Commit | Line | Data |
---|---|---|
920dae64 AT |
1 | |
2 | # This document contains text in Perl "POD" format. | |
3 | # Use a POD viewer like perldoc or perlman to render it. | |
4 | ||
5 | # This corrects some typoes in the previous release. | |
6 | ||
7 | =head1 NAME | |
8 | ||
9 | Locale::Maketext::TPJ13 -- article about software localization | |
10 | ||
11 | =head1 SYNOPSIS | |
12 | ||
13 | # This an article, not a module. | |
14 | ||
15 | =head1 DESCRIPTION | |
16 | ||
17 | The following article by Sean M. Burke and Jordan Lachler | |
18 | first appeared in I<The Perl | |
19 | Journal> #13 and is copyright 1999 The Perl Journal. It appears | |
20 | courtesy of Jon Orwant and The Perl Journal. This document may be | |
21 | distributed under the same terms as Perl itself. | |
22 | ||
23 | =head1 Localization and Perl: gettext breaks, Maketext fixes | |
24 | ||
25 | by Sean M. Burke and Jordan Lachler | |
26 | ||
27 | This article points out cases where gettext (a common system for | |
28 | localizing software interfaces -- i.e., making them work in the user's | |
29 | language of choice) fails because of basic differences between human | |
30 | languages. This article then describes Maketext, a new system capable | |
31 | of correctly treating these differences. | |
32 | ||
33 | =head2 A Localization Horror Story: It Could Happen To You | |
34 | ||
35 | =over | |
36 | ||
37 | "There are a number of languages spoken by human beings in this | |
38 | world." | |
39 | ||
40 | -- Harald Tveit Alvestrand, in RFC 1766, "Tags for the | |
41 | Identification of Languages" | |
42 | ||
43 | =back | |
44 | ||
45 | Imagine that your task for the day is to localize a piece of software | |
46 | -- and luckily for you, the only output the program emits is two | |
47 | messages, like this: | |
48 | ||
49 | I scanned 12 directories. | |
50 | ||
51 | Your query matched 10 files in 4 directories. | |
52 | ||
53 | So how hard could that be? You look at the code that | |
54 | produces the first item, and it reads: | |
55 | ||
56 | printf("I scanned %g directories.", | |
57 | $directory_count); | |
58 | ||
59 | You think about that, and realize that it doesn't even work right for | |
60 | English, as it can produce this output: | |
61 | ||
62 | I scanned 1 directories. | |
63 | ||
64 | So you rewrite it to read: | |
65 | ||
66 | printf("I scanned %g %s.", | |
67 | $directory_count, | |
68 | $directory_count == 1 ? | |
69 | "directory" : "directories", | |
70 | ); | |
71 | ||
72 | ...which does the Right Thing. (In case you don't recall, "%g" is for | |
73 | locale-specific number interpolation, and "%s" is for string | |
74 | interpolation.) | |
75 | ||
76 | But you still have to localize it for all the languages you're | |
77 | producing this software for, so you pull Locale::gettext off of CPAN | |
78 | so you can access the C<gettext> C functions you've heard are standard | |
79 | for localization tasks. | |
80 | ||
81 | And you write: | |
82 | ||
83 | printf(gettext("I scanned %g %s."), | |
84 | $dir_scan_count, | |
85 | $dir_scan_count == 1 ? | |
86 | gettext("directory") : gettext("directories"), | |
87 | ); | |
88 | ||
89 | But you then read in the gettext manual (Drepper, Miller, and Pinard 1995) | |
90 | that this is not a good idea, since how a single word like "directory" | |
91 | or "directories" is translated may depend on context -- and this is | |
92 | true, since in a case language like German or Russian, you'd may need | |
93 | these words with a different case ending in the first instance (where the | |
94 | word is the object of a verb) than in the second instance, which you haven't even | |
95 | gotten to yet (where the word is the object of a preposition, "in %g | |
96 | directories") -- assuming these keep the same syntax when translated | |
97 | into those languages. | |
98 | ||
99 | So, on the advice of the gettext manual, you rewrite: | |
100 | ||
101 | printf( $dir_scan_count == 1 ? | |
102 | gettext("I scanned %g directory.") : | |
103 | gettext("I scanned %g directories."), | |
104 | $dir_scan_count ); | |
105 | ||
106 | So, you email your various translators (the boss decides that the | |
107 | languages du jour are Chinese, Arabic, Russian, and Italian, so you | |
108 | have one translator for each), asking for translations for "I scanned | |
109 | %g directory." and "I scanned %g directories.". When they reply, | |
110 | you'll put that in the lexicons for gettext to use when it localizes | |
111 | your software, so that when the user is running under the "zh" | |
112 | (Chinese) locale, gettext("I scanned %g directory.") will return the | |
113 | appropriate Chinese text, with a "%g" in there where printf can then | |
114 | interpolate $dir_scan. | |
115 | ||
116 | Your Chinese translator emails right back -- he says both of these | |
117 | phrases translate to the same thing in Chinese, because, in linguistic | |
118 | jargon, Chinese "doesn't have number as a grammatical category" -- | |
119 | whereas English does. That is, English has grammatical rules that | |
120 | refer to "number", i.e., whether something is grammatically singular | |
121 | or plural; and one of these rules is the one that forces nouns to take | |
122 | a plural suffix (generally "s") when in a plural context, as they are when | |
123 | they follow a number other than "one" (including, oddly enough, "zero"). | |
124 | Chinese has no such rules, and so has just the one phrase where English | |
125 | has two. But, no problem, you can have this one Chinese phrase appear | |
126 | as the translation for the two English phrases in the "zh" gettext | |
127 | lexicon for your program. | |
128 | ||
129 | Emboldened by this, you dive into the second phrase that your software | |
130 | needs to output: "Your query matched 10 files in 4 directories.". You notice | |
131 | that if you want to treat phrases as indivisible, as the gettext | |
132 | manual wisely advises, you need four cases now, instead of two, to | |
133 | cover the permutations of singular and plural on the two items, | |
134 | $dir_count and $file_count. So you try this: | |
135 | ||
136 | printf( $file_count == 1 ? | |
137 | ( $directory_count == 1 ? | |
138 | gettext("Your query matched %g file in %g directory.") : | |
139 | gettext("Your query matched %g file in %g directories.") ) : | |
140 | ( $directory_count == 1 ? | |
141 | gettext("Your query matched %g files in %g directory.") : | |
142 | gettext("Your query matched %g files in %g directories.") ), | |
143 | $file_count, $directory_count, | |
144 | ); | |
145 | ||
146 | (The case of "1 file in 2 [or more] directories" could, I suppose, | |
147 | occur in the case of symlinking or something of the sort.) | |
148 | ||
149 | It occurs to you that this is not the prettiest code you've ever | |
150 | written, but this seems the way to go. You mail off to the | |
151 | translators asking for translations for these four cases. The | |
152 | Chinese guy replies with the one phrase that these all translate to in | |
153 | Chinese, and that phrase has two "%g"s in it, as it should -- but | |
154 | there's a problem. He translates it word-for-word back: "In %g | |
155 | directories contains %g files match your query." The %g | |
156 | slots are in an order reverse to what they are in English. You wonder | |
157 | how you'll get gettext to handle that. | |
158 | ||
159 | But you put it aside for the moment, and optimistically hope that the | |
160 | other translators won't have this problem, and that their languages | |
161 | will be better behaved -- i.e., that they will be just like English. | |
162 | ||
163 | But the Arabic translator is the next to write back. First off, your | |
164 | code for "I scanned %g directory." or "I scanned %g directories." | |
165 | assumes there's only singular or plural. But, to use linguistic | |
166 | jargon again, Arabic has grammatical number, like English (but unlike | |
167 | Chinese), but it's a three-term category: singular, dual, and plural. | |
168 | In other words, the way you say "directory" depends on whether there's | |
169 | one directory, or I<two> of them, or I<more than two> of them. Your | |
170 | test of C<($directory == 1)> no longer does the job. And it means | |
171 | that where English's grammatical category of number necessitates | |
172 | only the two permutations of the first sentence based on "directory | |
173 | [singular]" and "directories [plural]", Arabic has three -- and, | |
174 | worse, in the second sentence ("Your query matched %g file in %g | |
175 | directory."), where English has four, Arabic has nine. You sense | |
176 | an unwelcome, exponential trend taking shape. | |
177 | ||
178 | Your Italian translator emails you back and says that "I searched 0 | |
179 | directories" (a possible English output of your program) is stilted, | |
180 | and if you think that's fine English, that's your problem, but that | |
181 | I<just will not do> in the language of Dante. He insists that where | |
182 | $directory_count is 0, your program should produce the Italian text | |
183 | for "I I<didn't> scan I<any> directories.". And ditto for "I didn't | |
184 | match any files in any directories", although he says the last part | |
185 | about "in any directories" should probably just be left off. | |
186 | ||
187 | You wonder how you'll get gettext to handle this; to accomodate the | |
188 | ways Arabic, Chinese, and Italian deal with numbers in just these few | |
189 | very simple phrases, you need to write code that will ask gettext for | |
190 | different queries depending on whether the numerical values in | |
191 | question are 1, 2, more than 2, or in some cases 0, and you still haven't | |
192 | figured out the problem with the different word order in Chinese. | |
193 | ||
194 | Then your Russian translator calls on the phone, to I<personally> tell | |
195 | you the bad news about how really unpleasant your life is about to | |
196 | become: | |
197 | ||
198 | Russian, like German or Latin, is an inflectional language; that is, nouns | |
199 | and adjectives have to take endings that depend on their case | |
200 | (i.e., nominative, accusative, genitive, etc...) -- which is roughly a matter of | |
201 | what role they have in syntax of the sentence -- | |
202 | as well as on the grammatical gender (i.e., masculine, feminine, neuter) | |
203 | and number (i.e., singular or plural) of the noun, as well as on the | |
204 | declension class of the noun. But unlike with most other inflected languages, | |
205 | putting a number-phrase (like "ten" or "forty-three", or their Arabic | |
206 | numeral equivalents) in front of noun in Russian can change the case and | |
207 | number that noun is, and therefore the endings you have to put on it. | |
208 | ||
209 | He elaborates: In "I scanned %g directories", you'd I<expect> | |
210 | "directories" to be in the accusative case (since it is the direct | |
211 | object in the sentnce) and the plural number, | |
212 | except where $directory_count is 1, then you'd expect the singular, of | |
213 | course. Just like Latin or German. I<But!> Where $directory_count % | |
214 | 10 is 1 ("%" for modulo, remember), assuming $directory count is an | |
215 | integer, and except where $directory_count % 100 is 11, "directories" | |
216 | is forced to become grammatically singular, which means it gets the | |
217 | ending for the accusative singular... You begin to visualize the code | |
218 | it'd take to test for the problem so far, I<and still work for Chinese | |
219 | and Arabic and Italian>, and how many gettext items that'd take, but | |
220 | he keeps going... But where $directory_count % 10 is 2, 3, or 4 | |
221 | (except where $directory_count % 100 is 12, 13, or 14), the word for | |
222 | "directories" is forced to be genitive singular -- which means another | |
223 | ending... The room begins to spin around you, slowly at first... But | |
224 | with I<all other> integer values, since "directory" is an inanimate | |
225 | noun, when preceded by a number and in the nominative or accusative | |
226 | cases (as it is here, just your luck!), it does stay plural, but it is | |
227 | forced into the genitive case -- yet another ending... And | |
228 | you never hear him get to the part about how you're going to run into | |
229 | similar (but maybe subtly different) problems with other Slavic | |
230 | languages like Polish, because the floor comes up to meet you, and you | |
231 | fade into unconsciousness. | |
232 | ||
233 | ||
234 | The above cautionary tale relates how an attempt at localization can | |
235 | lead from programmer consternation, to program obfuscation, to a need | |
236 | for sedation. But careful evaluation shows that your choice of tools | |
237 | merely needed further consideration. | |
238 | ||
239 | =head2 The Linguistic View | |
240 | ||
241 | =over | |
242 | ||
243 | "It is more complicated than you think." | |
244 | ||
245 | -- The Eighth Networking Truth, from RFC 1925 | |
246 | ||
247 | =back | |
248 | ||
249 | The field of Linguistics has expended a great deal of effort over the | |
250 | past century trying to find grammatical patterns which hold across | |
251 | languages; it's been a constant process | |
252 | of people making generalizations that should apply to all languages, | |
253 | only to find out that, all too often, these generalizations fail -- | |
254 | sometimes failing for just a few languages, sometimes whole classes of | |
255 | languages, and sometimes nearly every language in the world except | |
256 | English. Broad statistical trends are evident in what the "average | |
257 | language" is like as far as what its rules can look like, must look | |
258 | like, and cannot look like. But the "average language" is just as | |
259 | unreal a concept as the "average person" -- it runs up against the | |
260 | fact no language (or person) is, in fact, average. The wisdom of past | |
261 | experience leads us to believe that any given language can do whatever | |
262 | it wants, in any order, with appeal to any kind of grammatical | |
263 | categories wants -- case, number, tense, real or metaphoric | |
264 | characteristics of the things that words refer to, arbitrary or | |
265 | predictable classifications of words based on what endings or prefixes | |
266 | they can take, degree or means of certainty about the truth of | |
267 | statements expressed, and so on, ad infinitum. | |
268 | ||
269 | Mercifully, most localization tasks are a matter of finding ways to | |
270 | translate whole phrases, generally sentences, where the context is | |
271 | relatively set, and where the only variation in content is I<usually> | |
272 | in a number being expressed -- as in the example sentences above. | |
273 | Translating specific, fully-formed sentences is, in practice, fairly | |
274 | foolproof -- which is good, because that's what's in the phrasebooks | |
275 | that so many tourists rely on. Now, a given phrase (whether in a | |
276 | phrasebook or in a gettext lexicon) in one language I<might> have a | |
277 | greater or lesser applicability than that phrase's translation into | |
278 | another language -- for example, strictly speaking, in Arabic, the | |
279 | "your" in "Your query matched..." would take a different form | |
280 | depending on whether the user is male or female; so the Arabic | |
281 | translation "your[feminine] query" is applicable in fewer cases than | |
282 | the corresponding English phrase, which doesn't distinguish the user's | |
283 | gender. (In practice, it's not feasable to have a program know the | |
284 | user's gender, so the masculine "you" in Arabic is usually used, by | |
285 | default.) | |
286 | ||
287 | But in general, such surprises are rare when entire sentences are | |
288 | being translated, especially when the functional context is restricted | |
289 | to that of a computer interacting with a user either to convey a fact | |
290 | or to prompt for a piece of information. So, for purposes of | |
291 | localization, translation by phrase (generally by sentence) is both the | |
292 | simplest and the least problematic. | |
293 | ||
294 | =head2 Breaking gettext | |
295 | ||
296 | =over | |
297 | ||
298 | "It Has To Work." | |
299 | ||
300 | -- First Networking Truth, RFC 1925 | |
301 | ||
302 | =back | |
303 | ||
304 | Consider that sentences in a tourist phrasebook are of two types: ones | |
305 | like "How do I get to the marketplace?" that don't have any blanks to | |
306 | fill in, and ones like "How much do these ___ cost?", where there's | |
307 | one or more blanks to fill in (and these are usually linked to a | |
308 | list of words that you can put in that blank: "fish", "potatoes", | |
309 | "tomatoes", etc.) The ones with no blanks are no problem, but the | |
310 | fill-in-the-blank ones may not be really straightforward. If it's a | |
311 | Swahili phrasebook, for example, the authors probably didn't bother to | |
312 | tell you the complicated ways that the verb "cost" changes its | |
313 | inflectional prefix depending on the noun you're putting in the blank. | |
314 | The trader in the marketplace will still understand what you're saying if | |
315 | you say "how much do these potatoes cost?" with the wrong | |
316 | inflectional prefix on "cost". After all, I<you> can't speak proper Swahili, | |
317 | I<you're> just a tourist. But while tourists can be stupid, computers | |
318 | are supposed to be smart; the computer should be able to fill in the | |
319 | blank, and still have the results be grammatical. | |
320 | ||
321 | In other words, a phrasebook entry takes some values as parameters | |
322 | (the things that you fill in the blank or blanks), and provides a value | |
323 | based on these parameters, where the way you get that final value from | |
324 | the given values can, properly speaking, involve an arbitrarily | |
325 | complex series of operations. (In the case of Chinese, it'd be not at | |
326 | all complex, at least in cases like the examples at the beginning of | |
327 | this article; whereas in the case of Russian it'd be a rather complex | |
328 | series of operations. And in some languages, the | |
329 | complexity could be spread around differently: while the act of | |
330 | putting a number-expression in front of a noun phrase might not be | |
331 | complex by itself, it may change how you have to, for example, inflect | |
332 | a verb elsewhere in the sentence. This is what in syntax is called | |
333 | "long-distance dependencies".) | |
334 | ||
335 | This talk of parameters and arbitrary complexity is just another way | |
336 | to say that an entry in a phrasebook is what in a programming language | |
337 | would be called a "function". Just so you don't miss it, this is the | |
338 | crux of this article: I<A phrase is a function; a phrasebook is a | |
339 | bunch of functions.> | |
340 | ||
341 | The reason that using gettext runs into walls (as in the above | |
342 | second-person horror story) is that you're trying to use a string (or | |
343 | worse, a choice among a bunch of strings) to do what you really need a | |
344 | function for -- which is futile. Preforming (s)printf interpolation | |
345 | on the strings which you get back from gettext does allow you to do I<some> | |
346 | common things passably well... sometimes... sort of; but, to paraphrase | |
347 | what some people say about C<csh> script programming, "it fools you | |
348 | into thinking you can use it for real things, but you can't, and you | |
349 | don't discover this until you've already spent too much time trying, | |
350 | and by then it's too late." | |
351 | ||
352 | =head2 Replacing gettext | |
353 | ||
354 | So, what needs to replace gettext is a system that supports lexicons | |
355 | of functions instead of lexicons of strings. An entry in a lexicon | |
356 | from such a system should I<not> look like this: | |
357 | ||
358 | "J'ai trouv\xE9 %g fichiers dans %g r\xE9pertoires" | |
359 | ||
360 | [\xE9 is e-acute in Latin-1. Some pod renderers would | |
361 | scream if I used the actual character here. -- SB] | |
362 | ||
363 | but instead like this, bearing in mind that this is just a first stab: | |
364 | ||
365 | sub I_found_X1_files_in_X2_directories { | |
366 | my( $files, $dirs ) = @_[0,1]; | |
367 | $files = sprintf("%g %s", $files, | |
368 | $files == 1 ? 'fichier' : 'fichiers'); | |
369 | $dirs = sprintf("%g %s", $dirs, | |
370 | $dirs == 1 ? "r\xE9pertoire" : "r\xE9pertoires"); | |
371 | return "J'ai trouv\xE9 $files dans $dirs."; | |
372 | } | |
373 | ||
374 | Now, there's no particularly obvious way to store anything but strings | |
375 | in a gettext lexicon; so it looks like we just have to start over and | |
376 | make something better, from scratch. I call my shot at a | |
377 | gettext-replacement system "Maketext", or, in CPAN terms, | |
378 | Locale::Maketext. | |
379 | ||
380 | When designing Maketext, I chose to plan its main features in terms of | |
381 | "buzzword compliance". And here are the buzzwords: | |
382 | ||
383 | =head2 Buzzwords: Abstraction and Encapsulation | |
384 | ||
385 | The complexity of the language you're trying to output a phrase in is | |
386 | entirely abstracted inside (and encapsulated within) the Maketext module | |
387 | for that interface. When you call: | |
388 | ||
389 | print $lang->maketext("You have [quant,_1,piece] of new mail.", | |
390 | scalar(@messages)); | |
391 | ||
392 | you don't know (and in fact can't easily find out) whether this will | |
393 | involve lots of figuring, as in Russian (if $lang is a handle to the | |
394 | Russian module), or relatively little, as in Chinese. That kind of | |
395 | abstraction and encapsulation may encourage other pleasant buzzwords | |
396 | like modularization and stratification, depending on what design | |
397 | decisions you make. | |
398 | ||
399 | =head2 Buzzword: Isomorphism | |
400 | ||
401 | "Isomorphism" means "having the same structure or form"; in discussions | |
402 | of program design, the word takes on the special, specific meaning that | |
403 | your implementation of a solution to a problem I<has the same | |
404 | structure> as, say, an informal verbal description of the solution, or | |
405 | maybe of the problem itself. Isomorphism is, all things considered, | |
406 | a good thing -- it's what problem-solving (and solution-implementing) | |
407 | should look like. | |
408 | ||
409 | What's wrong the with gettext-using code like this... | |
410 | ||
411 | printf( $file_count == 1 ? | |
412 | ( $directory_count == 1 ? | |
413 | "Your query matched %g file in %g directory." : | |
414 | "Your query matched %g file in %g directories." ) : | |
415 | ( $directory_count == 1 ? | |
416 | "Your query matched %g files in %g directory." : | |
417 | "Your query matched %g files in %g directories." ), | |
418 | $file_count, $directory_count, | |
419 | ); | |
420 | ||
421 | is first off that it's not well abstracted -- these ways of testing | |
422 | for grammatical number (as in the expressions like C<foo == 1 ? | |
423 | singular_form : plural_form>) should be abstracted to each language | |
424 | module, since how you get grammatical number is language-specific. | |
425 | ||
426 | But second off, it's not isomorphic -- the "solution" (i.e., the | |
427 | phrasebook entries) for Chinese maps from these four English phrases to | |
428 | the one Chinese phrase that fits for all of them. In other words, the | |
429 | informal solution would be "The way to say what you want in Chinese is | |
430 | with the one phrase 'For your question, in Y directories you would | |
431 | find X files'" -- and so the implemented solution should be, | |
432 | isomorphically, just a straightforward way to spit out that one | |
433 | phrase, with numerals properly interpolated. It shouldn't have to map | |
434 | from the complexity of other languages to the simplicity of this one. | |
435 | ||
436 | =head2 Buzzword: Inheritance | |
437 | ||
438 | There's a great deal of reuse possible for sharing of phrases between | |
439 | modules for related dialects, or for sharing of auxiliary functions | |
440 | between related languages. (By "auxiliary functions", I mean | |
441 | functions that don't produce phrase-text, but which, say, return an | |
442 | answer to "does this number require a plural noun after it?". Such | |
443 | auxiliary functions would be used in the internal logic of functions | |
444 | that actually do produce phrase-text.) | |
445 | ||
446 | In the case of sharing phrases, consider that you have an interface | |
447 | already localized for American English (probably by having been | |
448 | written with that as the native locale, but that's incidental). | |
449 | Localizing it for UK English should, in practical terms, be just a | |
450 | matter of running it past a British person with the instructions to | |
451 | indicate what few phrases would benefit from a change in spelling or | |
452 | possibly minor rewording. In that case, you should be able to put in | |
453 | the UK English localization module I<only> those phrases that are | |
454 | UK-specific, and for all the rest, I<inherit> from the American | |
455 | English module. (And I expect this same situation would apply with | |
456 | Brazilian and Continental Portugese, possbily with some I<very> | |
457 | closely related languages like Czech and Slovak, and possibly with the | |
458 | slightly different "versions" of written Mandarin Chinese, as I hear exist in | |
459 | Taiwan and mainland China.) | |
460 | ||
461 | As to sharing of auxiliary functions, consider the problem of Russian | |
462 | numbers from the beginning of this article; obviously, you'd want to | |
463 | write only once the hairy code that, given a numeric value, would | |
464 | return some specification of which case and number a given quanitified | |
465 | noun should use. But suppose that you discover, while localizing an | |
466 | interface for, say, Ukranian (a Slavic language related to Russian, | |
467 | spoken by several million people, many of whom would be relieved to | |
468 | find that your Web site's or software's interface is available in | |
469 | their language), that the rules in Ukranian are the same as in Russian | |
470 | for quantification, and probably for many other grammatical functions. | |
471 | While there may well be no phrases in common between Russian and | |
472 | Ukranian, you could still choose to have the Ukranian module inherit | |
473 | from the Russian module, just for the sake of inheriting all the | |
474 | various grammatical methods. Or, probably better organizationally, | |
475 | you could move those functions to a module called C<_E_Slavic> or | |
476 | something, which Russian and Ukranian could inherit useful functions | |
477 | from, but which would (presumably) provide no lexicon. | |
478 | ||
479 | =head2 Buzzword: Concision | |
480 | ||
481 | Okay, concision isn't a buzzword. But it should be, so I decree that | |
482 | as a new buzzword, "concision" means that simple common things should | |
483 | be expressible in very few lines (or maybe even just a few characters) | |
484 | of code -- call it a special case of "making simple things easy and | |
485 | hard things possible", and see also the role it played in the | |
486 | MIDI::Simple language, discussed elsewhere in this issue [TPJ#13]. | |
487 | ||
488 | Consider our first stab at an entry in our "phrasebook of functions": | |
489 | ||
490 | sub I_found_X1_files_in_X2_directories { | |
491 | my( $files, $dirs ) = @_[0,1]; | |
492 | $files = sprintf("%g %s", $files, | |
493 | $files == 1 ? 'fichier' : 'fichiers'); | |
494 | $dirs = sprintf("%g %s", $dirs, | |
495 | $dirs == 1 ? "r\xE9pertoire" : "r\xE9pertoires"); | |
496 | return "J'ai trouv\xE9 $files dans $dirs."; | |
497 | } | |
498 | ||
499 | You may sense that a lexicon (to use a non-committal catch-all term for a | |
500 | collection of things you know how to say, regardless of whether they're | |
501 | phrases or words) consisting of functions I<expressed> as above would | |
502 | make for rather long-winded and repetitive code -- even if you wisely | |
503 | rewrote this to have quantification (as we call adding a number | |
504 | expression to a noun phrase) be a function called like: | |
505 | ||
506 | sub I_found_X1_files_in_X2_directories { | |
507 | my( $files, $dirs ) = @_[0,1]; | |
508 | $files = quant($files, "fichier"); | |
509 | $dirs = quant($dirs, "r\xE9pertoire"); | |
510 | return "J'ai trouv\xE9 $files dans $dirs."; | |
511 | } | |
512 | ||
513 | And you may also sense that you do not want to bother your translators | |
514 | with having to write Perl code -- you'd much rather that they spend | |
515 | their I<very costly time> on just translation. And this is to say | |
516 | nothing of the near impossibility of finding a commercial translator | |
517 | who would know even simple Perl. | |
518 | ||
519 | In a first-hack implementation of Maketext, each language-module's | |
520 | lexicon looked like this: | |
521 | ||
522 | %Lexicon = ( | |
523 | "I found %g files in %g directories" | |
524 | => sub { | |
525 | my( $files, $dirs ) = @_[0,1]; | |
526 | $files = quant($files, "fichier"); | |
527 | $dirs = quant($dirs, "r\xE9pertoire"); | |
528 | return "J'ai trouv\xE9 $files dans $dirs."; | |
529 | }, | |
530 | ... and so on with other phrase => sub mappings ... | |
531 | ); | |
532 | ||
533 | but I immediately went looking for some more concise way to basically | |
534 | denote the same phrase-function -- a way that would also serve to | |
535 | concisely denote I<most> phrase-functions in the lexicon for I<most> | |
536 | languages. After much time and even some actual thought, I decided on | |
537 | this system: | |
538 | ||
539 | * Where a value in a %Lexicon hash is a contentful string instead of | |
540 | an anonymous sub (or, conceivably, a coderef), it would be interpreted | |
541 | as a sort of shorthand expression of what the sub does. When accessed | |
542 | for the first time in a session, it is parsed, turned into Perl code, | |
543 | and then eval'd into an anonymous sub; then that sub replaces the | |
544 | original string in that lexicon. (That way, the work of parsing and | |
545 | evaling the shorthand form for a given phrase is done no more than | |
546 | once per session.) | |
547 | ||
548 | * Calls to C<maketext> (as Maketext's main function is called) happen | |
549 | thru a "language session handle", notionally very much like an IO | |
550 | handle, in that you open one at the start of the session, and use it | |
551 | for "sending signals" to an object in order to have it return the text | |
552 | you want. | |
553 | ||
554 | So, this: | |
555 | ||
556 | $lang->maketext("You have [quant,_1,piece] of new mail.", | |
557 | scalar(@messages)); | |
558 | ||
559 | basically means this: look in the lexicon for $lang (which may inherit | |
560 | from any number of other lexicons), and find the function that we | |
561 | happen to associate with the string "You have [quant,_1,piece] of new | |
562 | mail" (which is, and should be, a functioning "shorthand" for this | |
563 | function in the native locale -- English in this case). If you find | |
564 | such a function, call it with $lang as its first parameter (as if it | |
565 | were a method), and then a copy of scalar(@messages) as its second, | |
566 | and then return that value. If that function was found, but was in | |
567 | string shorthand instead of being a fully specified function, parse it | |
568 | and make it into a function before calling it the first time. | |
569 | ||
570 | * The shorthand uses code in brackets to indicate method calls that | |
571 | should be performed. A full explanation is not in order here, but a | |
572 | few examples will suffice: | |
573 | ||
574 | "You have [quant,_1,piece] of new mail." | |
575 | ||
576 | The above code is shorthand for, and will be interpreted as, | |
577 | this: | |
578 | ||
579 | sub { | |
580 | my $handle = $_[0]; | |
581 | my(@params) = @_; | |
582 | return join '', | |
583 | "You have ", | |
584 | $handle->quant($params[1], 'piece'), | |
585 | "of new mail."; | |
586 | } | |
587 | ||
588 | where "quant" is the name of a method you're using to quantify the | |
589 | noun "piece" with the number $params[0]. | |
590 | ||
591 | A string with no brackety calls, like this: | |
592 | ||
593 | "Your search expression was malformed." | |
594 | ||
595 | is somewhat of a degerate case, and just gets turned into: | |
596 | ||
597 | sub { return "Your search expression was malformed." } | |
598 | ||
599 | However, not everything you can write in Perl code can be written in | |
600 | the above shorthand system -- not by a long shot. For example, consider | |
601 | the Italian translator from the beginning of this article, who wanted | |
602 | the Italian for "I didn't find any files" as a special case, instead | |
603 | of "I found 0 files". That couldn't be specified (at least not easily | |
604 | or simply) in our shorthand system, and it would have to be written | |
605 | out in full, like this: | |
606 | ||
607 | sub { # pretend the English strings are in Italian | |
608 | my($handle, $files, $dirs) = @_[0,1,2]; | |
609 | return "I didn't find any files" unless $files; | |
610 | return join '', | |
611 | "I found ", | |
612 | $handle->quant($files, 'file'), | |
613 | " in ", | |
614 | $handle->quant($dirs, 'directory'), | |
615 | "."; | |
616 | } | |
617 | ||
618 | Next to a lexicon full of shorthand code, that sort of sticks out like a | |
619 | sore thumb -- but this I<is> a special case, after all; and at least | |
620 | it's possible, if not as concise as usual. | |
621 | ||
622 | As to how you'd implement the Russian example from the beginning of | |
623 | the article, well, There's More Than One Way To Do It, but it could be | |
624 | something like this (using English words for Russian, just so you know | |
625 | what's going on): | |
626 | ||
627 | "I [quant,_1,directory,accusative] scanned." | |
628 | ||
629 | This shifts the burden of complexity off to the quant method. That | |
630 | method's parameters are: the numeric value it's going to use to | |
631 | quantify something; the Russian word it's going to quantify; and the | |
632 | parameter "accusative", which you're using to mean that this | |
633 | sentence's syntax wants a noun in the accusative case there, although | |
634 | that quantification method may have to overrule, for grammatical | |
635 | reasons you may recall from the beginning of this article. | |
636 | ||
637 | Now, the Russian quant method here is responsible not only for | |
638 | implementing the strange logic necessary for figuring out how Russian | |
639 | number-phrases impose case and number on their noun-phrases, but also | |
640 | for inflecting the Russian word for "directory". How that inflection | |
641 | is to be carried out is no small issue, and among the solutions I've | |
642 | seen, some (like variations on a simple lookup in a hash where all | |
643 | possible forms are provided for all necessary words) are | |
644 | straightforward but I<can> become cumbersome when you need to inflect | |
645 | more than a few dozen words; and other solutions (like using | |
646 | algorithms to model the inflections, storing only root forms and | |
647 | irregularities) I<can> involve more overhead than is justifiable for | |
648 | all but the largest lexicons. | |
649 | ||
650 | Mercifully, this design decision becomes crucial only in the hairiest | |
651 | of inflected languages, of which Russian is by no means the I<worst> case | |
652 | scenario, but is worse than most. Most languages have simpler | |
653 | inflection systems; for example, in English or Swahili, there are | |
654 | generally no more than two possible inflected forms for a given noun | |
655 | ("error/errors"; "kosa/makosa"), and the | |
656 | rules for producing these forms are fairly simple -- or at least, | |
657 | simple rules can be formulated that work for most words, and you can | |
658 | then treat the exceptions as just "irregular", at least relative to | |
659 | your ad hoc rules. A simpler inflection system (simpler rules, fewer | |
660 | forms) means that design decisions are less crucial to maintaining | |
661 | sanity, whereas the same decisions could incur | |
662 | overhead-versus-scalability problems in languages like Russian. It | |
663 | may I<also> be likely that code (possibly in Perl, as with | |
664 | Lingua::EN::Inflect, for English nouns) has already | |
665 | been written for the language in question, whether simple or complex. | |
666 | ||
667 | Moreover, a third possibility may even be simpler than anything | |
668 | discussed above: "Just require that all possible (or at least | |
669 | applicable) forms be provided in the call to the given language's quant | |
670 | method, as in:" | |
671 | ||
672 | "I found [quant,_1,file,files]." | |
673 | ||
674 | That way, quant just has to chose which form it needs, without having | |
675 | to look up or generate anything. While possibly not optimal for | |
676 | Russian, this should work well for most other languages, where | |
677 | quantification is not as complicated an operation. | |
678 | ||
679 | =head2 The Devil in the Details | |
680 | ||
681 | There's plenty more to Maketext than described above -- for example, | |
682 | there's the details of how language tags ("en-US", "i-pwn", "fi", | |
683 | etc.) or locale IDs ("en_US") interact with actual module naming | |
684 | ("BogoQuery/Locale/en_us.pm"), and what magic can ensue; there's the | |
685 | details of how to record (and possibly negotiate) what character | |
686 | encoding Maketext will return text in (UTF8? Latin-1? KOI8?). There's | |
687 | the interesting fact that Maketext is for localization, but nowhere | |
688 | actually has a "C<use locale;>" anywhere in it. For the curious, | |
689 | there's the somewhat frightening details of how I actually | |
690 | implement something like data inheritance so that searches across | |
691 | modules' %Lexicon hashes can parallel how Perl implements method | |
692 | inheritance. | |
693 | ||
694 | And, most importantly, there's all the practical details of how to | |
695 | actually go about deriving from Maketext so you can use it for your | |
696 | interfaces, and the various tools and conventions for starting out and | |
697 | maintaining individual language modules. | |
698 | ||
699 | That is all covered in the documentation for Locale::Maketext and the | |
700 | modules that come with it, available in CPAN. After having read this | |
701 | article, which covers the why's of Maketext, the documentation, | |
702 | which covers the how's of it, should be quite straightfoward. | |
703 | ||
704 | =head2 The Proof in the Pudding: Localizing Web Sites | |
705 | ||
706 | Maketext and gettext have a notable difference: gettext is in C, | |
707 | accessible thru C library calls, whereas Maketext is in Perl, and | |
708 | really can't work without a Perl interpreter (although I suppose | |
709 | something like it could be written for C). Accidents of history (and | |
710 | not necessarily lucky ones) have made C++ the most common language for | |
711 | the implementation of applications like word processors, Web browsers, | |
712 | and even many in-house applications like custom query systems. Current | |
713 | conditions make it somewhat unlikely that the next one of any of these | |
714 | kinds of applications will be written in Perl, albeit clearly more for | |
715 | reasons of custom and inertia than out of consideration of what is the | |
716 | right tool for the job. | |
717 | ||
718 | However, other accidents of history have made Perl a well-accepted | |
719 | language for design of server-side programs (generally in CGI form) | |
720 | for Web site interfaces. Localization of static pages in Web sites is | |
721 | trivial, feasable either with simple language-negotiation features in | |
722 | servers like Apache, or with some kind of server-side inclusions of | |
723 | language-appropriate text into layout templates. However, I think | |
724 | that the localization of Perl-based search systems (or other kinds of | |
725 | dynamic content) in Web sites, be they public or access-restricted, | |
726 | is where Maketext will see the greatest use. | |
727 | ||
728 | I presume that it would be only the exceptional Web site that gets | |
729 | localized for English I<and> Chinese I<and> Italian I<and> Arabic | |
730 | I<and> Russian, to recall the languages from the beginning of this | |
731 | article -- to say nothing of German, Spanish, French, Japanese, | |
732 | Finnish, and Hindi, to name a few languages that benefit from large | |
733 | numbers of programmers or Web viewers or both. | |
734 | ||
735 | However, the ever-increasing internationalization of the Web (whether | |
736 | measured in terms of amount of content, of numbers of content writers | |
737 | or programmers, or of size of content audiences) makes it increasingly | |
738 | likely that the interface to the average Web-based dynamic content | |
739 | service will be localized for two or maybe three languages. It is my | |
740 | hope that Maketext will make that task as simple as possible, and will | |
741 | remove previous barriers to localization for languages dissimilar to | |
742 | English. | |
743 | ||
744 | __END__ | |
745 | ||
746 | Sean M. Burke (sburkeE<64>cpan.org) has a Master's in linguistics | |
747 | from Northwestern University; he specializes in language technology. | |
748 | Jordan Lachler (lachlerE<64>unm.edu) is a PhD student in the Department of | |
749 | Linguistics at the University of New Mexico; he specializes in | |
750 | morphology and pedagogy of North American native languages. | |
751 | ||
752 | =head2 References | |
753 | ||
754 | Alvestrand, Harald Tveit. 1995. I<RFC 1766: Tags for the | |
755 | Identification of Languages.> | |
756 | C<ftp://ftp.isi.edu/in-notes/rfc1766.txt> | |
757 | [Now see RFC 3066.] | |
758 | ||
759 | Callon, Ross, editor. 1996. I<RFC 1925: The Twelve | |
760 | Networking Truths.> | |
761 | C<ftp://ftp.isi.edu/in-notes/rfc1925.txt> | |
762 | ||
763 | Drepper, Ulrich, Peter Miller, | |
764 | and FranE<ccedil>ois Pinard. 1995-2001. GNU | |
765 | C<gettext>. Available in C<ftp://prep.ai.mit.edu/pub/gnu/>, with | |
766 | extensive docs in the distribution tarball. [Since | |
767 | I wrote this article in 1998, I now see that the | |
768 | gettext docs are now trying more to come to terms with | |
769 | plurality. Whether useful conclusions have come from it | |
770 | is another question altogether. -- SMB, May 2001] | |
771 | ||
772 | Forbes, Nevill. 1964. I<Russian Grammar.> Third Edition, revised | |
773 | by J. C. Dumbreck. Oxford University Press. | |
774 | ||
775 | =cut | |
776 | ||
777 | #End | |
778 |