lisp/format.el

   1 ;;; format.el --- read and save files in multiple formats
   2
   3 ;; Copyright (c) 1994, 1995, 1997, 1999 Free Software Foundation
   4
   5 ;; Author: Boris Goldowsky <boris@gnu.org>
   6
   7 ;; This file is part of GNU Emacs.
   8
   9 ;; GNU Emacs is free software; you can redistribute it and/or modify
  10 ;; it under the terms of the GNU General Public License as published by
  11 ;; the Free Software Foundation; either version 2, or (at your option)
  12 ;; any later version.
  13
  14 ;; GNU Emacs is distributed in the hope that it will be useful,
  15 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
  16 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17 ;; GNU General Public License for more details.
  18
  19 ;; You should have received a copy of the GNU General Public License
  20 ;; along with GNU Emacs; see the file COPYING.  If not, write to the
  21 ;; Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  22 ;; Boston, MA 02111-1307, USA.
  23
  24 ;;; Commentary:
  25
  26 ;; This file defines a unified mechanism for saving & loading files stored
  27 ;; in different formats.  `format-alist' contains information that directs
  28 ;; Emacs to call an encoding or decoding function when reading or writing
  29 ;; files that match certain conditions.
  30 ;;
  31 ;; When a file is visited, its format is determined by matching the
  32 ;; beginning of the file against regular expressions stored in
  33 ;; `format-alist'.  If this fails, you can manually translate the buffer
  34 ;; using `format-decode-buffer'.  In either case, the formats used are
  35 ;; listed in the variable `buffer-file-format', and become the default
  36 ;; format for saving the buffer.  To save a buffer in a different format,
  37 ;; change this variable, or use `format-write-file'.
  38 ;;
  39 ;; Auto-save files are normally created in the same format as the visited
  40 ;; file, but the variable `auto-save-file-format' can be set to a
  41 ;; particularly fast or otherwise preferred format to be used for
  42 ;; auto-saving (or nil to do no encoding on auto-save files, but then you
  43 ;; risk losing any text-properties in the buffer).
  44 ;;
  45 ;; You can manually translate a buffer into or out of a particular format
  46 ;; with the functions `format-encode-buffer' and `format-decode-buffer'.
  47 ;; To translate just the region use the functions `format-encode-region'
  48 ;; and `format-decode-region'.
  49 ;;
  50 ;; You can define a new format by writing the encoding and decoding
  51 ;; functions, and adding an entry to `format-alist'.  See enriched.el for
  52 ;; an example of how to implement a file format.  There are various
  53 ;; functions defined in this file that may be useful for writing the
  54 ;; encoding and decoding functions:
  55 ;;  * `format-annotate-region' and `format-deannotate-region' allow a
  56 ;;     single alist of information to be used for encoding and decoding.
  57 ;;     The alist defines a correspondence between strings in the file
  58 ;;     ("annotations") and text-properties in the buffer.
  59 ;;  * `format-replace-strings' is similarly useful for doing simple
  60 ;;     string->string translations in a reversible manner.
  61
  62 ;;; Code:
  63
  64 (put 'buffer-file-format 'permanent-local t)
  65
  66 (defvar format-alist
  67   '((text/enriched "Extended MIME text/enriched format."
  68                    "Content-[Tt]ype:[ \t]*text/enriched"
  69                    enriched-decode enriched-encode t enriched-mode)
  70     (plain "ISO 8859-1 standard format, no text properties."
  71            ;; Plain only exists so that there is an obvious neutral choice in
  72            ;; the completion list.
  73            nil nil nil nil nil)
  74     (ibm   "IBM Code Page 850 (DOS)"
  75            nil                          ; The original "1\\(^\\)" is obscure.
  76            "recode -f ibm-pc:latin1" "recode -f latin1:ibm-pc" t nil)
  77     (mac   "Apple Macintosh"
  78            nil
  79            "recode -f mac:latin1" "recode -f latin1:mac" t nil)
  80     (hp    "HP Roman8"
  81            nil
  82            "recode -f roman8:latin1" "recode -f latin1:roman8" t nil)
  83     (TeX   "TeX (encoding)"
  84            nil
  85            iso-tex2iso iso-iso2tex t nil)
  86     (gtex  "German TeX (encoding)"
  87            nil
  88            iso-gtex2iso iso-iso2gtex t nil)
  89     (html  "HTML/SGML \"ISO 8879:1986//ENTITIES Added Latin 1//EN\" (encoding)"
  90            nil
  91            iso-sgml2iso iso-iso2sgml t nil)
  92     (rot13 "rot13"
  93            nil
  94            "tr a-mn-z n-za-m" "tr a-mn-z n-za-m" t nil)
  95     (duden "Duden Ersatzdarstellung"
  96            nil
  97            "diac" iso-iso2duden t nil)
  98     (de646 "German ASCII (ISO 646)"
  99            nil
 100            "recode -f iso646-ge:latin1" "recode -f latin1:iso646-ge" t nil)
 101     (denet "net German"
 102            nil
 103            iso-german iso-cvt-read-only t nil)
 104     (esnet "net Spanish"
 105            nil
 106            iso-spanish iso-cvt-read-only t nil))
 107   "List of information about understood file formats.
 108 Elements are of the form \(NAME DOC-STR REGEXP FROM-FN TO-FN MODIFY MODE-FN).
 109
 110 NAME    is a symbol, which is stored in `buffer-file-format'.
 111
 112 DOC-STR should be a single line providing more information about the
 113         format.  It is currently unused, but in the future will be shown to
 114         the user if they ask for more information.
 115
 116 REGEXP  is a regular expression to match against the beginning of the file;
 117         it should match only files in that format.  Use nil to avoid
 118         matching at all for formats for which this isn't appropriate to
 119         require explicit encoding/decoding.
 120
 121 FROM-FN is called to decode files in that format; it gets two args, BEGIN
 122         and END, and can make any modifications it likes, returning the new
 123         end.  It must make sure that the beginning of the file no longer
 124         matches REGEXP, or else it will get called again.
 125         Alternatively, FROM-FN can be a string, which specifies a shell command
 126         (including options) to be used as a filter to perform the conversion.
 127
 128 TO-FN   is called to encode a region into that format; it is passed three
 129         arguments: BEGIN, END, and BUFFER.  BUFFER is the original buffer that
 130         the data being written came from, which the function could use, for
 131         example, to find the values of local variables.  TO-FN should either
 132         return a list of annotations like `write-region-annotate-functions',
 133         or modify the region and return the new end.
 134         Alternatively, TO-FN can be a string, which specifies a shell command
 135         (including options) to be used as a filter to perform the conversion.
 136
 137 MODIFY, if non-nil, means the TO-FN wants to modify the region.  If nil,
 138         TO-FN will not make any changes but will instead return a list of
 139         annotations.
 140
 141 MODE-FN, if specified, is called when visiting a file with that format.")
 142
 143 ;;; Basic Functions (called from Lisp)
 144
 145 (defun format-encode-run-method (method from to &optional buffer)
 146   "Translate using function or shell script METHOD the text from FROM to TO.
 147 If METHOD is a string, it is a shell command;
 148 otherwise, it should be a Lisp function.
 149 BUFFER should be the buffer that the output originally came from."
 150   (if (stringp method)
 151       (let ((error-buff (get-buffer-create "*Format Errors*"))
 152             (coding-system-for-read 'no-conversion)
 153             format-alist)
 154         (with-current-buffer error-buff
 155           (widen)
 156           (erase-buffer))
 157         (if (and (zerop (save-window-excursion
 158                           (shell-command-on-region from to method t t
 159                                                    error-buff)))
 160                  ;; gzip gives zero exit status with bad args, for instance.
 161                  (zerop (with-current-buffer error-buff
 162                           (buffer-size))))
 163             (bury-buffer error-buff)
 164           (switch-to-buffer-other-window error-buff)
 165           (error "Format encoding failed")))
 166     (funcall method from to buffer)))
 167
 168 (defun format-decode-run-method (method from to &optional buffer)
 169   "Decode using function or shell script METHOD the text from FROM to TO.
 170 If METHOD is a string, it is a shell command; otherwise, it should be
 171 a Lisp function.  Decoding is done for the given BUFFER."
 172   (if (stringp method)
 173       (let ((error-buff (get-buffer-create "*Format Errors*"))
 174             (coding-system-for-write 'no-conversion)
 175             format-alist)
 176         (with-current-buffer error-buff
 177           (widen)
 178           (erase-buffer))
 179         ;; We should perhaps go via a temporary buffer and copy it
 180         ;; back, in case of errors.
 181         (if (and (zerop (save-window-excursion
 182                           (shell-command-on-region (point-min) (point-max)
 183                                                    method t t
 184                                                    error-buff)))
 185                  ;; gzip gives zero exit status with bad args, for instance.
 186                  (zerop (with-current-buffer error-buff
 187                           (buffer-size))))
 188             (bury-buffer error-buff)
 189           (switch-to-buffer-other-window error-buff)
 190           (error "Format decoding failed"))
 191         (point))
 192     (funcall method from to)))
 193
 194 (defun format-annotate-function (format from to orig-buf format-count)
 195   "Return annotations for writing region as FORMAT.
 196 FORMAT is a symbol naming one of the formats defined in `format-alist',
 197 it must be a single symbol, not a list like `buffer-file-format'.
 198 FROM and TO delimit the region to be operated on in the current buffer.
 199 ORIG-BUF is the original buffer that the data came from.
 200
 201 FORMAT-COUNT is an integer specifying how many times this function has
 202 been called in the process of decoding ORIG-BUF.
 203
 204 This function works like a function on `write-region-annotate-functions':
 205 it either returns a list of annotations, or returns with a different buffer
 206 current, which contains the modified text to write.  In the latter case,
 207 this function's value is nil.
 208
 209 For most purposes, consider using `format-encode-region' instead."
 210   ;; This function is called by write-region (actually
 211   ;; build_annotations) for each element of buffer-file-format.
 212   (let* ((info (assq format format-alist))
 213          (to-fn  (nth 4 info))
 214          (modify (nth 5 info)))
 215     (if to-fn
 216         (if modify
 217             ;; To-function wants to modify region.  Copy to safe place.
 218             (let ((copy-buf (get-buffer-create (format " *Format Temp %d*"
 219                                                        format-count))))
 220               (copy-to-buffer copy-buf from to)
 221               (set-buffer copy-buf)
 222               (format-insert-annotations write-region-annotations-so-far from)
 223               (format-encode-run-method to-fn (point-min) (point-max) orig-buf)
 224               nil)
 225           ;; Otherwise just call function, it will return annotations.
 226           (funcall to-fn from to orig-buf)))))
 227
 228 (defun format-decode (format length &optional visit-flag)
 229   ;; This function is called by insert-file-contents whenever a file is read.
 230   "Decode text from any known FORMAT.
 231 FORMAT is a symbol appearing in `format-alist' or a list of such symbols,
 232 or nil, in which case this function tries to guess the format of the data by
 233 matching against the regular expressions in `format-alist'.  After a match is
 234 found and the region decoded, the alist is searched again from the beginning
 235 for another match.
 236
 237 Second arg LENGTH is the number of characters following point to operate on.
 238 If optional third arg VISIT-FLAG is true, set `buffer-file-format'
 239 to the list of formats used, and call any mode functions defined for those
 240 formats.
 241
 242 Returns the new length of the decoded region.
 243
 244 For most purposes, consider using `format-decode-region' instead."
 245   (let ((mod (buffer-modified-p))
 246         (begin (point))
 247         (end (+ (point) length)))
 248     (unwind-protect
 249         (progn
 250           ;; Don't record undo information for the decoding.
 251
 252           (if (null format)
 253               ;; Figure out which format it is in, remember list in `format'.
 254               (let ((try format-alist))
 255                 (while try
 256                   (let* ((f (car try))
 257                          (regexp (nth 2 f))
 258                          (p (point)))
 259                     (if (and regexp (looking-at regexp)
 260                              (< (match-end 0) (+ begin length)))
 261                         (progn
 262                           (setq format (cons (car f) format))
 263                           ;; Decode it
 264                           (if (nth 3 f)
 265                               (setq end (format-decode-run-method (nth 3 f) begin end)))
 266                           ;; Call visit function if required
 267                           (if (and visit-flag (nth 6 f)) (funcall (nth 6 f) 1))
 268                           ;; Safeguard against either of the functions changing pt.
 269                           (goto-char p)
 270                           ;; Rewind list to look for another format
 271                           (setq try format-alist))
 272                       (setq try (cdr try))))))
 273             ;; Deal with given format(s)
 274             (or (listp format) (setq format (list format)))
 275             (let ((do format) f)
 276               (while do
 277                 (or (setq f (assq (car do) format-alist))
 278                     (error "Unknown format" (car do)))
 279                 ;; Decode:
 280                 (if (nth 3 f)
 281                     (setq end (format-decode-run-method (nth 3 f) begin end)))
 282                 ;; Call visit function if required
 283                 (if (and visit-flag (nth 6 f)) (funcall (nth 6 f) 1))
 284                 (setq do (cdr do)))))
 285           (if visit-flag
 286               (setq buffer-file-format format)))
 287
 288       (set-buffer-modified-p mod))
 289
 290       ;; Return new length of region
 291     (- end begin)))
 292
 293 ;;;
 294 ;;; Interactive functions & entry points
 295 ;;;
 296
 297 (defun format-decode-buffer (&optional format)
 298   "Translate the buffer from some FORMAT.
 299 If the format is not specified, this function attempts to guess.
 300 `buffer-file-format' is set to the format used, and any mode-functions
 301 for the format are called."
 302   (interactive
 303    (list (format-read "Translate buffer from format (default: guess): ")))
 304   (save-excursion
 305     (goto-char (point-min))
 306     (format-decode format (buffer-size) t)))
 307
 308 (defun format-decode-region (from to &optional format)
 309   "Decode the region from some format.
 310 Arg FORMAT is optional; if omitted the format will be determined by looking
 311 for identifying regular expressions at the beginning of the region."
 312   (interactive
 313    (list (region-beginning) (region-end)
 314          (format-read "Translate region from format (default: guess): ")))
 315   (save-excursion
 316     (goto-char from)
 317     (format-decode format (- to from) nil)))
 318
 319 (defun format-encode-buffer (&optional format)
 320   "Translate the buffer into FORMAT.
 321 FORMAT defaults to `buffer-file-format'.  It is a symbol naming one of the
 322 formats defined in `format-alist', or a list of such symbols."
 323   (interactive
 324    (list (format-read (format "Translate buffer to format (default %s): "
 325                               buffer-file-format))))
 326   (format-encode-region (point-min) (point-max) format))
 327
 328 (defun format-encode-region (beg end &optional format)
 329   "Translate the region into some FORMAT.
 330 FORMAT defaults to `buffer-file-format', it is a symbol naming
 331 one of the formats defined in `format-alist', or a list of such symbols."
 332   (interactive
 333    (list (region-beginning) (region-end)
 334          (format-read (format "Translate region to format (default %s): "
 335                               buffer-file-format))))
 336   (if (null format)    (setq format buffer-file-format))
 337   (if (symbolp format) (setq format (list format)))
 338   (save-excursion
 339     (goto-char end)
 340     (let ((cur-buf (current-buffer))
 341           (end (point-marker)))
 342       (while format
 343         (let* ((info (assq (car format) format-alist))
 344                (to-fn  (nth 4 info))
 345                (modify (nth 5 info))
 346                result)
 347           (if to-fn
 348               (if modify
 349                   (setq end (format-encode-run-method to-fn beg end
 350                                                       (current-buffer)))
 351                 (format-insert-annotations
 352                  (funcall to-fn beg end (current-buffer)))))
 353           (setq format (cdr format)))))))
 354
 355 (defun format-write-file (filename format)
 356   "Write current buffer into file FILENAME using some FORMAT.
 357 Makes buffer visit that file and sets the format as the default for future
 358 saves.  If the buffer is already visiting a file, you can specify a directory
 359 name as FILENAME, to write a file of the same old name in that directory."
 360   (interactive
 361    ;; Same interactive spec as write-file, plus format question.
 362    (let* ((file (if buffer-file-name
 363                     (read-file-name "Write file: "
 364                                     nil nil nil nil)
 365                   (read-file-name "Write file: "
 366                                   (cdr (assq 'default-directory
 367                                              (buffer-local-variables)))
 368                                   nil nil (buffer-name))))
 369           (fmt (format-read (format "Write file `%s' in format: "
 370                                     (file-name-nondirectory file)))))
 371      (list file fmt)))
 372   (setq buffer-file-format format)
 373   (write-file filename))
 374
 375 (defun format-find-file (filename format)
 376   "Find the file FILENAME using data format FORMAT.
 377 If FORMAT is nil then do not do any format conversion."
 378   (interactive
 379    ;; Same interactive spec as write-file, plus format question.
 380    (let* ((file (read-file-name "Find file: "))
 381           (fmt (format-read (format "Read file `%s' in format: "
 382                                     (file-name-nondirectory file)))))
 383      (list file fmt)))
 384   (let ((format-alist nil))
 385      (find-file filename))
 386   (if format
 387       (format-decode-buffer format)))
 388
 389 (defun format-insert-file (filename format &optional beg end)
 390   "Insert the contents of file FILENAME using data format FORMAT.
 391 If FORMAT is nil then do not do any format conversion.
 392 The optional third and fourth arguments BEG and END specify
 393 the part of the file to read.
 394
 395 The return value is like the value of `insert-file-contents':
 396 a list (ABSOLUTE-FILE-NAME . SIZE)."
 397   (interactive
 398    ;; Same interactive spec as write-file, plus format question.
 399    (let* ((file (read-file-name "Find file: "))
 400           (fmt (format-read (format "Read file `%s' in format: "
 401                                     (file-name-nondirectory file)))))
 402      (list file fmt)))
 403   (let (value size)
 404     (let ((format-alist nil))
 405       (setq value (insert-file-contents filename nil beg end))
 406       (setq size (nth 1 value)))
 407     (if format
 408         (setq size (format-decode format size)
 409               value (cons (car value) size)))
 410     value))
 411
 412 (defun format-read (&optional prompt)
 413   "Read and return the name of a format.
 414 Return value is a list, like `buffer-file-format'; it may be nil.
 415 Formats are defined in `format-alist'.  Optional arg is the PROMPT to use."
 416   (let* ((table (mapcar (lambda (x) (list (symbol-name (car x))))
 417                         format-alist))
 418          (ans (completing-read (or prompt "Format: ") table nil t)))
 419     (if (not (equal "" ans)) (list (intern ans)))))
 420
 421
 422 ;;;
 423 ;;; Below are some functions that may be useful in writing encoding and
 424 ;;; decoding functions for use in format-alist.
 425 ;;;
 426
 427 (defun format-replace-strings (alist &optional reverse beg end)
 428   "Do multiple replacements on the buffer.
 429 ALIST is a list of (from . to) pairs, which should be proper arguments to
 430 `search-forward' and `replace-match' respectively.
 431 Optional 2nd arg REVERSE, if non-nil, means the pairs are (to . from), so that
 432 you can use the same list in both directions if it contains only literal
 433 strings.
 434 Optional args BEG and END specify a region of the buffer on which to operate."
 435   (save-excursion
 436     (save-restriction
 437       (or beg (setq beg (point-min)))
 438       (if end (narrow-to-region (point-min) end))
 439       (while alist
 440         (let ((from (if reverse (cdr (car alist)) (car (car alist))))
 441               (to   (if reverse (car (car alist)) (cdr (car alist)))))
 442           (goto-char beg)
 443           (while (search-forward from nil t)
 444             (goto-char (match-beginning 0))
 445             (insert to)
 446             (set-text-properties (- (point) (length to)) (point)
 447                                  (text-properties-at (point)))
 448             (delete-region (point) (+ (point) (- (match-end 0)
 449                                                  (match-beginning 0)))))
 450           (setq alist (cdr alist)))))))
 451
 452 ;;; Some list-manipulation functions that we need.
 453
 454 (defun format-delq-cons (cons list)
 455   "Remove the given CONS from LIST by side effect and return the new LIST.
 456 Since CONS could be the first element of LIST, write
 457 `\(setq foo \(format-delq-cons element foo))' to be sure of changing
 458 the value of `foo'."
 459   (if (eq cons list)
 460       (cdr list)
 461     (let ((p list))
 462       (while (not (eq (cdr p) cons))
 463         (if (null p) (error "format-delq-cons: not an element."))
 464         (setq p (cdr p)))
 465       ;; Now (cdr p) is the cons to delete
 466       (setcdr p (cdr cons))
 467       list)))
 468
 469 (defun format-make-relatively-unique (a b)
 470   "Delete common elements of lists A and B, return as pair.
 471 Compares using `equal'."
 472   (let* ((acopy (copy-sequence a))
 473          (bcopy (copy-sequence b))
 474          (tail acopy))
 475     (while tail
 476       (let ((dup (member (car tail) bcopy))
 477             (next (cdr tail)))
 478         (if dup (setq acopy (format-delq-cons tail acopy)
 479                       bcopy (format-delq-cons dup  bcopy)))
 480         (setq tail next)))
 481     (cons acopy bcopy)))
 482
 483 (defun format-common-tail (a b)
 484   "Given two lists that have a common tail, return it.
 485 Compares with `equal', and returns the part of A that is equal to the
 486 equivalent part of B.  If even the last items of the two are not equal,
 487 returns nil."
 488   (let ((la (length a))
 489         (lb (length b)))
 490     ;; Make sure they are the same length
 491     (if (> la lb)
 492         (setq a (nthcdr (- la lb) a))
 493       (setq b (nthcdr (- lb la) b))))
 494   (while (not (equal a b))
 495     (setq a (cdr a)
 496           b (cdr b)))
 497   a)
 498
 499 (defun format-proper-list-p (list)
 500   "Return t if LIST is a proper list.
 501 A proper list is a list ending with a nil cdr, not with an atom "
 502   (when (listp list)
 503     (while (consp list)
 504       (setq list (cdr list)))
 505     (null list)))
 506
 507 (defun format-reorder (items order)
 508   "Arrange ITEMS to following partial ORDER.
 509 Elements of ITEMS equal to elements of ORDER will be rearranged to follow the
 510 ORDER.  Unmatched items will go last."
 511   (if order
 512       (let ((item (member (car order) items)))
 513         (if item
 514             (cons (car item)
 515                   (format-reorder (format-delq-cons item items)
 516                            (cdr order)))
 517           (format-reorder items (cdr order))))
 518     items))
 519
 520 (put 'face 'format-list-valued t)       ; These text-properties take values
 521 (put 'unknown 'format-list-valued t)    ; that are lists, the elements of which
 522                                         ; should be considered separately.
 523                                         ; See format-deannotate-region and
 524                                         ; format-annotate-region.
 525
 526 ;; This text property has list values, but they are treated atomically.
 527
 528 (put 'display 'format-list-atomic-p t)
 529
 530 ;;;
 531 ;;; Decoding
 532 ;;;
 533
 534 (defun format-deannotate-region (from to translations next-fn)
 535   "Translate annotations in the region into text properties.
 536 This sets text properties between FROM to TO as directed by the
 537 TRANSLATIONS and NEXT-FN arguments.
 538
 539 NEXT-FN is a function that searches forward from point for an annotation.
 540 It should return a list of 4 elements: \(BEGIN END NAME POSITIVE).  BEGIN and
 541 END are buffer positions bounding the annotation, NAME is the name searched
 542 for in TRANSLATIONS, and POSITIVE should be non-nil if this annotation marks
 543 the beginning of a region with some property, or nil if it ends the region.
 544 NEXT-FN should return nil if there are no annotations after point.
 545
 546 The basic format of the TRANSLATIONS argument is described in the
 547 documentation for the `format-annotate-region' function.  There are some
 548 additional things to keep in mind for decoding, though:
 549
 550 When an annotation is found, the TRANSLATIONS list is searched for a
 551 text-property name and value that corresponds to that annotation.  If the
 552 text-property has several annotations associated with it, it will be used only
 553 if the other annotations are also in effect at that point.  The first match
 554 found whose annotations are all present is used.
 555
 556 The text property thus determined is set to the value over the region between
 557 the opening and closing annotations.  However, if the text-property name has a
 558 non-nil `format-list-valued' property, then the value will be consed onto the
 559 surrounding value of the property, rather than replacing that value.
 560
 561 There are some special symbols that can be used in the \"property\" slot of
 562 the TRANSLATIONS list: PARAMETER and FUNCTION \(spelled in uppercase).
 563 Annotations listed under the pseudo-property PARAMETER are considered to be
 564 arguments of the immediately surrounding annotation; the text between the
 565 opening and closing parameter annotations is deleted from the buffer but saved
 566 as a string.  The surrounding annotation should be listed under the
 567 pseudo-property FUNCTION.  Instead of inserting a text-property for this
 568 annotation, the function listed in the VALUE slot is called to make whatever
 569 changes are appropriate.  The function's first two arguments are the START and
 570 END locations, and the rest of the arguments are any PARAMETERs found in that
 571 region.
 572
 573 Any annotations that are found by NEXT-FN but not defined by TRANSLATIONS
 574 are saved as values of the `unknown' text-property \(which is list-valued).
 575 The TRANSLATIONS list should usually contain an entry of the form
 576     \(unknown \(nil format-annotate-value))
 577 to write these unknown annotations back into the file."
 578   (save-excursion
 579     (save-restriction
 580       (narrow-to-region (point-min) to)
 581       (goto-char from)
 582       (let (next open-ans todo loc unknown-ans)
 583         (while (setq next (funcall next-fn))
 584           (let* ((loc      (nth 0 next))
 585                  (end      (nth 1 next))
 586                  (name     (nth 2 next))
 587                  (positive (nth 3 next))
 588                  (found    nil))
 589
 590             ;; Delete the annotation
 591             (delete-region loc end)
 592             (cond
 593              ;; Positive annotations are stacked, remembering location
 594              (positive (setq open-ans (cons `(,name ((,loc . nil))) open-ans)))
 595              ;; It is a negative annotation:
 596              ;; Close the top annotation & add its text property.
 597              ;; If the file's nesting is messed up, the close might not match
 598              ;; the top thing on the open-annotations stack.
 599              ;; If no matching annotation is open, just ignore the close.
 600              ((not (assoc name open-ans))
 601               (message "Extra closing annotation (%s) in file" name))
 602              ;; If one is open, but not on the top of the stack, close
 603              ;; the things in between as well.  Set `found' when the real
 604              ;; one is closed.
 605              (t
 606               (while (not found)
 607                 (let* ((top (car open-ans))     ; first on stack: should match.
 608                        (top-name (car top))     ; text property name
 609                        (top-extents (nth 1 top)) ; property regions
 610                        (params (cdr (cdr top))) ; parameters
 611                        (aalist translations)
 612                        (matched nil))
 613                   (if (equal name top-name)
 614                       (setq found t)
 615                     (message "Improper nesting in file."))
 616                   ;; Look through property names in TRANSLATIONS
 617                   (while aalist
 618                     (let ((prop (car (car aalist)))
 619                           (alist (cdr (car aalist))))
 620                       ;; And look through values for each property
 621                       (while alist
 622                         (let ((value (car (car alist)))
 623                               (ans (cdr (car alist))))
 624                           (if (member top-name ans)
 625                               ;; This annotation is listed, but still have to
 626                               ;; check if multiple annotations are satisfied
 627                               (if (member nil (mapcar (lambda (r)
 628                                                         (assoc r open-ans))
 629                                                       ans))
 630                                   nil   ; multiple ans not satisfied
 631                                 ;; If there are multiple annotations going
 632                                 ;; into one text property, split up the other
 633                                 ;; annotations so they apply individually to
 634                                 ;; the other regions.
 635                                 (setcdr (car top-extents) loc)
 636                                 (let ((to-split ans) this-one extents)
 637                                   (while to-split
 638                                     (setq this-one
 639                                           (assoc (car to-split) open-ans)
 640                                           extents (nth 1 this-one))
 641                                     (if (not (eq this-one top))
 642                                         (setcar (cdr this-one)
 643                                                 (format-subtract-regions
 644                                                  extents top-extents)))
 645                                     (setq to-split (cdr to-split))))
 646                                 ;; Set loop variables to nil so loop
 647                                 ;; will exit.
 648                                 (setq alist nil aalist nil matched t
 649                                       ;; pop annotation off stack.
 650                                       open-ans (cdr open-ans))
 651                                 (let ((extents top-extents)
 652                                       (start (car (car top-extents)))
 653                                       (loc (cdr (car top-extents))))
 654                                   (while extents
 655                                     (cond
 656                                      ;; Check for pseudo-properties
 657                                      ((eq prop 'PARAMETER)
 658                                       ;; A parameter of the top open ann:
 659                                       ;; delete text and use as arg.
 660                                       (if open-ans
 661                                           ;; (If nothing open, discard).
 662                                           (setq open-ans
 663                                                 (cons
 664                                                  (append (car open-ans)
 665                                                          (list
 666                                                           (buffer-substring
 667                                                            start loc)))
 668                                                  (cdr open-ans))))
 669                                       (delete-region start loc))
 670                                      ((eq prop 'FUNCTION)
 671                                       ;; Not a property, but a function.
 672                                       (let ((rtn
 673                                              (apply value start loc params)))
 674                                         (if rtn (setq todo (cons rtn todo)))))
 675                                      (t
 676                                       ;; Normal property/value pair
 677                                       (setq todo
 678                                             (cons (list start loc prop value)
 679                                                   todo))))
 680                                     (setq extents (cdr extents)
 681                                           start (car (car extents))
 682                                           loc (cdr (car extents))))))))
 683                         (setq alist (cdr alist))))
 684                     (setq aalist (cdr aalist)))
 685                   (if (not matched)
 686                       ;; Didn't find any match for the annotation:
 687                       ;; Store as value of text-property `unknown'.
 688                       (let ((extents top-extents)
 689                             (start (car (car top-extents)))
 690                             (loc (or (cdr (car top-extents)) loc)))
 691                         (while extents
 692                           (setq open-ans (cdr open-ans)
 693                                 todo (cons (list start loc 'unknown top-name)
 694                                            todo)
 695                                 unknown-ans (cons name unknown-ans)
 696                                 extents (cdr extents)
 697                                 start (car (car extents))
 698                                 loc (cdr (car extents))))))))))))
 699
 700         ;; Once entire file has been scanned, add the properties.
 701         (while todo
 702           (let* ((item (car todo))
 703                  (from (nth 0 item))
 704                  (to   (nth 1 item))
 705                  (prop (nth 2 item))
 706                  (val  (nth 3 item)))
 707
 708             (if (numberp val)   ; add to ambient value if numeric
 709                 (format-property-increment-region from to prop val 0)
 710               (put-text-property
 711                from to prop
 712                (cond ((get prop 'format-list-valued) ; value gets consed onto
 713                                                      ; list-valued properties
 714                       (let ((prev (get-text-property from prop)))
 715                         (cons val (if (listp prev) prev (list prev)))))
 716                      (t val))))) ; normally, just set to val.
 717           (setq todo (cdr todo)))
 718
 719         (if unknown-ans
 720             (message "Unknown annotations: %s" unknown-ans))))))
 721
 722 (defun format-subtract-regions (minu subtra)
 723   "Remove from the regions in MINUend the regions in SUBTRAhend.
 724 A region is a dotted pair (from . to).  Both parameters are lists of
 725 regions.  Each list must contain nonoverlapping, noncontiguous
 726 regions, in descending order.  The result is also nonoverlapping,
 727 noncontiguous, and in descending order.  The first element of MINUEND
 728 can have a cdr of nil, indicating that the end of that region is not
 729 yet known."
 730   (let* ((minuend (copy-alist minu))
 731          (subtrahend (copy-alist subtra))
 732          (m (car minuend))
 733          (s (car subtrahend))
 734          results)
 735     (while (and minuend subtrahend)
 736       (cond
 737        ;; The minuend starts after the subtrahend ends; keep it.
 738        ((> (car m) (cdr s))
 739         (setq results (cons m results)
 740               minuend (cdr minuend)
 741               m (car minuend)))
 742        ;; The minuend extends beyond the end of the subtrahend.  Chop it off.
 743        ((or (null (cdr m)) (> (cdr m) (cdr s)))
 744         (setq results (cons (cons (1+ (cdr s)) (cdr m)) results))
 745         (setcdr m (cdr s)))
 746        ;; The subtrahend starts after the minuend ends; throw it away.
 747        ((< (cdr m) (car s))
 748         (setq subtrahend (cdr subtrahend) s (car subtrahend)))
 749        ;; The subtrahend extends beyond the end of the minuend.  Chop it off.
 750        (t       ;(<= (cdr m) (cdr s)))
 751         (if (>= (car m) (car s))
 752             (setq minuend (cdr minuend) m (car minuend))
 753           (setcdr m (1- (car s)))
 754           (setq subtrahend (cdr subtrahend) s (car subtrahend))))))
 755     (nconc (nreverse results) minuend)))
 756
 757 ;; This should probably go somewhere other than format.el.  Then again,
 758 ;; indent.el has alter-text-property.  NOTE: We can also use
 759 ;; next-single-property-change instead of text-property-not-all, but then
 760 ;; we have to see if we passed TO.
 761 (defun format-property-increment-region (from to prop delta default)
 762   "Over the region between FROM and TO increment property PROP by amount DELTA.
 763 DELTA may be negative.  If property PROP is nil anywhere
 764 in the region, it is treated as though it were DEFAULT."
 765   (let ((cur from) val newval next)
 766     (while cur
 767       (setq val    (get-text-property cur prop)
 768             newval (+ (or val default) delta)
 769             next   (text-property-not-all cur to prop val))
 770       (put-text-property cur (or next to) prop newval)
 771       (setq cur next))))
 772
 773 ;;;
 774 ;;; Encoding
 775 ;;;
 776
 777 (defun format-insert-annotations (list &optional offset)
 778   "Apply list of annotations to buffer as `write-region' would.
 779 Inserts each element of the given LIST of buffer annotations at its
 780 appropriate place.  Use second arg OFFSET if the annotations' locations are
 781 not relative to the beginning of the buffer: annotations will be inserted
 782 at their location-OFFSET+1 \(ie, the offset is treated as the character number
 783 of the first character in the buffer)."
 784   (if (not offset)
 785       (setq offset 0)
 786     (setq offset (1- offset)))
 787   (let ((l (reverse list)))
 788     (while l
 789       (goto-char (- (car (car l)) offset))
 790       (insert (cdr (car l)))
 791       (setq l (cdr l)))))
 792
 793 (defun format-annotate-value (old new)
 794   "Return OLD and NEW as a \(close . open) annotation pair.
 795 Useful as a default function for TRANSLATIONS alist when the value of the text
 796 property is the name of the annotation that you want to use, as it is for the
 797 `unknown' text property."
 798   (cons (if old (list old))
 799         (if new (list new))))
 800
 801 (defun format-annotate-region (from to translations format-fn ignore)
 802   "Generate annotations for text properties in the region.
 803 Searches for changes between FROM and TO, and describes them with a list of
 804 annotations as defined by alist TRANSLATIONS and FORMAT-FN.  IGNORE lists text
 805 properties not to consider; any text properties that are neither ignored nor
 806 listed in TRANSLATIONS are warned about.
 807 If you actually want to modify the region, give the return value of this
 808 function to `format-insert-annotations'.
 809
 810 Format of the TRANSLATIONS argument:
 811
 812 Each element is a list whose car is a PROPERTY, and the following
 813 elements are VALUES of that property followed by the names of zero or more
 814 ANNOTATIONS.  Whenever the property takes on that value, the annotations
 815 \(as formatted by FORMAT-FN) are inserted into the file.
 816 When the property stops having that value, the matching negated annotation
 817 will be inserted \(it may actually be closed earlier and reopened, if
 818 necessary, to keep proper nesting).
 819
 820 If the property's value is a list, then each element of the list is dealt with
 821 separately.
 822
 823 If a VALUE is numeric, then it is assumed that there is a single annotation
 824 and each occurrence of it increments the value of the property by that number.
 825 Thus, given the entry \(left-margin \(4 \"indent\")), if the left margin
 826 changes from 4 to 12, two <indent> annotations will be generated.
 827
 828 If the VALUE is nil, then instead of annotations, a function should be
 829 specified.  This function is used as a default: it is called for all
 830 transitions not explicitly listed in the table.  The function is called with
 831 two arguments, the OLD and NEW values of the property.  It should return
 832 lists of annotations like `format-annotate-location' does.
 833
 834     The same structure can be used in reverse for reading files."
 835   (let ((all-ans nil)    ; All annotations - becomes return value
 836         (open-ans nil)   ; Annotations not yet closed
 837         (loc nil)        ; Current location
 838         (not-found nil)) ; Properties that couldn't be saved
 839     (while (or (null loc)
 840                (and (setq loc (next-property-change loc nil to))
 841                     (< loc to)))
 842       (or loc (setq loc from))
 843       (let* ((ans (format-annotate-location loc (= loc from) ignore translations))
 844              (neg-ans (format-reorder (aref ans 0) open-ans))
 845              (pos-ans (aref ans 1))
 846              (ignored (aref ans 2)))
 847         (setq not-found (append ignored not-found)
 848               ignore    (append ignored ignore))
 849         ;; First do the negative (closing) annotations
 850         (while neg-ans
 851           ;; Check if it's missing.  This can happen (eg, a numeric property
 852           ;; going negative can generate closing annotations before there are
 853           ;; any open).  Warn user & ignore.
 854           (if (not (member (car neg-ans) open-ans))
 855               (message "Can't close %s: not open." (car neg-ans))
 856             (while (not (equal (car neg-ans) (car open-ans)))
 857               ;; To close anno. N, need to first close ans 1 to N-1,
 858               ;; remembering to re-open them later.
 859               (setq pos-ans (cons (car open-ans) pos-ans))
 860               (setq all-ans
 861                     (cons (cons loc (funcall format-fn (car open-ans) nil))
 862                           all-ans))
 863               (setq open-ans (cdr open-ans)))
 864             ;; Now remove the one we're really interested in from open list.
 865             (setq open-ans (cdr open-ans))
 866             ;; And put the closing annotation here.
 867             (setq all-ans
 868                   (cons (cons loc (funcall format-fn (car neg-ans) nil))
 869                         all-ans)))
 870           (setq neg-ans (cdr neg-ans)))
 871         ;; Now deal with positive (opening) annotations
 872         (let ((p pos-ans))
 873           (while pos-ans
 874             (setq open-ans (cons (car pos-ans) open-ans))
 875             (setq all-ans
 876                   (cons (cons loc (funcall format-fn (car pos-ans) t))
 877                         all-ans))
 878             (setq pos-ans (cdr pos-ans))))))
 879
 880     ;; Close any annotations still open
 881     (while open-ans
 882       (setq all-ans
 883             (cons (cons to (funcall format-fn (car open-ans) nil))
 884                   all-ans))
 885       (setq open-ans (cdr open-ans)))
 886     (if not-found
 887         (message "These text properties could not be saved:\n    %s"
 888                  not-found))
 889     (nreverse all-ans)))
 890
 891 ;;; Internal functions for format-annotate-region.
 892
 893 (defun format-annotate-location (loc all ignore translations)
 894   "Return annotation(s) needed at location LOC.
 895 This includes any properties that change between LOC-1 and LOC.
 896 If ALL is true, don't look at previous location, but generate annotations for
 897 all non-nil properties.
 898 Third argument IGNORE is a list of text-properties not to consider.
 899 Use the TRANSLATIONS alist.
 900
 901 Return value is a vector of 3 elements:
 902 1. List of names of the annotations to close
 903 2. List of the names of annotations to open.
 904 3. List of properties that were ignored or couldn't be annotated."
 905   (let* ((prev-loc (1- loc))
 906          (before-plist (if all nil (text-properties-at prev-loc)))
 907          (after-plist (text-properties-at loc))
 908          p negatives positives prop props not-found)
 909     ;; make list of all property names involved
 910     (setq p before-plist)
 911     (while p
 912       (if (not (memq (car p) props))
 913           (setq props (cons (car p) props)))
 914       (setq p (cdr (cdr p))))
 915     (setq p after-plist)
 916     (while p
 917       (if (not (memq (car p) props))
 918           (setq props (cons (car p) props)))
 919       (setq p (cdr (cdr p))))
 920
 921     (while props
 922       (setq prop (car props)
 923             props (cdr props))
 924       (if (memq prop ignore)
 925           nil  ; If it's been ignored before, ignore it now.
 926         (let ((before (if all nil (car (cdr (memq prop before-plist)))))
 927               (after (car (cdr (memq prop after-plist)))))
 928           (if (equal before after)
 929               nil ; no change; ignore
 930             (let ((result (format-annotate-single-property-change
 931                            prop before after translations)))
 932               (if (not result)
 933                   (setq not-found (cons prop not-found))
 934                 (setq negatives (nconc negatives (car result))
 935                       positives (nconc positives (cdr result)))))))))
 936     (vector negatives positives not-found)))
 937
 938 (defun format-annotate-single-property-change (prop old new trans)
 939   "Return annotations for property PROP changing from OLD to NEW.
 940 These are searched for in the translations alist TRANS.
 941 If NEW does not appear in the list, but there is a default function, then that
 942 function is called.
 943 Annotations to open and to close are returned as a dotted pair."
 944   (let ((prop-alist (cdr (assoc prop trans)))
 945         default)
 946     (if (not prop-alist)
 947         nil
 948       ;; If either old or new is a list, have to treat both that way.
 949       (if (and (or (listp old) (listp new))
 950                (not (get prop 'format-list-atomic-p)))
 951           (if (or (not (format-proper-list-p old))
 952                   (not (format-proper-list-p new)))
 953               (format-annotate-atomic-property-change prop-alist old new)
 954             (let* ((old (if (listp old) old (list old)))
 955                    (new (if (listp new) new (list new)))
 956                    (tail (format-common-tail old new))
 957                    close open)
 958               (while old
 959                 (setq close
 960                       (append (car (format-annotate-atomic-property-change
 961                                     prop-alist (car old) nil))
 962                               close)
 963                       old (cdr old)))
 964               (while new
 965                 (setq open
 966                       (append (cdr (format-annotate-atomic-property-change
 967                                     prop-alist nil (car new)))
 968                               open)
 969                       new (cdr new)))
 970               (format-make-relatively-unique close open)))
 971         (format-annotate-atomic-property-change prop-alist old new)))))
 972
 973 (defun format-annotate-atomic-property-change (prop-alist old new)
 974   "Internal function annotate a single property change.
 975 PROP-ALIST is the relevant segment of a TRANSLATIONS list.
 976 OLD and NEW are the values."
 977   (let (num-ann)
 978     ;; If old and new values are numbers,
 979     ;; look for a number in PROP-ALIST.
 980     (if (and (or (null old) (numberp old))
 981              (or (null new) (numberp new)))
 982         (progn
 983           (setq num-ann prop-alist)
 984           (while (and num-ann (not (numberp (car (car num-ann)))))
 985             (setq num-ann (cdr num-ann)))))
 986     (if num-ann
 987         ;; Numerical annotation - use difference
 988         (progn
 989           ;; If property is numeric, nil means 0
 990           (cond ((and (numberp old) (null new))
 991                  (setq new 0))
 992                 ((and (numberp new) (null old))
 993                  (setq old 0)))
 994
 995           (let* ((entry (car num-ann))
 996                  (increment (car entry))
 997                  (n (ceiling (/ (float (- new old)) (float increment))))
 998                  (anno (car (cdr entry))))
 999             (if (> n 0)
1000                 (cons nil (make-list n anno))
1001               (cons (make-list (- n) anno) nil))))
1002
1003       ;; Standard annotation
1004       (let ((close (and old (cdr (assoc old prop-alist))))
1005             (open  (and new (cdr (assoc new prop-alist)))))
1006         (if (or close open)
1007             (format-make-relatively-unique close open)
1008           ;; Call "Default" function, if any
1009           (let ((default (assq nil prop-alist)))
1010             (if default
1011                 (funcall (car (cdr default)) old new))))))))
1012
1013 (provide 'format)
1014
1015 ;;; format.el ends here