3 PROPOSAL FOR HOW THIS ALL OUGHT TO WORK
4 this isn't implemented yet, but this is the plan-in-progress
6 In general, it's accepted that the best way to internationalize is for all
7 messages to be referred to by a symbolic name (or number) and come out of a
8 table or tables, which are easy to change.
10 However, with Emacs, we've got the task of internationalizing a huge body
11 of existing code, which already contains messages internally.
13 For the C code we've got two options:
15 - Use a Sun-like gettext() form, which takes an "english" string which
16 appears literally in the source, and uses that as a hash key to find
18 - Rip all of the strings out and put them in a table.
20 In this case, it's desirable to make as few changes as possible to the C
21 code, to make it easier to merge the code with the FSF version of emacs
22 which won't ever have these changes made to it. So we should go with the
25 The way it has been done (between 19.8 and 19.9) was to use gettext(), but
26 *also* to make massive changes to the source code. The goal now is to use
27 gettext() at run-time and yet not require a textual change to every line
28 in the C code which contains a string constant. A possible way to do this
31 (gettext() can be implemented in terms of catgets() for non-Sun systems, so
32 that in itself isn't a problem.)
34 For the Lisp code, we've got basically the same options: put everything in
35 a table, or translate things implicitly.
37 Another kink that lisp code introduces is that there are thousands of third-
38 party packages, so changing the source for all of those is simply not an
41 Is it a goal that if some third party package displays a message which is
42 one we know how to translate, then we translate it? I think this is a
43 worthy goal. It remains to be seen how well it will work in practice.
45 So, we should endeavor to minimize the impact on the lisp code. Certain
46 primitive lisp routines (the stuff in lisp/prim/, and especially in
47 cmdloop.el and minibuf.el) may need to be changed to know about translation,
48 but that's an ideologically clean thing to do because those are considered
49 a part of the emacs substrate.
51 However, if we find ourselves wanting to make changes to, say, RMAIL, then
52 something has gone wrong. (Except to do things like remove assumptions
53 about the order of words within a sentence, or how pluralization works.)
55 There are two parts to the task of displaying translated strings to the
56 user: the first is to extract the strings which need to be translated from
57 the sources; and the second is to make some call which will translate those
58 strings before they are presented to the user.
60 The old way was to use the same form to do both, that is, GETTEXT() was both
61 the tag that we searched for to build a catalog, and was the form which did
62 the translation. The new plan is to separate these two things more: the
63 tags that we search for to build the catalog will be stuff that was in there
64 already, and the translation will get done in some more centralized, lower
67 This program (make-msgfile.c) addresses the first part, extracting the
70 For the emacs C code, we need to recognize the following patterns:
72 message ("string" ... )
74 report_file_error ("string" ... )
75 signal_simple_error ("string" ... )
76 signal_simple_error_2 ("string" ... )
78 build_translated_string ("string")
79 #### add this and use it instead of build_string() in some places.
81 yes_or_no_p ("string" ... )
82 #### add this instead of funcalling Qyes_or_no_p directly.
84 barf_or_query_if_file_exists #### restructure this
85 check all callers of Fsignal #### restructure these
86 signal_error (Qerror ... ) #### change all of these to error()
88 And we also parse out the `interactive' prompts from DEFUN() forms.
90 #### When we've got a string which is a candidate for translation, we
91 should ignore it if it contains only format directives, that is, if
92 there are no alphabetic characters in it that are not a part of a `%'
93 directive. (Careful not to translate either "%s%s" or "%s: ".)
95 For the emacs Lisp code, we need to recognize the following patterns:
97 (message "string" ... )
99 (format "string" ... )
100 (read-from-minibuffer "string" ... )
101 (read-shell-command "string" ... )
102 (y-or-n-p "string" ... )
103 (yes-or-no-p "string" ... )
104 (read-file-name "string" ... )
105 (temp-minibuffer-message "string")
106 (query-replace-read-args "string" ... )
108 I expect there will be a lot like the above; basically, any function which
109 is a commonly used wrapper around an eventual call to `message' or
110 `read-from-minibuffer' needs to be recognized by this program.
112 (dgettext "domain-name" "string") #### do we still need this?
114 things that should probably be restructured:
115 `princ' in cmdloop.el
118 help.el, syntax.el all messed up
120 Menu descriptors: one way to extract the strings in menu labels would be
121 to teach this program about "^(defvar .*menu\n" forms; that's probably
122 kind of hard, though, so perhaps a better approach would be to make this
123 program recognize lines of the form
125 "string" ... ;###translate
127 where the magic token ";###translate" on a line means that the string
128 constant on this line should go into the message catalog. This is analogous
129 to the magic ";###autoload" comments, and to the magic comments used in the
130 EPSF structuring conventions.
133 So this program manages to build up a catalog of strings to be translated.
134 To address the second part of the problem, of actually looking up the
135 translations, there are hooks in a small number of low level places in
138 Assume the existence of a C function gettext(str) which returns the
139 translation of `str' if there is one, otherwise returns `str'.
141 - message() takes a char* as its argument, and always filters it through
142 gettext() before displaying it.
144 - errors are printed by running the lisp function `display-error' which
145 doesn't call `message' directly (it princ's to streams), so it must be
146 carefully coded to translate its arguments. This is only a few lines
149 - Fread_minibuffer_internal() is the lowest level interface to all minibuf
150 interactions, so it is responsible for translating the value that will go
151 into Vminibuf_prompt.
153 - Fpopup_menu filters the menu titles through gettext().
155 The above take care of 99% of all messages the user ever sees.
157 - The lisp function temp-minibuffer-message translates its arg.
159 - query-replace-read-args is funny; it does
160 (setq from (read-from-minibuffer (format "%s: " string) ... ))
161 (setq to (read-from-minibuffer (format "%s %s with: " string from) ... ))
163 What should we do about this? We could hack query-replace-read-args to
164 translate its args, but might this be a more general problem? I don't
165 think we ought to translate all calls to format. We could just change
166 the calling sequence, since this is odd in that the first %s wants to be
167 translated but the second doesn't.
169 Solving the "translating too much" problem:
170 The concern has been raised that in this situation:
171 - "Help" is a string for which we know a translation;
172 - someone visits a file called Help, and someone does something
173 contrived like (error buffer-file-name)
174 then we would display the translation of Help, which would not be correct.
175 We can solve this by adding a bit to Lisp_String objects which identifies
176 them as having been read as literal constants from a .el or .elc file (as
177 opposed to having been constructed at run time as it would in the above
178 case.) To solve this:
180 - Fmessage() takes a lisp string as its first argument.
181 If that string is a constant, that is, was read from a source file
182 as a literal, then it calls message() with it, which translates.
183 Otherwise, it calls message_no_translate(), which does not translate.
185 - Ferror() (actually, Fsignal() when condition is Qerror) works similarly.
188 /* Scan specified C and Lisp files, extracting the following messages:
193 DEFUN interactive prompts
196 (dgettext "domain-name" ...)
200 The arguments given to this program are all the C and Lisp source files
201 of GNU Emacs. .el and .c files are allowed. There is no support for .elc
202 files at this time, but they may be specified; the corresponding .el file
203 will be used. Similarly, .o files can also be specified, and the corresponding
204 .c file will be used. This helps the makefile pass the correct list of files.
206 The results, which go to standard output or to a file specified with -a or -o
207 (-a to append, -o to start from nothing), are quoted strings wrapped in
208 gettext(...). The results can be passed to xgettext to produce a .po message
216 #define GET_LINE fgets (line, LINESIZE, infile)
217 #define CHECK_EOL(p) if (*(p) == '\0') (p) = GET_LINE
218 #define SKIP_BLANKS(p) while ((*p) == ' ' || (*p) == '\t') (p)++
220 enum filetype { C_FILE, LISP_FILE, INVALID_FILE };
221 /* some brain-dead headers define this ... */
224 enum boolean { FALSE, TRUE };
230 void scan_file(char *filename);
231 void process_C_file(void);
232 void process_Lisp_file(void);
233 char *copy_up_to_paren(register char *p);
234 char *copy_quoted_string(register char *p);
235 enum boolean no_interactive_prompt(register char *q);
236 char *skip_blanks(register char *p);
238 main(int argc, char *argv[])
244 /* If first two args are -o FILE, output to FILE. */
246 if (argc > i + 1 && strcmp(argv[i], "-o") == 0) {
247 outfile = fopen(argv[++i], "w");
250 /* ...Or if args are -a FILE, append to FILE. */
251 if (argc > i + 1 && strcmp(argv[i], "-a") == 0) {
252 outfile = fopen(argv[++i], "a");
256 fprintf(stderr, "Unable to open output file %s\n", argv[--i]);
260 for (; i < argc; i++)
266 void scan_file(char *filename)
268 enum filetype type = INVALID_FILE;
269 register char *p = filename + strlen(filename);
271 if (strcmp(p - 4, ".elc") == 0) {
272 *--p = '\0'; /* Use .el file instead */
274 } else if (strcmp(p - 3, ".el") == 0)
276 else if (strcmp(p - 2, ".o") == 0) {
277 *--p = 'c'; /* Use .c file instead */
279 } else if (strcmp(p - 2, ".c") == 0)
282 if (type == INVALID_FILE) {
283 fprintf(stderr, "File %s being ignored\n", filename);
286 infile = fopen(filename, "r");
288 fprintf(stderr, "Unable to open input file %s\n", filename);
292 fprintf(outfile, "/* %s */\n", filename);
297 fputc('\n', outfile);
302 void process_C_file(void)
305 char *gettext, *defun;
307 while (p = GET_LINE) {
308 gettext = strstr(p, "GETTEXT");
309 defun = strstr(p, "DEFUN");
310 if (gettext || defun) {
313 p += 7; /* Skip over "GETTEXT" */
316 p += 5; /* Skip over "DEFUN" */
326 for (i = 0; i < 5; i++) /* Skip over commas to doc string */
334 if (*p != '\"') /* Make sure there is a quoted string */
337 if (defun && no_interactive_prompt(p))
340 fprintf(outfile, "gettext(");
342 p = copy_up_to_paren(p);
344 p = copy_quoted_string(p);
345 fprintf(outfile, ")\n");
350 void process_Lisp_file(void)
353 char *gettext, *interactive;
354 enum boolean dgettext = FALSE;
356 while (p = GET_LINE) {
357 gettext = strstr(p, "gettext");
358 interactive = strstr(p, "(interactive");
359 if (gettext || interactive) {
364 else if (gettext < interactive) {
373 if (p > line && *(p - 1) == 'd')
375 p += 7; /* Skip over "gettext" */
377 p += 12; /* Skip over "(interactive" */
380 if (*p != '\"') /* Make sure there is a quoted string */
383 if (dgettext) { /* Skip first quoted string (domain name) */
384 while (*++p != '"') ; /* null statement */
387 if (*p != '\"') /* Check for second quoted string (message) */
391 if (interactive && no_interactive_prompt(p))
394 fprintf(outfile, "gettext(");
395 p = copy_up_to_paren(p);
396 fprintf(outfile, ")\n");
401 /* Assuming p points to some character beyond an opening parenthesis, copy
402 everything to outfile up to but not including the closing parenthesis.
404 char *copy_up_to_paren(register char *p)
407 SKIP_BLANKS(p); /* We don't call skip_blanks() in order to */
408 CHECK_EOL(p); /* preserve blanks at the beginning of the line */
413 p = copy_quoted_string(p);
415 fputc(*p++, outfile);
420 /* Assuming p points to a quote character, copy the quoted string to outfile.
422 char *copy_quoted_string(register char *p)
426 fputc(*p++, outfile);
427 fputc(*p++, outfile);
429 } while (*p != '\"');
431 fputc(*p++, outfile);
435 /* Return TRUE if the interactive specification consists only
436 of code letters and no prompt.
438 enum boolean no_interactive_prompt(register char *q)
440 while (++q, *q == '*' || *q == '@') ; /* null statement */
446 if (*q == '\\' && *++q == 'n') {
448 goto skip_code_letter;
453 char *skip_blanks(register char *p)
455 while (*p == ' ' || *p == '\t' || *p == '\n') {