[i18n] future 2.x plan

[vicb]

Features planned for 2.0:

• gender
• pluralization
• runtime compiler
• offline compiler
• xliff support
• hash message ids
• i18n messages

fix issues:

• xmb/xtb support [Feature][i18n] .xmb/.xtb support #9067
• do not extract empty msgs [i18n] do not extract message when there is not child nodes #8802
• allow escaping ICU messages - [i18n] escaping in ICU message #9286
• fix markup for plural/gender [i18n] {messages.length, plural, ...} should not be expanded to a list #9072 (PR feat(I18N Expander): do not add extra <ul> & <li> around ICU mess… #9283)
• improve the extractor [i18n]: add support for implicit translation #9071 (PR feat(I18nExtractor): Add file paths to error messages #9177)
• fix ICU format [i18n] ICU format for plural #9094 (fixed in 5267115)

fix post 2.0:

• Template Parser: Error when passing object literal as pipe parameter. False-Positive: Parser Error: Missing expected } at the end of the expression #9571
• Improve Html Parser Html Parser improvements #9340
• Opt out of translation I18n: Opting out of translation #7814
• Provide NgLocalization implementation per locale
• ICU plural format [i18n] ICU plural format - support # and offset #9117
• allow ICU messages in attributes
• implement ICU ordinal messages
• use translation strings outside of a template i18n: Able to use translation strings outside a template #11405

[thelgevold]

@vicb I have seen a few postings about i18n in the angular repo. How does this relate to https://github.com/angular/i18n? Is this a new initiative?

[robwormald]

@thelgevold like most things in ng2, the standalone repos existed for early prototyping. major work is typically done inside of this one.

[marcalj]

Hi, there's any document about how it will work?
I'm wondering if I can write keys in markup and normal text in translation files like I'm using with angular-translate.

{{ 'HOMEPAGE__ACTION__SIGNUP' | translate }}

Coding like this, one external person can edit all supported languages (since they reside in JSON/po file, one file per language). If the main language is in the markup, it blocks all revisions until a developer can change the string in the markup code.

Thanks!

[vicb]

@marcalj The (outdated) spec is https://docs.google.com/document/d/1mwyOFsAD-bPoXTk3Hthq0CAcGXCUw-BtTJMR4nGTY-0/edit.

Your use case is supported.

[marcalj]

Thanks Victor! Any plans to include number abbreviations? It's something i18n context aware.

[vicb]

@marcalj could you please create a separate issue with more details ? Thanks (but probably not for 2.0 anyways)

[dagerber]

Is or will there be support for dynamic translations (loaded from a server)?
Any plans to document i18n in the near future?
Thanks!

[vicb]

The current plan is to generate a binary per locale.

We will work on the doc as the features become available. We should definitely have a minimal doc for the 2.0 release.

[jaska45]

What do you mean by "binary per locale". Is the deployed resource file going to be binary little bit like PO/MO where MO is used insted of PO to improve performance. There are also bad examples in binary deployment such as XAML/BAML.

I am not very happy to the current Anguar localization methods. These are the reasons:

• Several different APIs and methods.
• You have to modify your templates so much from original that templates become hard to read and maintain.
• You have to manually update the original resource file and make sure it is in sync with the IDs used in the template

What we really need is

• One single official localization method with solid resource format (XML or JSON based)
• Minimal affect to templates. For example only add "localize" attribute into root element and possible "noloc" attributes into those elements that are not localized.
• Ability to add localization comments to the strings in the templates.
• A tool that scans source code extracting orignal strings and comments into resource file. Use as permanent ids as possible. For example use id attributes or if such is not provided use combination of location and original string value.
• This resource file should be able to store plain strings, strings with patterns ({{...}}), pluralized strings and gender strings little bit like Android's value files.
• Easy way to package localized resources and to download them as needed
• Ability to change language without reload.

I am ready to help if you guys need some help.

[vicb]

@jaska45 most your wishes should come true. One exception: "Ability to change language without reload.". Changing a language will imply bootstrapping a new app.

[jaska45]

Thanks. Sound very good. Language change is something that occurs seldom and in that case it is acceptable that some state is lost. Definetely lowest priority in my list and not worth spending too much efforts.

[choeller]

@vicb I'm not sure if I understand this "binary per locale" correctly, but would that mean, that it's not possible to add translations at runtime? We are currently building SAAS with tenant support, where each tenant can extend our system by providing their own set of Form-Definitions and their own set of translations, so in our case it would be really important to be able to add translations at runtime.

[nickspoons]

The @angular/common DatePipe (date_pipe.ts) currently contains this TODO on line 15:

// TODO: move to a global configurable location along with other i18n components.
var defaultLocale: string = 'en-US';

Is the ability to pass in a locale string to the DatePipe included in the i18n plans for the 2.0 release?

[marcalj]

@vicb Our app sets the frontend language once we fetch user object from network after bootstrapping the application. So, did you define if this process of re-boostrapping would be easy?

[MrCroft]

There's one bit of information I can't find anywhere, just to confirm and be able to sleep at night :)
The official i18n implementation will support gettext/po files, right? That's very common, so I'm guessing it should be available, we won't be stuck with jsons and xtb only, am I right? I mean gettext is so common that it doesn't even need mentioning, it's always there out of the box in any i18n implementation, that's what I'm thinking.

[vicb]

There is no plan to support po for 2.0. You should be able to convert from/to xliff.

sorry for your nights !

[intellix]

What's the thought behind storing strings in XML format? Never ran into any issues extracting to POT and compiling PO to JSON in angular-gettext. It's served us extremely well:

// en-GB.json, cached into localStorage after first request
"en-GB": {
    "Hello {{firstName}}": "{{firstName}}, Gutentag"
}

Random question: any thoughts about translated URLs in the router?

[vicb]

Random question: any thoughts about translated URLs in the router?

@vsavkin what are the plans ?

[vsavkin]

We are not planning to provide any support for translated URLs until the final is out. We had some ideas on ways to do that after that, but we haven't made any decisions.

[concept-hf]

ng2 i18n gave me a headache. I was looking for i18n directive in exports :D to no end... Looking through the test specs, and code the lightning struck me:

The template compiler itself will transform and localize the template.

Am I right, that the basic workflow for i18n will be like this:

1. Mark elements with i18n or enclose html parts in i18n comments.
2. Run somehow the extractor (ng cli?) to produce xliff.
3. Pass that to translation and receive back the translated binary.
4. Configure angular at bootstrap somehow to use the wanted language binary file.
5. The template compiler itself will transform and localize the template.

Because if it is, for the time being I will use a custom "i18n" directive to mimic the very-very basics of translation until 2.0 hits.

Is this the intended workflow for i18n in angular2 in the future?

[vicb]

@concept-hf that is right, see the doc that I linker earlier in this discussion.

You'll run ng-xi18n to create the xliff.

We have a couple of directives and filter for ICU messages

[marcalj]

@vicb Will be available in the upcoming RC5? :)

[vicb]

Depends on when RC5 is... otherwise will be in RC6

[esinek]

What version of XLIFF will be supported? 1.2 or 2.0 or both?

[ghidoz]

I found a problem with plurals and select: are you sure it is correctly implemented?

If I have just normal string, I can export them with the ng-xi18n command, but if I add a string with plurals or select, it does not work.

Using

<span i18n>{user.receivedReports, plural, =0 {No spam/scam reports} =1 {One spam/scam report} other {# spam/scam reports}}</span>

and

<span i18n>{currentConversation.user.online, select, true {Online} false {Offline}}</span>

I get many parsing errors like this:

Unexpected closing tag "li" ("plural, =0 {No spam/scam reports} =1 {One spam/scam report} other {# spam/scam reports}}</span>
    [ERROR ->]</li>

Using the comments syntax I get no error while extracting:

<!--i18n: select test-->
{user.online, select, true {Online} false {Offline}}
<!--/i18n-->

<!--i18n: plural test-->
{user.receivedReports, plural, =0 {No spam/scam reports} =1 {One spam/scam report} other {# spam/scam reports}}
<!--/i18n-->

but then what I get in the XLF file is just empty tags:

<trans-unit id="2b10577df0a3d78b7e65c07d5760524f881f390a" datatype="html">
    <source>
        <x id="ICU"/>
    </source>
    <target/>
    <note priority="1" from="description">plural test</note>
</trans-unit>

<trans-unit id="e61681024eadd7f72e33f053f04c128910c419a2" datatype="html">
    <source>
        <x id="ICU"/>
    </source>
    <target/>
    <note priority="1" from="description">select test</note>
</trans-unit>

I am doing something wrong?

Edit: using --i18nFormat=xmb throws the same error using the attribute version, while it seems to work with the comment version:

<msg id="3c8aef599bb96e303600aa44617acd0a7d602147">{user.receivedReports, plural, =0 {No spam/scam reports} =1 {One spam/scam report} other {# spam/scam reports} }</msg>
<msg id="06fe711b55463e3ee28d72239d77e9a50da62d86">{currentConversation.user.online, select, true {Online} false {Offline} }</msg> 

[zh99998]

how to translate Date/Time ?

[martinmcwhorter]

The date is translated using the date pipe

…

On Wed, 4 Jan 2017, 08:12 zh99998, ***@***.***> wrote: how to translate Date/Time ? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#9104 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABI5-gHZiWOREPwucdqwRhVIs9xwMgorks5rO1RngaJpZM4IxhKY> .

[dietergeerts]

@feloy , your example only let's you switch the lang on the login page, what if you want to switch the lang when logged in, and without a browser refresh? How to manually destroy the application and bootstrap it again?

[rolandoldengarm]

@dietergeerts that's only possible with JIT, not with AOT.

[dietergeerts]

@rolandoldengarm , it's also possible with AOT if you load your translations on application load. I'm currently using something in between ngx-translate and the official i18n, in the sense that we use pipes to get the translations (pure ones), but we only load one language at startup, and switching the language requires a reboot of the application (If only we had routing reload, as the i18n setup is done in one of the root resolves)

[Thomas-Kuipers]

@vicb

I'll try to take a look at what's needed e/o this week and come with a plan.

Do you have any updates on the roadmap? Specifically for an i18n implementation for translations outside of templates.

[thelgevold]

I am also very interested in the capability to translate strings in TypeScript. Any ETA on this?

[ocombe]

@Thomas-Kuipers @thelgevold I'm working on it, and it will be available in 4.1 or 4.2 (most likely)

[rayer4u]

@ocombe
how to deal with static text in interpolation, which can't use select translation.

<div i18n>{{exercise._desc ? exercise._desc : 'empty'}}</div>

string 'empty' can't be translate

[ocombe]

@rayer4u this is not the subject of this issue, for support questions you should use stackoverflow or gitter.
I'll answer this one anyway, but don't ask further questions here that are unrelated to i18n future plans please.

The answer is: you can't do it, dynamic translations are not supported by Angular i18n since it works at compile or build time, and not at runtime.

[martinmcwhorter]

Off Topic: I Apologize
@rayer4u The solution is to use static translations:

<div *ngIf="exercise?._desc">{{ exercise?._desc }}</div>
<div *ngIf="!exercise?._desc" i18n>empty</div>

[rayer4u]

@ocombe I Apologize for off Topic, no need to reply. I just have this need and can't find any quick way to discuss it. I think this may be concerned by angular i18n plan. As Angular i18n plan don't deal with interpolation, a extra mark/find/replace is better build time i18n solution for me right now.
@martinmcwhorter Thanks for your reply. We write a lot mixed if/else logic to output diffrent i18n text in template, this is the easiest one. Maybe put runtime text logic in template is a bad choice.

[elvisbegovic]

maybe this is bad idea but can't angular add automatically @@id in our i18n attributes when we extract first time ?

[ocombe]

No because we would have to maintain this when you edit the code.
Let's say that you have two blocks of i18n code that are the same, we add the ids, and then you change one of them, the ids don't match which means that the next extraction will skip on of those...

[elvisbegovic]

I confirm it's bad idea, can you explain how angular know to attribut same id to string when we extract again and again, how they keep same ID on string. so powerful but I don't undersand :s

@ocombe

[ocombe]

This is what we use: https://github.com/angular/angular/blob/master/packages/compiler/src/i18n/digest.ts#L13-L15

[ocombe]

We are now at v4 so I'm closing this issue, please follow up in #16477 for our plans for v4 and beyond.

[angular-automatic-lock-bot]

This issue has been automatically locked due to inactivity.
Please file a new issue if you are encountering a similar or related problem.

Read more about our automatic conversation locking policy.

_{This action has been performed automatically by a bot.}