPT (buf, pos);
1703 SET_GAP_SENTINEL (buf);
1704 #ifdef ERROR_CHECK_EXTENTS
1705 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
1707 sledgehammer_extent_check (make_buffer (mbuf));
1714 gap_right (struct buffer *buf, Bytind pos)
1719 struct buffer *mbuf;
1720 Lisp_Object bufcons;
1722 to = BUF_GPT_ADDR (buf);
1723 from = to + BUF_GAP_SIZE (buf);
1724 new_s1 = BI_BUF_GPT (buf);
1726 /* Now copy the characters. To move the gap up,
1727 copy characters down. */
1731 /* I gets number of characters left to copy. */
1735 /* If a quit is requested, stop copying now.
1736 Change POS to be where we have actually moved the gap to. */
1742 /* Move at most GAP_MOVE_CHUNK chars before checking again for a quit. */
1743 if (i > GAP_MOVE_CHUNK)
1749 memmove (to, from, i);
1762 int gsize = BUF_GAP_SIZE (buf);
1763 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
1765 adjust_markers (mbuf, BI_BUF_GPT (mbuf) + gsize, pos + gsize, - gsize);
1767 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
1769 adjust_extents (make_buffer (mbuf), BI_BUF_GPT (mbuf) + gsize,
1770 pos + gsize, - gsize);
1772 SET_BI_BUF_GPT (buf, pos);
1773 SET_GAP_SENTINEL (buf);
1774 #ifdef ERROR_CHECK_EXTENTS
1775 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
1777 sledgehammer_extent_check (make_buffer (mbuf));
1781 if (pos == BI_BUF_Z (buf))
1783 /* merge gap with end gap */
1785 SET_BUF_GAP_SIZE (buf, BUF_GAP_SIZE (buf) + BUF_END_GAP_SIZE (buf));
1786 SET_BUF_END_GAP_SIZE (buf, 0);
1787 SET_END_SENTINEL (buf);
1793 /* Move gap to position `pos'.
1794 Note that this can quit! */
1797 move_gap (struct buffer *buf, Bytind pos)
1799 if (! BUF_BEG_ADDR (buf))
1801 if (pos < BI_BUF_GPT (buf))
1802 gap_left (buf, pos);
1803 else if (pos > BI_BUF_GPT (buf))
1804 gap_right (buf, pos);
1807 /* Merge the end gap into the gap */
1810 merge_gap_with_end_gap (struct buffer *buf)
1813 Bytind real_gap_loc;
1814 Bytecount old_gap_size;
1815 Bytecount increment;
1817 increment = BUF_END_GAP_SIZE (buf);
1818 SET_BUF_END_GAP_SIZE (buf, 0);
1822 /* Prevent quitting in move_gap. */
1823 tem = Vinhibit_quit;
1826 real_gap_loc = BI_BUF_GPT (buf);
1827 old_gap_size = BUF_GAP_SIZE (buf);
1829 /* Pretend the end gap is the gap */
1830 SET_BI_BUF_GPT (buf, BI_BUF_Z (buf) + BUF_GAP_SIZE (buf));
1831 SET_BUF_GAP_SIZE (buf, increment);
1833 /* Move the new gap down to be consecutive with the end of the old one.
1834 This adjusts the markers properly too. */
1835 gap_left (buf, real_gap_loc + old_gap_size);
1837 /* Now combine the two into one large gap. */
1838 SET_BUF_GAP_SIZE (buf, BUF_GAP_SIZE (buf) + old_gap_size);
1839 SET_BI_BUF_GPT (buf, real_gap_loc);
1840 SET_GAP_SENTINEL (buf);
1842 /* We changed the total size of the buffer (including gap),
1843 so we need to fix up the end sentinel. */
1844 SET_END_SENTINEL (buf);
1846 Vinhibit_quit = tem;
1850 /* Make the gap INCREMENT bytes longer. */
1853 make_gap (struct buffer *buf, Bytecount increment)
1857 Bytind real_gap_loc;
1858 Bytecount old_gap_size;
1860 /* If we have to get more space, get enough to last a while. We use
1861 a geometric progression that saves on realloc space. */
1862 increment += 2000 + ((BI_BUF_Z (buf) - BI_BUF_BEG (buf)) / 8);
1864 if (increment > BUF_END_GAP_SIZE (buf))
1866 /* Don't allow a buffer size that won't fit in an int
1867 even if it will fit in a Lisp integer.
1868 That won't work because so many places use `int'. */
1870 if (BUF_Z (buf) - BUF_BEG (buf) + BUF_GAP_SIZE (buf) + increment
1872 error ("Maximum buffer size exceeded");
1874 result = BUFFER_REALLOC (buf->text->beg,
1875 BI_BUF_Z (buf) - BI_BUF_BEG (buf) +
1876 BUF_GAP_SIZE (buf) + increment +
1877 BUF_END_SENTINEL_SIZE);
1881 SET_BUF_BEG_ADDR (buf, result);
1884 increment = BUF_END_GAP_SIZE (buf);
1886 /* Prevent quitting in move_gap. */
1887 tem = Vinhibit_quit;
1890 real_gap_loc = BI_BUF_GPT (buf);
1891 old_gap_size = BUF_GAP_SIZE (buf);
1893 /* Call the newly allocated space a gap at the end of the whole space. */
1894 SET_BI_BUF_GPT (buf, BI_BUF_Z (buf) + BUF_GAP_SIZE (buf));
1895 SET_BUF_GAP_SIZE (buf, increment);
1897 SET_BUF_END_GAP_SIZE (buf, 0);
1899 /* Move the new gap down to be consecutive with the end of the old one.
1900 This adjusts the markers properly too. */
1901 gap_left (buf, real_gap_loc + old_gap_size);
1903 /* Now combine the two into one large gap. */
1904 SET_BUF_GAP_SIZE (buf, BUF_GAP_SIZE (buf) + old_gap_size);
1905 SET_BI_BUF_GPT (buf, real_gap_loc);
1906 SET_GAP_SENTINEL (buf);
1908 /* We changed the total size of the buffer (including gap),
1909 so we need to fix up the end sentinel. */
1910 SET_END_SENTINEL (buf);
1912 Vinhibit_quit = tem;
1916 /************************************************************************/
1917 /* Before/after-change processing */
1918 /************************************************************************/
1920 /* Those magic changes ... */
1923 buffer_signal_changed_region (struct buffer *buf, Bufpos start,
1926 /* The changed region is recorded as the number of unchanged
1927 characters from the beginning and from the end of the
1928 buffer. This obviates much of the need of shifting the
1929 region around to compensate for insertions and deletions.
1931 if (buf->changes->begin_unchanged < 0 ||
1932 buf->changes->begin_unchanged > start - BUF_BEG (buf))
1933 buf->changes->begin_unchanged = start - BUF_BEG (buf);
1934 if (buf->changes->end_unchanged < 0 ||
1935 buf->changes->end_unchanged > BUF_Z (buf) - end)
1936 buf->changes->end_unchanged = BUF_Z (buf) - end;
1940 buffer_extent_signal_changed_region (struct buffer *buf, Bufpos start,
1943 if (buf->changes->begin_extent_unchanged < 0 ||
1944 buf->changes->begin_extent_unchanged > start - BUF_BEG (buf))
1945 buf->changes->begin_extent_unchanged = start - BUF_BEG (buf);
1946 if (buf->changes->end_extent_unchanged < 0 ||
1947 buf->changes->end_extent_unchanged > BUF_Z (buf) - end)
1948 buf->changes->end_extent_unchanged = BUF_Z (buf) - end;
1952 buffer_reset_changes (struct buffer *buf)
1954 buf->changes->begin_unchanged = -1;
1955 buf->changes->end_unchanged = -1;
1956 buf->changes->begin_extent_unchanged = -1;
1957 buf->changes->end_extent_unchanged = -1;
1958 buf->changes->newline_was_deleted = 0;
1962 signal_after_change (struct buffer *buf, Bufpos start, Bufpos orig_end,
1966 /* Call the after-change-functions according to the changes made so far
1967 and treat all further changes as single until the outermost
1968 multiple change exits. This is called when the outermost multiple
1969 change exits and when someone is trying to make a change that violates
1970 the constraints specified in begin_multiple_change(), typically
1971 when nested multiple-change sessions occur. (There are smarter ways of
1972 dealing with nested multiple changes, but these rarely occur so there's
1973 probably no point in it.) */
1975 /* #### This needs to keep track of what actually changed and only
1976 call the after-change functions on that region. */
1979 cancel_multiple_change (struct buffer *buf)
1981 /* This function can GC */
1982 /* Call the after-change-functions except when they've already been
1983 called or when there were no changes made to the buffer at all. */
1984 if (buf->text->changes->mc_begin != 0 &&
1985 buf->text->changes->mc_begin_signaled)
1987 Bufpos real_mc_begin = buf->text->changes->mc_begin;
1988 buf->text->changes->mc_begin = 0;
1990 signal_after_change (buf, real_mc_begin, buf->text->changes->mc_orig_end,
1991 buf->text->changes->mc_new_end);
1995 buf->text->changes->mc_begin = 0;
1999 /* this is an unwind_protect, to ensure that the after-change-functions
2000 get called even in a non-local exit. */
2003 multiple_change_finish_up (Lisp_Object buffer)
2005 struct buffer *buf = XBUFFER (buffer);
2007 /* #### I don't know whether or not it should even be possible to
2008 get here with a dead buffer (though given how it is called I can
2009 see how it might be). In any case, there isn't time before 19.14
2011 if (!BUFFER_LIVE_P (buf))
2014 /* This function can GC */
2015 buf->text->changes->in_multiple_change = 0; /* do this first so that
2016 errors in the after-change
2017 functions don't mess things
2019 cancel_multiple_change (buf);
2023 /* Call this function when you're about to make a number of buffer changes
2024 that should be considered a single change. (e.g. `replace-match' calls
2025 this.) You need to specify the START and END of the region that is
2026 going to be changed so that the before-change-functions are called
2027 with the correct arguments. The after-change region is calculated
2028 automatically, however, and if changes somehow or other happen outside
2029 of the specified region, that will also be handled correctly.
2031 begin_multiple_change() returns a number (actually a specpdl depth)
2032 that you must pass to end_multiple_change() when you are done.
2034 FSF Emacs 20 implements a similar feature, accessible from Lisp
2035 through a `combine-after-change-calls' special form, which is
2036 essentially equivalent to this function. We should consider
2037 whether we want to introduce a similar Lisp form. */
2040 begin_multiple_change (struct buffer *buf, Bufpos start, Bufpos end)
2042 /* This function can GC */
2044 if (buf->text->changes->in_multiple_change)
2046 if (buf->text->changes->mc_begin != 0 &&
2047 (start < buf->text->changes->mc_begin ||
2048 end > buf->text->changes->mc_new_end))
2049 cancel_multiple_change (buf);
2055 buf->text->changes->mc_begin = start;
2056 buf->text->changes->mc_orig_end = buf->text->changes->mc_new_end = end;
2057 buf->text->changes->mc_begin_signaled = 0;
2058 count = specpdl_depth ();
2059 XSETBUFFER (buffer, buf);
2060 record_unwind_protect (multiple_change_finish_up, buffer);
2062 buf->text->changes->in_multiple_change++;
2063 /* We don't call before-change-functions until signal_before_change()
2064 is called, in case there is a read-only or other error. */
2069 end_multiple_change (struct buffer *buf, int count)
2071 assert (buf->text->changes->in_multiple_change > 0);
2072 buf->text->changes->in_multiple_change--;
2073 if (!buf->text->changes->in_multiple_change)
2074 unbind_to (count, Qnil);
2077 static int inside_change_hook;
2080 change_function_restore (Lisp_Object buffer)
2082 /* We should first reset the variable and then change the buffer,
2083 because Fset_buffer() can throw. */
2084 inside_change_hook = 0;
2085 if (XBUFFER (buffer) != current_buffer)
2086 Fset_buffer (buffer);
2090 static int in_first_change;
2093 first_change_hook_restore (Lisp_Object buffer)
2095 in_first_change = 0;
2096 Fset_buffer (buffer);
2100 /* Signal an initial modification to the buffer. */
2103 signal_first_change (struct buffer *buf)
2105 /* This function can GC */
2107 XSETBUFFER (buffer, current_buffer);
2109 if (!in_first_change)
2111 if (!NILP (symbol_value_in_buffer (Qfirst_change_hook, buffer)))
2113 int speccount = specpdl_depth ();
2114 record_unwind_protect (first_change_hook_restore, buffer);
2115 set_buffer_internal (buf);
2116 in_first_change = 1;
2117 run_hook (Qfirst_change_hook);
2118 unbind_to (speccount, Qnil);
2123 /* Signal a change to the buffer immediately before it happens.
2124 START and END are the bounds of the text to be changed. */
2127 signal_before_change (struct buffer *buf, Bufpos start, Bufpos end)
2129 /* This function can GC */
2130 struct buffer *mbuf;
2131 Lisp_Object bufcons;
2133 if (!inside_change_hook)
2138 /* Are we in a multiple-change session? */
2139 if (buf->text->changes->in_multiple_change &&
2140 buf->text->changes->mc_begin != 0)
2142 /* If we're violating the constraints of the session,
2143 call the after-change-functions as necessary for the
2144 changes already made and treat further changes as
2146 if (start < buf->text->changes->mc_begin ||
2147 end > buf->text->changes->mc_new_end)
2148 cancel_multiple_change (buf);
2149 /* Do nothing if this is not the first change in the session. */
2150 else if (buf->text->changes->mc_begin_signaled)
2154 /* First time through; call the before-change-functions
2155 specifying the entire region to be changed. (Note that
2156 we didn't call before-change-functions in
2157 begin_multiple_change() because the buffer might be
2159 start = buf->text->changes->mc_begin;
2160 end = buf->text->changes->mc_new_end;
2164 /* If buffer is unmodified, run a special hook for that case. */
2165 if (BUF_SAVE_MODIFF (buf) >= BUF_MODIFF (buf))
2167 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2169 signal_first_change (mbuf);
2173 /* Now in any case run the before-change-functions if any. */
2174 speccount = specpdl_depth ();
2175 record_unwind_protect (change_function_restore, Fcurrent_buffer ());
2176 inside_change_hook = 1;
2178 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2180 XSETBUFFER (buffer, mbuf);
2181 if (!NILP (symbol_value_in_buffer (Qbefore_change_functions, buffer))
2182 /* Obsolete, for compatibility */
2183 || !NILP (symbol_value_in_buffer (Qbefore_change_function, buffer)))
2185 set_buffer_internal (buf);
2186 va_run_hook_with_args (Qbefore_change_functions, 2,
2187 make_int (start), make_int (end));
2188 /* Obsolete, for compatibility */
2189 va_run_hook_with_args (Qbefore_change_function, 2,
2190 make_int (start), make_int (end));
2194 /* Make sure endpoints remain valid. before-change-functions
2195 might have modified the buffer. */
2196 if (start < BUF_BEGV (buf)) start = BUF_BEGV (buf);
2197 if (start > BUF_ZV (buf)) start = BUF_ZV (buf);
2198 if (end < BUF_BEGV (buf)) end = BUF_BEGV (buf);
2199 if (end > BUF_ZV (buf)) end = BUF_ZV (buf);
2201 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2203 XSETBUFFER (buffer, mbuf);
2204 report_extent_modification (buffer, start, end, 0);
2206 unbind_to (speccount, Qnil);
2208 /* Only now do we indicate that the before-change-functions have
2209 been called, in case some function throws out. */
2210 buf->text->changes->mc_begin_signaled = 1;
2214 /* Signal a change immediately after it happens.
2215 START is the bufpos of the start of the changed text.
2216 ORIG_END is the bufpos of the end of the before-changed text.
2217 NEW_END is the bufpos of the end of the after-changed text.
2221 signal_after_change (struct buffer *buf, Bufpos start, Bufpos orig_end,
2224 /* This function can GC */
2225 struct buffer *mbuf;
2226 Lisp_Object bufcons;
2228 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2230 /* always do this. */
2231 buffer_signal_changed_region (mbuf, start, new_end);
2233 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2235 /* #### This seems inefficient. Wouldn't it be better to just
2236 keep one cache per base buffer? */
2237 font_lock_maybe_update_syntactic_caches (mbuf, start, orig_end, new_end);
2240 if (!inside_change_hook)
2245 if (buf->text->changes->in_multiple_change &&
2246 buf->text->changes->mc_begin != 0)
2248 assert (start >= buf->text->changes->mc_begin &&
2249 start <= buf->text->changes->mc_new_end);
2250 assert (orig_end >= buf->text->changes->mc_begin &&
2251 orig_end <= buf->text->changes->mc_new_end);
2252 buf->text->changes->mc_new_end += new_end - orig_end;
2253 return; /* after-change-functions signalled when all changes done */
2256 speccount = specpdl_depth ();
2257 record_unwind_protect (change_function_restore, Fcurrent_buffer ());
2258 inside_change_hook = 1;
2259 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2261 XSETBUFFER (buffer, mbuf);
2263 if (!NILP (symbol_value_in_buffer (Qafter_change_functions, buffer))
2264 /* Obsolete, for compatibility */
2265 || !NILP (symbol_value_in_buffer (Qafter_change_function, buffer)))
2267 set_buffer_internal (buf);
2268 /* The actual after-change functions take slightly
2269 different arguments than what we were passed. */
2270 va_run_hook_with_args (Qafter_change_functions, 3,
2271 make_int (start), make_int (new_end),
2272 make_int (orig_end - start));
2273 /* Obsolete, for compatibility */
2274 va_run_hook_with_args (Qafter_change_function, 3,
2275 make_int (start), make_int (new_end),
2276 make_int (orig_end - start));
2280 /* Make sure endpoints remain valid. after-change-functions
2281 might have modified the buffer. */
2282 if (start < BUF_BEGV (buf)) start = BUF_BEGV (buf);
2283 if (start > BUF_ZV (buf)) start = BUF_ZV (buf);
2284 if (new_end < BUF_BEGV (buf)) new_end = BUF_BEGV (buf);
2285 if (new_end > BUF_ZV (buf)) new_end = BUF_ZV (buf);
2286 if (orig_end < BUF_BEGV (buf)) orig_end = BUF_BEGV (buf);
2287 if (orig_end > BUF_ZV (buf)) orig_end = BUF_ZV (buf);
2289 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2291 XSETBUFFER (buffer, mbuf);
2292 report_extent_modification (buffer, start, new_end, 1);
2294 unbind_to (speccount, Qnil); /* sets inside_change_hook back to 0 */
2298 /* Call this if you're about to change the region of BUFFER from START
2299 to END. This checks the read-only properties of the region, calls
2300 the necessary modification hooks, and warns the next redisplay that
2301 it should pay attention to that area. */
2304 prepare_to_modify_buffer (struct buffer *buf, Bufpos start, Bufpos end,
2307 /* This function can GC */
2308 /* dmoore - This function can also kill the buffer buf, the current
2309 buffer, and do anything it pleases. So if you call it, be
2311 struct buffer *mbuf;
2312 Lisp_Object buffer, bufcons;
2313 struct gcpro gcpro1;
2315 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2317 barf_if_buffer_read_only (mbuf, start, end);
2320 /* if this is the first modification, see about locking the buffer's
2322 XSETBUFFER (buffer, buf);
2324 if (!NILP (buf->filename) && lockit &&
2325 BUF_SAVE_MODIFF (buf) >= BUF_MODIFF (buf))
2327 /* At least warn if this file has changed on disk since it was visited.*/
2328 if (NILP (Fverify_visited_file_modtime (buffer))
2329 && !NILP (Ffile_exists_p (buf->filename)))
2330 call1_in_buffer (buf, intern ("ask-user-about-supersession-threat"),
2332 #ifdef CLASH_DETECTION
2333 if (!NILP (buf->file_truename))
2334 /* Make binding buffer-file-name to nil effective. */
2335 lock_file (buf->file_truename);
2336 #endif /* not CLASH_DETECTION */
2340 /* #### dmoore - is this reasonable in case of buf being killed above? */
2341 if (!BUFFER_LIVE_P (buf))
2344 signal_before_change (buf, start, end);
2346 #ifdef REGION_CACHE_NEEDS_WORK
2347 if (buf->newline_cache)
2348 invalidate_region_cache (buf,
2350 start - BUF_BEG (buf), BUF_Z (buf) - end);
2351 if (buf->width_run_cache)
2352 invalidate_region_cache (buf,
2353 buf->width_run_cache,
2354 start - BUF_BEG (buf), BUF_Z (buf) - end);
2358 Vdeactivate_mark = Qt;
2361 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2363 mbuf->point_before_scroll = Qnil;
2368 /************************************************************************/
2369 /* Insertion of strings */
2370 /************************************************************************/
2373 fixup_internal_substring (const Bufbyte *nonreloc, Lisp_Object reloc,
2374 Bytecount offset, Bytecount *len)
2376 assert ((nonreloc && NILP (reloc)) || (!nonreloc && STRINGP (reloc)));
2381 *len = strlen ((const char *) nonreloc) - offset;
2383 *len = XSTRING_LENGTH (reloc) - offset;
2385 #ifdef ERROR_CHECK_BUFPOS
2387 if (STRINGP (reloc))
2389 assert (offset >= 0 && offset <= XSTRING_LENGTH (reloc));
2390 assert (offset + *len <= XSTRING_LENGTH (reloc));
2395 /* Insert a string into BUF at Bufpos POS. The string data comes
2396 from one of two sources: constant, non-relocatable data (specified
2397 in NONRELOC), or a Lisp string object (specified in RELOC), which
2398 is relocatable and may have extent data that needs to be copied
2399 into the buffer. OFFSET and LENGTH specify the substring of the
2400 data that is actually to be inserted. As a special case, if POS
2401 is -1, insert the string at point and move point to the end of the
2404 Normally, markers at the insertion point end up before the
2405 inserted string. If INSDEL_BEFORE_MARKERS is set in flags, however,
2406 they end up after the string.
2408 INSDEL_NO_LOCKING is kludgy and is used when insert-file-contents is
2409 visiting a new file; it inhibits the locking checks normally done
2410 before modifying a buffer. Similar checks were already done
2411 in the higher-level Lisp functions calling insert-file-contents. */
2414 buffer_insert_string_1 (struct buffer *buf, Bufpos pos,
2415 const Bufbyte *nonreloc, Lisp_Object reloc,
2416 Bytecount offset, Bytecount length,
2419 /* This function can GC */
2420 struct gcpro gcpro1;
2424 struct buffer *mbuf;
2425 Lisp_Object bufcons;
2427 /* Defensive steps just in case a buffer gets deleted and a calling
2428 function doesn't notice it. */
2429 if (!BUFFER_LIVE_P (buf))
2432 fixup_internal_substring (nonreloc, reloc, offset, &length);
2441 /* #### See the comment in print_internal(). If this buffer is marked
2442 as translatable, then Fgettext() should be called on obj if it
2446 /* Make sure that point-max won't exceed the size of an emacs int. */
2447 if ((length + BUF_Z (buf)) > EMACS_INT_MAX)
2448 error ("Maximum buffer size exceeded");
2450 /* theoretically not necessary -- caller should GCPRO.
2451 #### buffer_insert_from_buffer_1() doesn't! */
2454 prepare_to_modify_buffer (buf, pos, pos, !(flags & INSDEL_NO_LOCKING));
2456 /* Defensive steps in case the before-change-functions fuck around */
2457 if (!BUFFER_LIVE_P (buf))
2460 /* Bad bad pre-change function. */
2464 /* Make args be valid again. prepare_to_modify_buffer() might have
2465 modified the buffer. */
2466 if (pos < BUF_BEGV (buf))
2467 pos = BUF_BEGV (buf);
2468 if (pos > BUF_ZV (buf))
2471 /* string may have been relocated up to this point */
2472 if (STRINGP (reloc))
2473 nonreloc = XSTRING_DATA (reloc);
2475 ind = bufpos_to_bytind (buf, pos);
2476 cclen = bytecount_to_charcount (nonreloc + offset, length);
2478 if (ind != BI_BUF_GPT (buf))
2479 /* #### if debug-on-quit is invoked and the user changes the
2480 buffer, bad things can happen. This is a rampant problem
2482 move_gap (buf, ind); /* may QUIT */
2483 if (! GAP_CAN_HOLD_SIZE_P (buf, length))
2485 if (BUF_END_GAP_SIZE (buf) >= length)
2486 merge_gap_with_end_gap (buf);
2488 make_gap (buf, length - BUF_GAP_SIZE (buf));
2491 insert_invalidate_line_number_cache (buf, pos, nonreloc + offset, length);
2493 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2495 record_insert (mbuf, pos, cclen);
2499 MARK_BUFFERS_CHANGED;
2501 /* string may have been relocated up to this point */
2502 if (STRINGP (reloc))
2503 nonreloc = XSTRING_DATA (reloc);
2505 memcpy (BUF_GPT_ADDR (buf), nonreloc + offset, length);
2507 SET_BUF_GAP_SIZE (buf, BUF_GAP_SIZE (buf) - length);
2508 SET_BI_BUF_GPT (buf, BI_BUF_GPT (buf) + length);
2509 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2511 SET_BOTH_BUF_ZV (mbuf, BUF_ZV (mbuf) + cclen, BI_BUF_ZV (mbuf) + length);
2513 SET_BOTH_BUF_Z (buf, BUF_Z (buf) + cclen, BI_BUF_Z (buf) + length);
2514 SET_GAP_SENTINEL (buf);
2517 buffer_mule_signal_inserted_region (buf, pos, length, cclen);
2520 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2522 process_extents_for_insertion (make_buffer (mbuf), ind, length);
2525 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2527 /* We know the gap is at IND so the cast is OK. */
2528 adjust_markers_for_insert (mbuf, (Memind) ind, length);
2531 /* Point logically doesn't move, but may need to be adjusted because
2532 it's a byte index. point-marker doesn't change because it's a
2534 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2536 if (BI_BUF_PT (mbuf) > ind)
2537 JUST_SET_POINT (mbuf, BUF_PT (mbuf) + cclen,
2538 BI_BUF_PT (mbuf) + length);
2541 /* Well, point might move. */
2543 BI_BUF_SET_PT (buf, ind + length);
2545 if (STRINGP (reloc))
2547 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2549 splice_in_string_extents (reloc, mbuf, ind, length, offset);
2553 if (flags & INSDEL_BEFORE_MARKERS)
2555 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2557 /* ind - 1 is correct because the FROM argument is exclusive.
2558 I formerly used DEC_BYTIND() but that caused problems at the
2559 beginning of the buffer. */
2560 adjust_markers (mbuf, ind - 1, ind, length);
2564 signal_after_change (buf, pos, pos, pos + cclen);
2572 /* The following functions are interfaces onto the above function,
2573 for inserting particular sorts of data. In all the functions,
2574 BUF and POS specify the buffer and location where the insertion is
2575 to take place. (If POS is -1, text is inserted at point and point
2576 moves forward past the text.) FLAGS is as above. */
2579 buffer_insert_raw_string_1 (struct buffer *buf, Bufpos pos,
2580 const Bufbyte *nonreloc, Bytecount length,
2583 /* This function can GC */
2584 return buffer_insert_string_1 (buf, pos, nonreloc, Qnil, 0, length,
2589 buffer_insert_lisp_string_1 (struct buffer *buf, Bufpos pos, Lisp_Object str,
2592 /* This function can GC */
2593 #ifdef ERROR_CHECK_TYPECHECK
2594 assert (STRINGP (str));
2596 return buffer_insert_string_1 (buf, pos, 0, str, 0,
2597 XSTRING_LENGTH (str),
2601 /* Insert the null-terminated string S (in external format). */
2604 buffer_insert_c_string_1 (struct buffer *buf, Bufpos pos, const char *s,
2607 /* This function can GC */
2608 const char *translated = GETTEXT (s);
2609 return buffer_insert_string_1 (buf, pos, (const Bufbyte *) translated, Qnil,
2610 0, strlen (translated), flags);
2614 buffer_insert_emacs_char_1 (struct buffer *buf, Bufpos pos, Emchar ch,
2617 /* This function can GC */
2618 Bufbyte str[MAX_EMCHAR_LEN];
2619 Bytecount len = set_charptr_emchar (str, ch);
2620 return buffer_insert_string_1 (buf, pos, str, Qnil, 0, len, flags);
2624 buffer_insert_c_char_1 (struct buffer *buf, Bufpos pos, char c,
2627 /* This function can GC */
2628 return buffer_insert_emacs_char_1 (buf, pos, (Emchar) (unsigned char) c,
2633 buffer_insert_from_buffer_1 (struct buffer *buf, Bufpos pos,
2634 struct buffer *buf2, Bufpos pos2,
2635 Charcount length, int flags)
2637 /* This function can GC */
2638 Lisp_Object str = make_string_from_buffer (buf2, pos2, length);
2639 return buffer_insert_string_1 (buf, pos, 0, str, 0,
2640 XSTRING_LENGTH (str), flags);
2644 /************************************************************************/
2645 /* Deletion of ranges */
2646 /************************************************************************/
2648 /* Delete characters in buffer from FROM up to (but not including) TO. */
2651 buffer_delete_range (struct buffer *buf, Bufpos from, Bufpos to, int flags)
2653 /* This function can GC */
2655 Bytind bi_from, bi_to;
2656 Bytecount bc_numdel;
2658 struct buffer *mbuf;
2659 Lisp_Object bufcons;
2661 /* Defensive steps just in case a buffer gets deleted and a calling
2662 function doesn't notice it. */
2663 if (!BUFFER_LIVE_P (buf))
2666 /* Make args be valid */
2667 if (from < BUF_BEGV (buf))
2668 from = BUF_BEGV (buf);
2669 if (to > BUF_ZV (buf))
2671 if ((numdel = to - from) <= 0)
2674 prepare_to_modify_buffer (buf, from, to, !(flags & INSDEL_NO_LOCKING));
2676 /* Defensive steps in case the before-change-functions fuck around */
2677 if (!BUFFER_LIVE_P (buf))
2678 /* Bad bad pre-change function. */
2681 /* Make args be valid again. prepare_to_modify_buffer() might have
2682 modified the buffer. */
2683 if (from < BUF_BEGV (buf))
2684 from = BUF_BEGV (buf);
2685 if (to > BUF_ZV (buf))
2687 if ((numdel = to - from) <= 0)
2690 /* Redisplay needs to know if a newline was in the deleted region.
2691 If we've already marked the changed region as having a deleted
2692 newline there is no use in performing the check. */
2693 if (!buf->changes->newline_was_deleted)
2695 scan_buffer (buf, '\n', from, to, 1, &shortage, 1);
2698 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2700 mbuf->changes->newline_was_deleted = 1;
2705 bi_from = bufpos_to_bytind (buf, from);
2706 bi_to = bufpos_to_bytind (buf, to);
2707 bc_numdel = bi_to - bi_from;
2709 delete_invalidate_line_number_cache (buf, from, to);
2711 if (to == BUF_Z (buf) &&
2712 bi_from > BI_BUF_GPT (buf))
2714 /* avoid moving the gap just to delete from the bottom. */
2716 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2718 record_delete (mbuf, from, numdel);
2721 MARK_BUFFERS_CHANGED;
2723 /* #### Point used to be modified here, but this causes problems
2724 with MULE, as point is used to calculate bytinds, and if the
2725 offset in bc_numdel causes point to move to a non first-byte
2726 location, causing some other function to throw an assertion
2727 in ASSERT_VALID_BYTIND. I've moved the code to right after
2728 the other movements and adjustments, but before the gap is
2729 moved. -- jh 970813 */
2731 /* Detach any extents that are completely within the range [FROM, TO],
2732 if the extents are detachable.
2734 This must come AFTER record_delete(), so that the appropriate
2735 extents will be present to be recorded, and BEFORE the gap
2736 size is increased, as otherwise we will be confused about
2737 where the extents end. */
2738 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2740 process_extents_for_deletion (make_buffer (mbuf), bi_from, bi_to, 0);
2743 /* Relocate all markers pointing into the new, larger gap to
2744 point at the end of the text before the gap. */
2745 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2747 adjust_markers (mbuf,
2748 (bi_to + BUF_GAP_SIZE (mbuf)),
2749 (bi_to + BUF_GAP_SIZE (mbuf)),
2753 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2755 /* Relocate any extent endpoints just like markers. */
2756 adjust_extents_for_deletion (make_buffer (mbuf), bi_from, bi_to,
2757 BUF_GAP_SIZE (mbuf), bc_numdel, 0);
2760 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2762 /* Relocate point as if it were a marker. */
2763 if (bi_from < BI_BUF_PT (mbuf))
2765 if (BI_BUF_PT (mbuf) < bi_to)
2766 JUST_SET_POINT (mbuf, from, bi_from);
2768 JUST_SET_POINT (mbuf, BUF_PT (mbuf) - numdel,
2769 BI_BUF_PT (mbuf) - bc_numdel);
2773 SET_BUF_END_GAP_SIZE (buf, BUF_END_GAP_SIZE (buf) + bc_numdel);
2775 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2777 SET_BOTH_BUF_ZV (mbuf, BUF_ZV (mbuf) - numdel,
2778 BI_BUF_ZV (mbuf) - bc_numdel);
2780 SET_BOTH_BUF_Z (buf, BUF_Z (buf) - numdel, BI_BUF_Z (buf) - bc_numdel);
2781 SET_GAP_SENTINEL (buf);
2785 /* Make sure the gap is somewhere in or next to what we are deleting. */
2786 if (bi_to < BI_BUF_GPT (buf))
2787 gap_left (buf, bi_to);
2788 if (bi_from > BI_BUF_GPT (buf))
2789 gap_right (buf, bi_from);
2791 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2793 record_delete (mbuf, from, numdel);
2796 MARK_BUFFERS_CHANGED;
2798 /* #### Point used to be modified here, but this causes problems
2799 with MULE, as point is used to calculate bytinds, and if the
2800 offset in bc_numdel causes point to move to a non first-byte
2801 location, causing some other function to throw an assertion
2802 in ASSERT_VALID_BYTIND. I've moved the code to right after
2803 the other movements and adjustments, but before the gap is
2804 moved. -- jh 970813 */
2806 /* Detach any extents that are completely within the range [FROM, TO],
2807 if the extents are detachable.
2809 This must come AFTER record_delete(), so that the appropriate extents
2810 will be present to be recorded, and BEFORE the gap size is increased,
2811 as otherwise we will be confused about where the extents end. */
2812 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2814 process_extents_for_deletion (make_buffer (mbuf), bi_from, bi_to, 0);
2817 /* Relocate all markers pointing into the new, larger gap to
2818 point at the end of the text before the gap. */
2819 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2821 adjust_markers (mbuf,
2822 (bi_to + BUF_GAP_SIZE (mbuf)),
2823 (bi_to + BUF_GAP_SIZE (mbuf)),
2824 (- bc_numdel - BUF_GAP_SIZE (mbuf)));
2827 /* Relocate any extent endpoints just like markers. */
2828 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2830 adjust_extents_for_deletion (make_buffer (mbuf), bi_from, bi_to,
2831 BUF_GAP_SIZE (mbuf),
2832 bc_numdel, BUF_GAP_SIZE (mbuf));
2835 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2837 /* Relocate point as if it were a marker. */
2838 if (bi_from < BI_BUF_PT (mbuf))
2840 if (BI_BUF_PT (mbuf) < bi_to)
2841 JUST_SET_POINT (mbuf, from, bi_from);
2843 JUST_SET_POINT (mbuf, BUF_PT (mbuf) - numdel,
2844 BI_BUF_PT (mbuf) - bc_numdel);
2848 SET_BUF_GAP_SIZE (buf, BUF_GAP_SIZE (buf) + bc_numdel);
2849 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2851 SET_BOTH_BUF_ZV (mbuf, BUF_ZV (mbuf) - numdel,
2852 BI_BUF_ZV (mbuf) - bc_numdel);
2854 SET_BOTH_BUF_Z (buf, BUF_Z (buf) - numdel, BI_BUF_Z (buf) - bc_numdel);
2855 SET_BI_BUF_GPT (buf, bi_from);
2856 SET_GAP_SENTINEL (buf);
2860 buffer_mule_signal_deleted_region (buf, from, to, bi_from, bi_to);
2863 #ifdef ERROR_CHECK_EXTENTS
2864 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2866 sledgehammer_extent_check (make_buffer (mbuf));
2870 signal_after_change (buf, from, to, from);
2874 /************************************************************************/
2875 /* Replacement of characters */
2876 /************************************************************************/
2878 /* Replace the character at POS in buffer B with CH. */
2881 buffer_replace_char (struct buffer *buf, Bufpos pos, Emchar ch,
2882 int not_real_change, int force_lock_check)
2884 /* This function can GC */
2885 Bufbyte curstr[MAX_EMCHAR_LEN];
2886 Bufbyte newstr[MAX_EMCHAR_LEN];
2887 Bytecount curlen, newlen;
2889 /* Defensive steps just in case a buffer gets deleted and a calling
2890 function doesn't notice it. */
2891 if (!BUFFER_LIVE_P (buf))
2894 curlen = BUF_CHARPTR_COPY_CHAR (buf, pos, curstr);
2895 newlen = set_charptr_emchar (newstr, ch);
2897 if (curlen == newlen)
2899 struct buffer *mbuf;
2900 Lisp_Object bufcons;
2902 /* then we can just replace the text. */
2903 prepare_to_modify_buffer (buf, pos, pos + 1,
2904 !not_real_change || force_lock_check);
2905 /* Defensive steps in case the before-change-functions fuck around */
2906 if (!BUFFER_LIVE_P (buf))
2907 /* Bad bad pre-change function. */
2910 /* Make args be valid again. prepare_to_modify_buffer() might have
2911 modified the buffer. */
2912 if (pos < BUF_BEGV (buf))
2913 pos = BUF_BEGV (buf);
2914 if (pos >= BUF_ZV (buf))
2915 pos = BUF_ZV (buf) - 1;
2916 if (pos < BUF_BEGV (buf))
2917 /* no more characters in buffer! */
2920 if (BUF_FETCH_CHAR (buf, pos) == '\n')
2922 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2924 mbuf->changes->newline_was_deleted = 1;
2927 MARK_BUFFERS_CHANGED;
2928 if (!not_real_change)
2930 MAP_INDIRECT_BUFFERS (buf, mbuf, bufcons)
2932 record_change (mbuf, pos, 1);
2936 memcpy (BUF_BYTE_ADDRESS (buf, pos), newstr, newlen);
2938 signal_after_change (buf, pos, pos + 1, pos + 1);
2940 /* We do not have to adjust the Mule data; we just replaced a
2941 character with another of the same number of bytes. */
2946 * Must implement as deletion followed by insertion.
2948 * Make a note to move point forward later in the one situation
2949 * where it is needed, a delete/insert one position behind
2950 * point. Point will drift backward by one position and stay
2953 int movepoint = (pos == BUF_PT (buf) - 1);
2955 buffer_delete_range (buf, pos, pos + 1, 0);
2956 /* Defensive steps in case the before-change-functions fuck around */
2957 if (!BUFFER_LIVE_P (buf))
2958 /* Bad bad pre-change function. */
2961 /* Make args be valid again. prepare_to_modify_buffer() might have
2962 modified the buffer. */
2963 if (pos < BUF_BEGV (buf))
2964 pos = BUF_BEGV (buf);
2965 if (pos >= BUF_ZV (buf))
2966 pos = BUF_ZV (buf) - 1;
2967 if (pos < BUF_BEGV (buf))
2968 /* no more characters in buffer! */
2971 * -1 as the pos argument means to move point forward with the
2972 * insertion, which we must do if the deletion moved point
2973 * backward so that it now equals the insertion point.
2975 buffer_insert_string_1 (buf, (movepoint ? -1 : pos),
2976 newstr, Qnil, 0, newlen, 0);
2981 /************************************************************************/
2982 /* Other functions */
2983 /************************************************************************/
2985 /* Make a string from a buffer. This needs to take into account the gap,
2986 and add any necessary extents from the buffer. */
2989 make_string_from_buffer_1 (struct buffer *buf, Bufpos pos, Charcount length,
2992 /* This function can GC */
2993 Bytind bi_ind = bufpos_to_bytind (buf, pos);
2994 Bytecount bi_len = bufpos_to_bytind (buf, pos + length) - bi_ind;
2995 Lisp_Object val = make_uninit_string (bi_len);
2997 struct gcpro gcpro1;
3001 add_string_extents (val, buf, bi_ind, bi_len);
3004 Bytecount len1 = BI_BUF_GPT (buf) - bi_ind;
3005 Bufbyte *start1 = BI_BUF_BYTE_ADDRESS (buf, bi_ind);
3006 Bufbyte *dest = XSTRING_DATA (val);
3010 /* Completely after gap */
3011 memcpy (dest, start1, bi_len);
3013 else if (bi_len <= len1)
3015 /* Completely before gap */
3016 memcpy (dest, start1, bi_len);
3021 Bytind pos2 = bi_ind + len1;
3022 Bufbyte *start2 = BI_BUF_BYTE_ADDRESS (buf, pos2);
3024 memcpy (dest, start1, len1);
3025 memcpy (dest + len1, start2, bi_len - len1);
3034 make_string_from_buffer (struct buffer *buf, Bufpos pos, Charcount length)
3036 return make_string_from_buffer_1 (buf, pos, length, 0);
3040 make_string_from_buffer_no_extents (struct buffer *buf, Bufpos pos,
3043 return make_string_from_buffer_1 (buf, pos, length, 1);
3047 barf_if_buffer_read_only (struct buffer *buf, Bufpos from, Bufpos to)
3052 XSETBUFFER (buffer, buf);
3054 iro = (buf == current_buffer ? Vinhibit_read_only :
3055 symbol_value_in_buffer (Qinhibit_read_only, buffer));
3058 if (NILP (iro) && !NILP (buf->read_only))
3060 Fsignal (Qbuffer_read_only, (list1 (buffer)));
3067 verify_extent_modification (buffer,
3068 bufpos_to_bytind (buf, from),
3069 bufpos_to_bytind (buf, to),
3075 find_charsets_in_bufbyte_string (unsigned char *charsets, const Bufbyte *str,
3079 /* Telescope this. */
3082 const Bufbyte *strend = str + len;
3083 memset (charsets, 0, NUM_LEADING_BYTES);
3085 /* #### SJT doesn't like this. */
3088 charsets[XCHARSET_LEADING_BYTE (Vcharset_ascii) - 128] = 1;
3092 while (str < strend)
3094 charsets[CHAR_LEADING_BYTE (charptr_emchar (str)) - 128] = 1;
3101 find_charsets_in_emchar_string (unsigned char *charsets, const Emchar *str,
3105 /* Telescope this. */
3110 memset (charsets, 0, NUM_LEADING_BYTES);
3112 /* #### SJT doesn't like this. */
3115 charsets[XCHARSET_LEADING_BYTE (Vcharset_ascii) - 128] = 1;
3119 for (i = 0; i < len; i++)
3121 charsets[CHAR_LEADING_BYTE (str[i]) - 128] = 1;
3127 bufbyte_string_displayed_columns (const Bufbyte *str, Bytecount len)
3130 const Bufbyte *end = str + len;
3135 Emchar ch = charptr_emchar (str);
3136 cols += XCHARSET_COLUMNS (CHAR_CHARSET (ch));
3147 emchar_string_displayed_columns (const Emchar *str, Charcount len)
3153 for (i = 0; i < len; i++)
3154 cols += XCHARSET_COLUMNS (CHAR_CHARSET (str[i]));
3157 #else /* not MULE */
3162 /* NOTE: Does not reset the Dynarr. */
3165 convert_bufbyte_string_into_emchar_dynarr (const Bufbyte *str, Bytecount len,
3168 const Bufbyte *strend = str + len;
3170 while (str < strend)
3172 Emchar ch = charptr_emchar (str);
3173 Dynarr_add (dyn, ch);
3179 convert_bufbyte_string_into_emchar_string (const Bufbyte *str, Bytecount len,
3182 const Bufbyte *strend = str + len;
3183 Charcount newlen = 0;
3184 while (str < strend)
3186 Emchar ch = charptr_emchar (str);
3193 /* Convert an array of Emchars into the equivalent string representation.
3194 Store into the given Bufbyte dynarr. Does not reset the dynarr.
3195 Does not add a terminating zero. */
3198 convert_emchar_string_into_bufbyte_dynarr (Emchar *arr, int nels,
3199 Bufbyte_dynarr *dyn)
3201 Bufbyte str[MAX_EMCHAR_LEN];
3204 for (i = 0; i < nels; i++)
3206 Bytecount len = set_charptr_emchar (str, arr[i]);
3207 Dynarr_add_many (dyn, str, len);
3211 /* Convert an array of Emchars into the equivalent string representation.
3212 Malloc the space needed for this and return it. If LEN_OUT is not a
3213 NULL pointer, store into LEN_OUT the number of Bufbytes in the
3214 malloc()ed string. Note that the actual number of Bufbytes allocated
3215 is one more than this: the returned string is zero-terminated. */
3218 convert_emchar_string_into_malloced_string (Emchar *arr, int nels,
3221 /* Damn zero-termination. */
3222 Bufbyte *str = (Bufbyte *) alloca (nels * MAX_EMCHAR_LEN + 1);
3223 Bufbyte *strorig = str;
3228 for (i = 0; i < nels; i++)
3229 str += set_charptr_emchar (str, arr[i]);
3231 len = str - strorig;
3232 str = (Bufbyte *) xmalloc (1 + len);
3233 memcpy (str, strorig, 1 + len);
3240 /************************************************************************/
3241 /* initialization */
3242 /************************************************************************/
3245 reinit_vars_of_insdel (void)
3249 inside_change_hook = 0;
3250 in_first_change = 0;
3252 for (i = 0; i <= MAX_BYTIND_GAP_SIZE_3; i++)
3253 three_to_one_table[i] = i / 3;
3257 vars_of_insdel (void)
3259 reinit_vars_of_insdel ();
3263 init_buffer_text (struct buffer *b)
3265 if (!b->base_buffer)
3267 SET_BUF_GAP_SIZE (b, 20);
3268 BUFFER_ALLOC (b->text->beg, BUF_GAP_SIZE (b) + BUF_END_SENTINEL_SIZE);
3269 if (! BUF_BEG_ADDR (b))
3272 SET_BUF_END_GAP_SIZE (b, 0);
3273 SET_BI_BUF_GPT (b, 1);
3274 SET_BOTH_BUF_Z (b, 1, 1);
3275 SET_GAP_SENTINEL (b);
3276 SET_END_SENTINEL (b);
3281 b->text->mule_bufmin = b->text->mule_bufmax = 1;
3282 b->text->mule_bytmin = b->text->mule_bytmax = 1;
3283 b->text->mule_shifter = 0;
3284 b->text->mule_three_p = 0;
3286 for (i = 0; i < 16; i++)
3288 b->text->mule_bufpos_cache[i] = 1;
3289 b->text->mule_bytind_cache[i] = 1;
3293 b->text->line_number_cache = Qnil;
3296 BUF_SAVE_MODIFF (b) = 1;
3298 JUST_SET_POINT (b, 1, 1);
3299 SET_BOTH_BUF_BEGV (b, 1, 1);
3300 SET_BOTH_BUF_ZV (b, 1, 1);
3302 b->text->changes = xnew_and_zero (struct buffer_text_change_data);
3306 JUST_SET_POINT (b, BUF_PT (b->base_buffer), BI_BUF_PT (b->base_buffer));
3307 SET_BOTH_BUF_BEGV (b, BUF_BEGV (b->base_buffer),
3308 BI_BUF_BEGV (b->base_buffer));
3309 SET_BOTH_BUF_ZV (b, BUF_ZV (b->base_buffer),
3310 BI_BUF_ZV (b->base_buffer));
3313 b->changes = xnew_and_zero (struct each_buffer_change_data);
3314 BUF_FACECHANGE (b) = 1;
3316 #ifdef REGION_CACHE_NEEDS_WORK
3317 b->newline_cache = 0;
3318 b->width_run_cache = 0;
3319 b->width_table = Qnil;
3324 uninit_buffer_text (struct buffer *b)
3326 if (!b->base_buffer)
3328 BUFFER_FREE (b->text->beg);
3329 xfree (b->text->changes);
3333 #ifdef REGION_CACHE_NEEDS_WORK
3334 if (b->newline_cache)
3336 free_region_cache (b->newline_cache);
3337 b->newline_cache = 0;
3339 if (b->width_run_cache)
3341 free_region_cache (b->width_run_cache);
3342 b->width_run_cache = 0;
3344 b->width_table = Qnil;