[web-devel] [Yesod] rendering different templates for different languages

Michael Snoyman michael at snoyman.com
Mon Feb 21 09:21:27 CET 2011


There's no way to automatically pluralize an arbitrary word. English
is the best example of this: cat -> cats, octopus -> octopi, cherry ->
cherries. There's no universal rule governing this; each word needs to
be specified manually by the translator. Hebrew is a little easier,
but still has plenty of exceptions: sometimes male nouns have female
plural endings and vice-versa. And don't even get into compound nouns.

As tempting as it is, I don't think there are any ways to make this a
generic process: each language likely needs to have its own specific
logic applied.

Michael

On Mon, Feb 21, 2011 at 9:37 AM, Ian Duncan <iand675 at gmail.com> wrote:
> My reasoning for typeclasses is this:
> Keep in mind this is just a mental exercise, not saying that this would
> compile...
> Suppose for your example we wanted to translate it into french:
> -- Nouns for languages have some way to decide gender, if that matters for
> the given language
> class Translate a where
> toPlural :: Noun a -> Int -> [Words a]
> data French = French {
>     ... stuff ...
>     languageCode = "FR"
> }
> instance Translate French where
> toPlural = frenchPluralizationEngine
> From there we go with something like your .trans files: we give each
> language the extension according to their language code. So basket.fr is the
> french translation:
> Vous avez #{toPlural 'chien' (maleDogs myBasket)} et #{femaleCats 'chatte'
> myBasket)} dans votre charrette.
> This would be pluralized by chien being run through a pluralization engine
> if necessary
> I know my thinking is rough so far since you've obviously given it a lot of
> thought, but does my concept make sense?
> --
> Ian Duncan
>
> On Monday, February 21, 2011 at 1:20 AM, Michael Snoyman wrote:
>
> Let's look at a more concrete example: you have an online store
> selling male dogs and female cats. So you would have:
>
> data Basket = Basket { maleDogs :: Int, femaleCats :: Int }
>
> What you need is a function such as:
>
> renderBasket :: Basket -> String
>
> for each language. In English, this could be something like:
>
> pluralize :: Int -> (String, String) -> String
> pluralize 1 (x, _) = x
> pluralize _ (_, x) = x
>
> renderBasket (Basket dogs cats) = concat
> [ "You have decided to purchase "
> , show dogs
> , pluralize dogs ("dog", "dogs")
> , " and "
> , show cats
> , pluralize cats ("cat", "cats")
> ]
>
> In Hebrew, some words (like years) have a singular, plural *and* dual
> form, so pluralize for Hebrew may look like:
>
> pluralize :: Int -> (String, Maybe String, String) -> String
> pluralize 1 (x, _, _) = x ++ " אחד" -- in Hebrew, the "one" comes
> after the word, all other numbers before
> pluralize 2 (_, Just x, _) = x -- for dual form, you never show
> the number, it is assumed
> pluralize i (_, _, x) = show i ++ " " ++ x -- for the plural, put
> the number before
>
> If we could build up a library in Haskell of such helper functions, I
> think it would make translating applications much simpler. But this is
> the point where we would need a lot of collaboration: I can help out
> on English and Hebrew (and if I still remember it, Spanish), but I
> don't know a thing about Japanese, Russian, or most other languages in
> the world.
>
> I'm not sure how much it would really help to use typeclasses here,
> however. I think for the most part it will just be an issue of having
> a separate module for each language. What I'd *really* like to figure
> out is how to make a nice, easy-to-use wrapper around all of this for
> translators, who will likely not know any Haskell. Perhaps a language
> similar to Hamlet:
>
> # strings-english.trans
> Hello: Hello
> Person name age: ##{name} is #{age} #{pluralize age "year" "years"} old.
> Basket dogs cats: You have purchased #{dogs} #{pluralize dogs
> "dog" "dogs"} and #{cats} #{pluralize cats "cat" cats"}.
>
> Michael
>
> On Mon, Feb 21, 2011 at 8:57 AM, Ian Duncan <iand675 at gmail.com> wrote:
>
> And of course in some languages such as Japanese, there are barely any
> gender distinctions or such things as pluralization at all. Perhaps we need
> pluralization, conjugation, and 'genderization' typeclasses with instances
> defined for different language datatypes?
>
> --
> Ian Duncan
>
> On Monday, February 21, 2011 at 12:46 AM, Michael Snoyman wrote:
>
> The other day I was speaking with a woman on the train. She was
> telling me about her daughters. I wanted to ask her how old they are,
> but I got the pluralization wrong and instead of saying "bnot kama"
> (plural) I said "bat kama," (singular) to which she responded 36.
>
> tl;dr: You can offend people just was well with pluralization issues
> as with gender issues.
>
> Michael
>
> On Mon, Feb 21, 2011 at 8:40 AM, Max Cantor <mxcantor at gmail.com> wrote:
>
> Of course, you just pointed out one of the big difficulties with i18n.  I
> dont think you're wife would take kindly to you referring to her in the male
> gender.  so now, you need the person's gender too.  i18n is hard :(  the
> whole would should switch to esperanto.
>
> max
>
> On Feb 21, 2011, at 2:25 PM, Michael Snoyman wrote:
>
> A proper i18n solution is high on my wish list right now, but I've
> purposely avoided implementing one so far since I'd rather wait until
> I think we have a good solution as opposed to implementing an
> acceptable solution now. But let me share my ideas, it might help you
> out here.
>
> In general, it's very uncommon that you need a completely separate set
> of templates for each language. Your markup, classes, styles, and
> logic will likely be identical for each language, and creating a
> separate template for each will just result in a lot of pain in the
> long run. Instead, you're likely better off having a single template
> and just translating strings.
>
> I've blogged about this before[1]. My idea is to use a datatype for
> your translatable strings, and then have a function that takes a
> language and a value and returns the translated string. A simple
> example:
>
>    data Strings = Hello | Person String Int
>    toEnglish Hello = "Hello"
>    toEnglish (Person name age) = name ++ " is " ++ show age ++ "
> years old" -- obviously need to check if person is 1 year old and
> correct
>
>    toHebrew Hello = "שלום"
>    toHebrew (Person name age) = name ++ " הוא בן " ++ show age ++ " שנים"
>
> The nice thing about this approach is you have the full power of
> Haskell to address typical translation issues, such as pluralization,
> word order and gender matching. (As a counter example, at work, we use
> XSLT for this, and then you get the full power of XSLT for solving the
> problem ::cringe::.)
>
> You can then use the languages[2] function from Yesod to help you out:
>
>    getRenderString = chooseFunc `fmap` languages
>       where
>         chooseFunc [] = toEnglish -- default language
>         chooseFunc ("en":_) = toEnglish
>         chooseFunc ("he":_) = toHebrew
>         chooseFunc (_:x) = chooseFunc x
>
> Then you can write a handler function like:
>
> getPersonR name age = do
>    render <- getRenderString
>    defaultLayout [$hamlet|
> <h1>#{render Hello}
> <p>#{render $ Person name age}
> |]
>
> Which will work for English and Hebrew just fine. Ideally, I would
> like to add support to Hamlet for this directly, involving a String
> rendering function similar to the URL rendering function already in
> place. But for the moment, this should work.
>
> I'd love to hear peoples opinions about this.
>
> Michael
>
> [1] http://docs.yesodweb.com/blog/i18n-in-haskell
> [2]
> http://hackage.haskell.org/packages/archive/yesod-core/0.7.0.1/doc/html/Yesod-Request.html#v:languages
>
> On Sun, Feb 20, 2011 at 11:19 PM, Dmitry Kurochkin
> <dmitry.kurochkin at gmail.com> wrote:
>
> Hi all.
>
> I want a handler to render different templates for different languages.
> I have getCurrentLanguage function and now I try to do something like:
>
>    getRootR = do
>        currentLanguage <- getCurrentLanguage
>        defaultLayout $ do
>            addWidget $(widgetFile $ currentLanguage ++ "/homepage")
>
> This results in:
>
>    GHC stage restriction: `currentLanguage'
>      is used in a top-level splice or annotation,
>      and must be imported, not defined locally
>
> This makes sense to me, because TH is calculated at compile time. I
> would like to hear ideas how to work around this restriction. Perhaps
> there is an existing solution in Yesod?
>
> At the moment, the best I could think of is smth like this:
>
>    getRootR = do
>        currentLanguage <- getCurrentLanguage
>        defaultLayout $ do
>            case currentLanguage of
>                "en" -> addWidget $(widgetFile  "en/homepage")
>                ... and so on for each language ...
>
> Obviously, this is not a solution taking in account that there are many
> languages and many handlers.
>
> I was considering creating a global (template file name -> rendered
> template) map. But I am not sure this is really feasible.
>
> Regards,
>  Dmitry
>
> _______________________________________________
> web-devel mailing list
> web-devel at haskell.org
> http://www.haskell.org/mailman/listinfo/web-devel
>
> _______________________________________________
> web-devel mailing list
> web-devel at haskell.org
> http://www.haskell.org/mailman/listinfo/web-devel
>
> _______________________________________________
> web-devel mailing list
> web-devel at haskell.org
> http://www.haskell.org/mailman/listinfo/web-devel
>
>



More information about the web-devel mailing list