137695 - Mac Japaense IME do not support "Reconversion" (control+1)

Status: VERIFIED FIXED

(Core :: Internationalization, defect, P3)

Description

[Frank Tang]

Currently, both the mozilla and PPEmbed does not support "Reconversion" in the
japaense IME. 
Reproduce procdure
1. open File:New:Blank page to edit
2. switch to Japanese input method
3. type "nihongo" , hit space- now the IME will underline the peredit text and
in conversion mode.
4.  hit return to commit the text
5. select the commited text
6. hit control + 1 (or select the last item in the pencel menu "Reconversion ^1")
expect result:
the ime will go back to conversion mode like step 3 and keep hitting space can
let the user reconvert, candidte widnow may popup
actual result:
nothing happen.

The reason that we don't reconvert is we have not implement the kGetSelectedText
AppleEvent . This Apple event is documented in develop issue 29 


http://developer.apple.com/dev/techsupport/develop/issue29/griffith.html


Table 1. Definition of the Get Text event

Event class kTextServiceClass
Event ID kGetText (= 'gtxt')
Requested action Returns the current selection as the return parameter
(Keyword / Descriptor type : Data)
Required parameters : 
keyDocumentRefcon / typeLongInteger : 
 Standard refCon parameter
keyAEServerInstance / typeComponentInstance : 
 Standard component instance parameter

Optional parameters: 
keyAEBufferSize (= 'buff') / typeLongInteger :
 Maximum number of bytes the input method can receive

Return parameters:
keyAETheData / typeText :
The text specified by the current selection. The maximum byte length is
specified by the keyAEBufferSize parameter. Any portion of the text that exceeds
the buffer length shouldn't be returned.

according to
http://developer.apple.com/techpubs/macos8/pdf/SupportingUnicodeInput.pdf (page 21)

The Get Selected Text Event
Your applicatoin's Apple event handler for the kGetSelectedTextApple Event must
return the current text selection as Unicode text. ... Supporting this event is
optional, but recommended.

Comment 1

[Frank Tang]

look at how window implement reconversion at 
/widget/src/windows/nsWindow.cpp, line 5828 -
send a NS_RECONVERSION_QUERY message :
5823 nsReconversionEvent event;
5824 nsPoint point;
5825 
5826 point.x = 0;
5827 point.y = 0;
5828 InitEvent(event, NS_RECONVERSION_QUERY, &point);
5829 event.theReply.mReconversionString = NULL;
5830 DispatchWindowEvent(&event);
5831 
5832 mIMEReconvertUnicode = event.theReply.mReconversionString;
5833 NS_RELEASE(event.widget);

Comment 2

[Frank Tang]

Attachment: working log. not ready for review

Comment 3

[Frank Tang]

Attachment: patch v1.0

Comment 4

[Frank Tang]

yokoyama, can you r= this one. Low priority for now. Not for nsbeta1

Comment 5

[Frank Tang]

kathy brade: can you review this for me. 

Comment 6

[Kathleen :Brade]

Comment on attachment 80503 [details] [diff] [review]
patch v1.0

r=brade

note for this block (@@ -2442,5 +2469,5 @@) in nsMacEventHandler.cpp there
appears to be a tab or spaces added after the "}" which should be removed

Comment 7

[Simon Fraser [no longer active]]

Comments:

+  returnSize = outString.Length()*2;

outString.Length() * sizeof(PRUnichar) would be more meaningful.

+  if((maxReturnSize >0) && (returnSize > maxReturnSize))

Fix spacing to be consistent with the rest of the file (space after 'if').

+    returnSize = maxReturnSize & ~1L;

Add a comment saying what this does, so I don't have to think about it.
// Round down
would be fine.

+  err = AEPutParamPtr(reply, keyAETheData, typeUnicodeText, outString.get(),
outString.Length()*sizeof(PRUnichar));

Don't you want to use 'returnSize' here, not outString.Length()*sizeof(PRUnichar)?

Comment 8

[hippytrail]

Perhaps I'm wrong but if sizeof(PRUnichar) == 2, doesn't this mean that it
holds one UCS-2 or UTF-16 codepoint?
Since surrogate charcters can require two codepoints, won't this calculation
fail in the case of the string containing characters which map to the surrogate
area?  Or is this known to be impossible?

Comment 9

[Frank Tang]

>Perhaps I'm wrong but if sizeof(PRUnichar) == 2, doesn't this mean that it
>holds one UCS-2 or UTF-16 codepoint?
>Since surrogate charcters can require two codepoints, won't this calculation
>fail in the case of the string containing characters which map to the surrogate
>area?  Or is this known to be impossible?
no, this is a correct cacluation about the size. 

Comment 10

[Frank Tang]

Attachment: v1.1 address sfraser's comment

Comment 11

[Frank Tang]

Comment on attachment 94106 [details] [diff] [review]
v1.1 address sfraser's comment

carry last r=

Comment 12

[Simon Fraser [no longer active]]

Comment on attachment 94106 [details] [diff] [review]
v1.1 address sfraser's comment

sr=sfraser

Comment 13

[Frank Tang]

fixed and check in .

Comment 14

[Rui Xu]

Verified fixed with trunk 2002-08-15-08.

Comment 15

[Marek Z. Jeziorek]

batch: adding topembed per Gecko2 document
http://rocknroll.mcom.com/users/marek/publish/Gecko/Gecko2Tasks.html

Comment 16

[Yuying Long]

The reconversion with "Control+1" is not resolved, at least not completed resolved:

11-11 trunk build / Mac 10.2.1:
When I highlighted a commited kannji character "Japan"(nihonn), then select the
"reconversion" under Japanese IME pencel menu, I'll get hiragana "nihonn" which
I think it's correct brhavior.
However, when I use "Control+1" to do the same thing, I don't get the "nihonn"
hiragana but a wrong kanji character with un-commit mode (red underline), and if
I press space to bring the candidate window, it will bring up a wrong candidate
window.

11-11 trunk build / Mac 10.1.5:
Similar result as in 10.2.1, but instead of give me a wrong un-commint kanji
character, it will show the original kanji (not hiragana) character, the
candidate window only can see that kanji character by pressing space key.

Re-open this bug for this "Control+1" doesn't work properly, although the
reconversion by select under IME pancel menu works fine.

Change QA contact to myself.

Comment 17

[Rui Xu]

There are some differences between MacOS 10.1 and MacOS 10.2 for "Reconversion" 
in Japanese IME pencel menu and short-cut keys, e.g. the correct short-cut for 
JA "Reconversion" on MacOS 10.2 should be "Control + 2", but not "Control + 1" 
on MacOS 10.1. "Control + 1" on MacOS 10.2 is assigned to a different function 
other than "Reconversion".

It might be better to have a system app (e.g. TextEdit) as a reference of the 
correct behavior.

Kepping as Fixed.

Comment 18

[Yuying Long]

Thanks.  Mark as verified based on the same behave as in TextEditor.