[chromium-reviews] Add getLanguage function to tabs extension

Add getLanguage function to tab extension.

BUG=none
TEST=enable extensions using the toolstip.html code (added with this CL) and load pages in different languages. The corresponding language should appear in the bottom left after the page is loadedm or when the button is clicked, or when you navigate back to that tab after visiting some other tab.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=20378

[sidchat (Google), 2009-06-29 22:14:05]

The getLanguage function has to be called explicitly to get the language of the
currently active tab. If the page is not loading, it gets the language
immediately. Otherwise, it waits till the page is loaded, and then sends message
async to the extension module with the language of the page. A sample
toolstip.html file, provided by Rafael (who also provided the boiler plate code
to plumb tabs.getLanguage()) has been included.

[sidchat (Google), 2009-06-30 01:37:33]

I made changes to the architecture. Language detection now takes place in the
browser, and the renderer just provides the text content of the tab selected.
Sorry if you had started reviewing the previous patch. Please review the latest
one. Thanx.

-Sid

On 2009/06/29 22:14:05, sidchat (Google) wrote:
> The getLanguage function has to be called explicitly to get the language of
the
> currently active tab. If the page is not loading, it gets the language
> immediately. Otherwise, it waits till the page is loaded, and then sends
message
> async to the extension module with the language of the page. A sample
> toolstip.html file, provided by Rafael (who also provided the boiler plate
code
> to plumb tabs.getLanguage()) has been included.

[Aaron Boodman, 2009-06-30 21:53:16]

Very cool.

http://codereview.chromium.org/150062/diff/1086/66
File chrome/browser/extensions/extension_tabs_module.cc (right):

http://codereview.chromium.org/150062/diff/1086/66#newcode626
Line 626: args_->GetAsInteger(&tab_id);
If the argument is not null and is not an integer, tab_id will have garbage
here. You should initialize tab_id to zero, but also you should ues
EXTENSION_FUNCTION_VALIDATE to verify that the received message has the expected
structure. Like this:

if (args->IsType(Value::TYPE_INTEGER))
  get integer
else
  EXTENSION_FUNCTION_VALIDATE(args->IsType(Value::NULL)

http://codereview.chromium.org/150062/diff/1086/66#newcode641
Line 641: // Figure out what language |contents| contains. This sends an async
This comment looks like it can be rewrapped.

http://codereview.chromium.org/150062/diff/1086/66#newcode644
Line 644: // after the tab loads (it may not loaded at all!) to the browser,
which in
We should try to ensure that we either send back success or failure eventually.
It is bad form to never send any response at all, because then developers have
to write timeout code.

http://codereview.chromium.org/150062/diff/1086/66#newcode664
Line 664: }
You should DCHECK that you don't receive any other notifications.

http://codereview.chromium.org/150062/diff/1086/70
File chrome/browser/renderer_host/render_view_host.cc (left):

http://codereview.chromium.org/150062/diff/1086/70#oldcode44
Line 44: 
Nit: put this line back.

http://codereview.chromium.org/150062/diff/1086/70
File chrome/browser/renderer_host/render_view_host.cc (right):

http://codereview.chromium.org/150062/diff/1086/70#newcode48
Line 48: // TODO(port): The compact language detection library works only for
Win.
What is the long-term plan for other platforms? We should avoid extension APIs
that only work on one platform.

Nit: Avoid abreviations: Win->Windows

http://codereview.chromium.org/150062/diff/1086/70#newcode401
Line 401: // in the page, use that to determine the language it is in.
This comment is not necessary.

http://codereview.chromium.org/150062/diff/1086/70#newcode1105
Line 1105: DWORD error_code;
Nit: I think we're supposed to avoid platform specific types, so this would be
uint32, not DWORD.

http://codereview.chromium.org/150062/diff/1086/70#newcode1107
Line 1107: const char* language_char = LanguageName(DetectLanguageOfUnicodeText(
Should we check the error code?

http://codereview.chromium.org/150062/diff/1086/70#newcode1107
Line 1107: const char* language_char = LanguageName(DetectLanguageOfUnicodeText(
Who owns this memory (language_char)?

http://codereview.chromium.org/150062/diff/1086/71
File chrome/browser/renderer_host/render_view_host.h (right):

http://codereview.chromium.org/150062/diff/1086/71#newcode204
Line 204: // a message to the render view to know the content of the page as
text.
Grammar nit: replace "know" with "get".

http://codereview.chromium.org/150062/diff/1086/71#newcode205
Line 205: void GetPageLanguage();
It's worth mentioning here how a caller gets the result (via the notification
service).

http://codereview.chromium.org/150062/diff/1086/71#newcode495
Line 495: // information. On getting the text, the compact language detection
(CLD)
You don't need comments for these message handler methods.

http://codereview.chromium.org/150062/diff/1086/73
File chrome/common/notification_type.h (right):

http://codereview.chromium.org/150062/diff/1086/73#newcode215
Line 215: // has been determined for the focussed tab. The source is render view
host.
You don't need so much detail on what will happen after this notification is
received. The notification could conceivably be used for something else, later,
not having to do with extensions.

I'd just say:

This notification is sent after a call to RenderViewHost::DeterminePageLanguage.
The details are xxx and the source is yyy.

http://codereview.chromium.org/150062/diff/1086/77
File chrome/test/data/extensions/samples/cld/toolstrip.html (right):

http://codereview.chromium.org/150062/diff/1086/77#newcode1
Line 1: <html>
Put the standard Chromium license header here (in an HTML comment).

http://codereview.chromium.org/150062/diff/1086/77#newcode27
Line 27: <div id="languageDiv" class="toolstrip-button"
onclick="refreshLanguage();"></div>
line length.

[sidchat (Google), 2009-07-01 18:27:40]

http://codereview.chromium.org/150062/diff/1086/66
File chrome/browser/extensions/extension_tabs_module.cc (right):

http://codereview.chromium.org/150062/diff/1086/66#newcode626
Line 626: args_->GetAsInteger(&tab_id);
I had to move stuff in Run() to RunImpl() because EXTENSION_FUNCTION_VALIDATE
expects to run in a method which returns a bool - hence seems to be always used
in a RunImpl() method.

On 2009/06/30 21:53:16, Aaron Boodman wrote:
> If the argument is not null and is not an integer, tab_id will have garbage
> here. You should initialize tab_id to zero, but also you should ues
> EXTENSION_FUNCTION_VALIDATE to verify that the received message has the
expected
> structure. Like this:
> 
> if (args->IsType(Value::TYPE_INTEGER))
>   get integer
> else
>   EXTENSION_FUNCTION_VALIDATE(args->IsType(Value::NULL)

http://codereview.chromium.org/150062/diff/1086/66#newcode641
Line 641: // Figure out what language |contents| contains. This sends an async
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> This comment looks like it can be rewrapped.

Done.

http://codereview.chromium.org/150062/diff/1086/66#newcode644
Line 644: // after the tab loads (it may not loaded at all!) to the browser,
which in
It seem response is always sent through the CapturePageInfo method in
renderview. If the user navigates away from the page before that, a fresh
request for language detection is sent, so this should not be an issue.

On 2009/06/30 21:53:16, Aaron Boodman wrote:
> We should try to ensure that we either send back success or failure
eventually.
> It is bad form to never send any response at all, because then developers have
> to write timeout code.

http://codereview.chromium.org/150062/diff/1086/66#newcode664
Line 664: }
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> You should DCHECK that you don't receive any other notifications.

Done.

http://codereview.chromium.org/150062/diff/1086/70
File chrome/browser/renderer_host/render_view_host.cc (right):

http://codereview.chromium.org/150062/diff/1086/70#newcode48
Line 48: // TODO(port): The compact language detection library works only for
Win.
Porting efforts are ongoing. I will talk to the toolbar folks to figure out
quick ports for Linux and Mac next quarter.

On 2009/06/30 21:53:16, Aaron Boodman wrote:
> What is the long-term plan for other platforms? We should avoid extension APIs
> that only work on one platform.
> 
> Nit: Avoid abreviations: Win->Windows

http://codereview.chromium.org/150062/diff/1086/70#newcode401
Line 401: // in the page, use that to determine the language it is in.
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> This comment is not necessary.

Done.

http://codereview.chromium.org/150062/diff/1086/70#newcode1105
Line 1105: DWORD error_code;
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> Nit: I think we're supposed to avoid platform specific types, so this would be
> uint32, not DWORD.

Done.

http://codereview.chromium.org/150062/diff/1086/70#newcode1107
Line 1107: const char* language_char = LanguageName(DetectLanguageOfUnicodeText(
Checking for error code is not required here - it seems we can pass a NULL. The
API is built such that it returns "UNKNOWN" if language could not be determined.

On 2009/06/30 21:53:16, Aaron Boodman wrote:
> Should we check the error code?

http://codereview.chromium.org/150062/diff/1086/70#newcode1107
Line 1107: const char* language_char = LanguageName(DetectLanguageOfUnicodeText(
RenderViewHost owns it. I am guessing it will be alive when
GetTabLanguageFunction gets the notification.

On 2009/06/30 21:53:16, Aaron Boodman wrote:
> Who owns this memory (language_char)?

http://codereview.chromium.org/150062/diff/1086/71
File chrome/browser/renderer_host/render_view_host.h (right):

http://codereview.chromium.org/150062/diff/1086/71#newcode204
Line 204: // a message to the render view to know the content of the page as
text.
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> Grammar nit: replace "know" with "get".

Done.

http://codereview.chromium.org/150062/diff/1086/71#newcode205
Line 205: void GetPageLanguage();
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> It's worth mentioning here how a caller gets the result (via the notification
> service).

Done.

http://codereview.chromium.org/150062/diff/1086/73
File chrome/common/notification_type.h (right):

http://codereview.chromium.org/150062/diff/1086/73#newcode215
Line 215: // has been determined for the focussed tab. The source is render view
host.
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> You don't need so much detail on what will happen after this notification is
> received. The notification could conceivably be used for something else,
later,
> not having to do with extensions.
> 
> I'd just say:
> 
> This notification is sent after a call to
RenderViewHost::DeterminePageLanguage.
> The details are xxx and the source is yyy.

Done.

http://codereview.chromium.org/150062/diff/1086/77
File chrome/test/data/extensions/samples/cld/toolstrip.html (right):

http://codereview.chromium.org/150062/diff/1086/77#newcode1
Line 1: <html>
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> Put the standard Chromium license header here (in an HTML comment).

Done.

http://codereview.chromium.org/150062/diff/1086/77#newcode27
Line 27: <div id="languageDiv" class="toolstrip-button"
onclick="refreshLanguage();"></div>
On 2009/06/30 21:53:16, Aaron Boodman wrote:
> line length.

Done.

[Aaron Boodman, 2009-07-04 09:37:56]

Getting closer.

Ping me on monday if my refactor suggestion doesn't make sense.

http://codereview.chromium.org/150062/diff/1119/1124
File chrome/browser/extensions/extension_tabs_module.cc (right):

http://codereview.chromium.org/150062/diff/1119/1124#newcode619
Line 619: get_tab_language_initiated_ = false;
Do we need "get_tab_language_" in the name of this member? Seems like it can
just be initiated_, since we know that it is specific to this class already.

http://codereview.chromium.org/150062/diff/1119/1124#newcode622
Line 622: NotificationService::AllSources());
indent.

http://codereview.chromium.org/150062/diff/1119/1124#newcode641
Line 641: EXTENSION_FUNCTION_VALIDATE(args_->GetAsInteger(&tab_id));
This EXTENSION_FUNCTION_VALIDATE isn't doing anything because it sets
bad_message_ to true, which is only processed by SendResponse(), which we aren't
calling in the failure case.

I think it will be common to want to do preflight validation of
AsyncExtnesionFunctions, so how about refactoring things so that this works
like:

* Add a bool RunImpl() method to AsyncExtensionFunction, like
SyncExtensionFunction has.
* Add an implementation of AsyncExtensionFunction::Run(), similar to
SyncExtensionFunction's, that does: if (!RunImpl()) SendResponse(false).
* Move the contents of GetTabFunction::Run() to ::RunImpl().

http://codereview.chromium.org/150062/diff/1119/1124#newcode642
Line 642: if (!GetTabById(tab_id, profile(), &browser, NULL, &contents, NULL,
&error_))
line length

http://codereview.chromium.org/150062/diff/1119/1135
File chrome/test/data/extensions/samples/cld/toolstrip.html (right):

http://codereview.chromium.org/150062/diff/1119/1135#newcode1
Line 1: <!-- Copyright (c) 2009 The Chromium Authors. All rights reserved. -->
Nit: use this format:

<!--
copyright copyright copyright copyright copyright copyright
copyright copyright copyright copyright copyright copyright
-->

[sidchat (Google), 2009-07-06 19:00:22]

Great suggestions - made changes accordingly - please have a look again.

-Sid

On 2009/07/04 09:37:56, Aaron Boodman wrote:
> Getting closer.
> 
> Ping me on monday if my refactor suggestion doesn't make sense.
> 
> http://codereview.chromium.org/150062/diff/1119/1124
> File chrome/browser/extensions/extension_tabs_module.cc (right):
> 
> http://codereview.chromium.org/150062/diff/1119/1124#newcode619
> Line 619: get_tab_language_initiated_ = false;
> Do we need "get_tab_language_" in the name of this member? Seems like it can
> just be initiated_, since we know that it is specific to this class already.
> 
> http://codereview.chromium.org/150062/diff/1119/1124#newcode622
> Line 622: NotificationService::AllSources());
> indent.
> 
> http://codereview.chromium.org/150062/diff/1119/1124#newcode641
> Line 641: EXTENSION_FUNCTION_VALIDATE(args_->GetAsInteger(&tab_id));
> This EXTENSION_FUNCTION_VALIDATE isn't doing anything because it sets
> bad_message_ to true, which is only processed by SendResponse(), which we
aren't
> calling in the failure case.
> 
> I think it will be common to want to do preflight validation of
> AsyncExtnesionFunctions, so how about refactoring things so that this works
> like:
> 
> * Add a bool RunImpl() method to AsyncExtensionFunction, like
> SyncExtensionFunction has.
> * Add an implementation of AsyncExtensionFunction::Run(), similar to
> SyncExtensionFunction's, that does: if (!RunImpl()) SendResponse(false).
> * Move the contents of GetTabFunction::Run() to ::RunImpl().
> 
> http://codereview.chromium.org/150062/diff/1119/1124#newcode642
> Line 642: if (!GetTabById(tab_id, profile(), &browser, NULL, &contents, NULL,
> &error_))
> line length
> 
> http://codereview.chromium.org/150062/diff/1119/1135
> File chrome/test/data/extensions/samples/cld/toolstrip.html (right):
> 
> http://codereview.chromium.org/150062/diff/1119/1135#newcode1
> Line 1: <!-- Copyright (c) 2009 The Chromium Authors. All rights reserved. -->
> Nit: use this format:
> 
> <!--
> copyright copyright copyright copyright copyright copyright
> copyright copyright copyright copyright copyright copyright
> -->

[sidchat (Google), 2009-07-06 19:02:59]

Forgot to add - I don't know how to initialize the boolean |initiate_| in class
GetTabLanguageFunction before RunImpl() is first called. It is false by default
I guess, which makes it work.

On 2009/07/06 19:00:22, sidchat (Google) wrote:
> Great suggestions - made changes accordingly - please have a look again.
> 
> -Sid
> 
> On 2009/07/04 09:37:56, Aaron Boodman wrote:
> > Getting closer.
> > 
> > Ping me on monday if my refactor suggestion doesn't make sense.
> > 
> > http://codereview.chromium.org/150062/diff/1119/1124
> > File chrome/browser/extensions/extension_tabs_module.cc (right):
> > 
> > http://codereview.chromium.org/150062/diff/1119/1124#newcode619
> > Line 619: get_tab_language_initiated_ = false;
> > Do we need "get_tab_language_" in the name of this member? Seems like it can
> > just be initiated_, since we know that it is specific to this class already.
> > 
> > http://codereview.chromium.org/150062/diff/1119/1124#newcode622
> > Line 622: NotificationService::AllSources());
> > indent.
> > 
> > http://codereview.chromium.org/150062/diff/1119/1124#newcode641
> > Line 641: EXTENSION_FUNCTION_VALIDATE(args_->GetAsInteger(&tab_id));
> > This EXTENSION_FUNCTION_VALIDATE isn't doing anything because it sets
> > bad_message_ to true, which is only processed by SendResponse(), which we
> aren't
> > calling in the failure case.
> > 
> > I think it will be common to want to do preflight validation of
> > AsyncExtnesionFunctions, so how about refactoring things so that this works
> > like:
> > 
> > * Add a bool RunImpl() method to AsyncExtensionFunction, like
> > SyncExtensionFunction has.
> > * Add an implementation of AsyncExtensionFunction::Run(), similar to
> > SyncExtensionFunction's, that does: if (!RunImpl()) SendResponse(false).
> > * Move the contents of GetTabFunction::Run() to ::RunImpl().
> > 
> > http://codereview.chromium.org/150062/diff/1119/1124#newcode642
> > Line 642: if (!GetTabById(tab_id, profile(), &browser, NULL, &contents,
NULL,
> > &error_))
> > line length
> > 
> > http://codereview.chromium.org/150062/diff/1119/1135
> > File chrome/test/data/extensions/samples/cld/toolstrip.html (right):
> > 
> > http://codereview.chromium.org/150062/diff/1119/1135#newcode1
> > Line 1: <!-- Copyright (c) 2009 The Chromium Authors. All rights reserved.
-->
> > Nit: use this format:
> > 
> > <!--
> > copyright copyright copyright copyright copyright copyright
> > copyright copyright copyright copyright copyright copyright
> > -->

[Aaron Boodman, 2009-07-06 22:08:55]

Ok, getting really close now.

Before I give this the LG, can you add some unit tests and a browser test.

Please see extension_api_client_unittest.cc and extension_browsertests_misc.cc
for examples.

In the browser tests, look at the toolstrip test for an example of calling some
javascript in a toolstrip. You should be able to extend this to call the
language detection code an ensure it returns a good result.

http://codereview.chromium.org/150062/diff/3003/3008
File chrome/browser/extensions/extension_tabs_module.cc (right):

http://codereview.chromium.org/150062/diff/3003/3008#newcode625
Line 625: int tab_id;
Please initialize values. This is scary when you're reading the code for the
first time.

http://codereview.chromium.org/150062/diff/3003/3008#newcode666
Line 666: SendResponse(RunImpl());
It is weird to enter back into RunImpl() here to set the value. Instead just do
set result_ directly here and kill the initiated_ value.

http://codereview.chromium.org/150062/diff/3003/3012
File chrome/browser/renderer_host/render_view_host.cc (right):

http://codereview.chromium.org/150062/diff/3003/3012#newcode1024
Line 1024: int num_languages;
Please initialize values :)

[sidchat (Google), 2009-07-08 18:50:12]

Made changes as per your suggestions. Also, added unit tests and a browser test.
Please have a look again.

-Sid

On 2009/07/06 22:08:55, Aaron Boodman wrote:
> Ok, getting really close now.
> 
> Before I give this the LG, can you add some unit tests and a browser test.
> 
> Please see extension_api_client_unittest.cc and extension_browsertests_misc.cc
> for examples.
> 
> In the browser tests, look at the toolstrip test for an example of calling
some
> javascript in a toolstrip. You should be able to extend this to call the
> language detection code an ensure it returns a good result.
> 
> http://codereview.chromium.org/150062/diff/3003/3008
> File chrome/browser/extensions/extension_tabs_module.cc (right):
> 
> http://codereview.chromium.org/150062/diff/3003/3008#newcode625
> Line 625: int tab_id;
> Please initialize values. This is scary when you're reading the code for the
> first time.
> 
> http://codereview.chromium.org/150062/diff/3003/3008#newcode666
> Line 666: SendResponse(RunImpl());
> It is weird to enter back into RunImpl() here to set the value. Instead just
do
> set result_ directly here and kill the initiated_ value.
> 
> http://codereview.chromium.org/150062/diff/3003/3012
> File chrome/browser/renderer_host/render_view_host.cc (right):
> 
> http://codereview.chromium.org/150062/diff/3003/3012#newcode1024
> Line 1024: int num_languages;
> Please initialize values :)

[Aaron Boodman, 2009-07-08 18:56:37]

Great!

One question below, then lgtm

http://codereview.chromium.org/150062/diff/4047/3060
File
chrome/test/data/extensions/good/Extensions/behllobkkfkfnphdnhnkndlbkcpglgmj/1.0.0.0/toolstrip1.html
(right):

http://codereview.chromium.org/150062/diff/4047/3060#newcode17
Line 17: window.domAutomationController.send(language == 'FRENCH');
Is it possible to use the standard ISO language codes (like en, fr). It seems it
would be more interoperable with other software that way.

[sidchat (Google), 2009-07-08 19:04:28]

> http://codereview.chromium.org/150062/diff/4047/3060#newcode17
> Line 17: window.domAutomationController.send(language == 'FRENCH');
> Is it possible to use the standard ISO language codes (like en, fr). It seems
it
> would be more interoperable with other software that way.
Yes, I have to convert these languages to ISO languages to hook up with
translation, spellcheck auto language detection etc. I will do that in a
separate patch, when 'FRENCH' will be returned as 'fr', etc.

[jungshik at Google, 2009-07-10 00:38:42]

http://codereview.chromium.org/150062/diff/3087/4160
File chrome/renderer/resources/extension_process_bindings.js (right):

http://codereview.chromium.org/150062/diff/3087/4160#newcode27
Line 27: native function GetTabLanguage();
I'm afraid Get*Language gives a wrong impression (even though we'll document
about that). How about DetectTabLanguage() to make clear that it's not for
reading off the value of 'lang' or 'xml:lang' or 'Content-Language'?

Also, CLD returns the top 3 languages with the proportion of each in the input.
It appears that this function just returns the top language. Do we want to have
another function for a more advanced usage?

[sidchat (Google), 2009-07-10 18:12:54]

> I'm afraid Get*Language gives a wrong impression (even though we'll document
> about that). How about DetectTabLanguage() to make clear that it's not for
> reading off the value of 'lang' or 'xml:lang' or 'Content-Language'?
Yes, that can done - I will make a separate CL for that.

> 
> Also, CLD returns the top 3 languages with the proportion of each in the
input.
> It appears that this function just returns the top language. Do we want to
have
> another function for a more advanced usage?
I don't see any use of the advanced feature where all the three languages are
returned. When translating a web page, a single "from" language is required. Of
course, the page may be multilingual - even then, since translation occurs for
each block of text, it still would require the dominant single language to
translate from.

Having said that, I may require the advanced API for getting all the three
languages for my advanced-spell-check stuff - but they will be internal, and not
be part of extensions.

[xji, 2009-07-13 19:49:44]

On 2009/07/10 18:12:54, sidchat (Google) wrote:
> > I'm afraid Get*Language gives a wrong impression (even though we'll document
> > about that). How about DetectTabLanguage() to make clear that it's not for
> > reading off the value of 'lang' or 'xml:lang' or 'Content-Language'?
> Yes, that can done - I will make a separate CL for that.
> 
> > 
> > Also, CLD returns the top 3 languages with the proportion of each in the
> input.
> > It appears that this function just returns the top language. Do we want to
> have
> > another function for a more advanced usage?
> I don't see any use of the advanced feature where all the three languages are
> returned. When translating a web page, a single "from" language is required.
Of
> course, the page may be multilingual - even then, since translation occurs for
> each block of text, it still would require the dominant single language to
> translate from.
> 
> Having said that, I may require the advanced API for getting all the three
> languages for my advanced-spell-check stuff - but they will be internal, and
not
> be part of extensions.

Hi Sid:

I plan to work on dictionary extension, which will  need CLD extension, and it
might need the top-3 results (and with reliability) because the language of the
word might be difference from the language of the page, especially when the
detection reliability is low (for mix-language pages).

This is just a thought. Not final yet. But such information would be useful.

Thanks,
Xiaomei

[sidchat (Google), 2009-07-14 18:08:55]

I see - OK - from popular requests, let me work on an advanced function that
returns all the three languages with percentages etc.

-Sid

On 2009/07/13 19:49:44, xji wrote:
> On 2009/07/10 18:12:54, sidchat (Google) wrote:
> > > I'm afraid Get*Language gives a wrong impression (even though we'll
document
> > > about that). How about DetectTabLanguage() to make clear that it's not for
> > > reading off the value of 'lang' or 'xml:lang' or 'Content-Language'?
> > Yes, that can done - I will make a separate CL for that.
> > 
> > > 
> > > Also, CLD returns the top 3 languages with the proportion of each in the
> > input.
> > > It appears that this function just returns the top language. Do we want to
> > have
> > > another function for a more advanced usage?
> > I don't see any use of the advanced feature where all the three languages
are
> > returned. When translating a web page, a single "from" language is required.
> Of
> > course, the page may be multilingual - even then, since translation occurs
for
> > each block of text, it still would require the dominant single language to
> > translate from.
> > 
> > Having said that, I may require the advanced API for getting all the three
> > languages for my advanced-spell-check stuff - but they will be internal, and
> not
> > be part of extensions.
> 
> Hi Sid:
> 
> I plan to work on dictionary extension, which will  need CLD extension, and it
> might need the top-3 results (and with reliability) because the language of
the
> word might be difference from the language of the page, especially when the
> detection reliability is low (for mix-language pages).
> 
> This is just a thought. Not final yet. But such information would be useful.
> 
> Thanks,
> Xiaomei