Screen Reader NVDA Japanese

Fork

• howto
• nvdajp
• nvdajpmiscdep

accessibleObjectFromEvent in MSAAHandler now uses the accessibleObjectFromEvent window api call, rather than just accessibleObjectFromWindow and using the given child ID. This now fixes the problem with giving focus to menu bar items or menu popup items reporting as role menuBar or role menu popup.

*Added getRoleText and getStateText to MSAAHandler. The use the real windows api functions from oleacc.dll.
*Changed getRoleName and getStateName in NVDAObjects.py to see if the role or state is in the dicionary and use that, but if not, use the name returned by getRoleText or getStateText.
*Deleted most of the entries out of the role and state dictionaries, meaning that most roles and states will be reported by their proper windows names (most were labeled coorectly anyways, so not much change will be noticed) but role_system_text is left in as edit, role_system_statictext is text, and role_system_pushBUTTON is button.
*Changed excel appModule back to asking the excel7 window for the native object model with comtypes. Yes this takes time to generate, but at least it always works. Eventually the generation stuff will be fixed up so it doesn't have to.

All needed windows api functions have been created in newly added winUser.py and winKernel.py using ctypes. All references to python win32 extentions have been removed and now winUser and winKernel modules are used instead. The only reasons for installing the win32 extentions are for pyHook which will be replaced hopefully soon, and some win32com calls which hopefully also will be replaced with comtypes calls.

Excel appModule again uses a non-dynamic com interface, this is temporary, but for now the dynamic stuff is just to flimzy. Hav contacted comtypes author and am asking for assistance with how to use dynamic IDispatch that does not need to generate type info.

Wordpad bug fixed ... with a time.sleep in getDocumentObjectModel in the nvdaObject_ITextDocument ... it seems that now NVDA is just a little too responsive and we get a focus to the richEditControl before ITextDocument com interface is completely up.

No pyAA anymore. All MSAA is implemented through ctypes and comtypes directly. All is going well, all methods work, though there is one bug where NVDA freezes when starting wordpad, yet it handles worpad fine if it had already been started.

Renamed keyEventHandler to keyboardHandler, MSAAEventHandler to MSAAHandler and mouseEventHandler to mouseHandler. One reason being is that MSAAHandler will soon contain some other MSAA functions that arn't to do with watching events.

First part of changing NVDA to be independent of pyAA and pyHook. This change now uses ctypes and windows api hook functions to watch for MSAA events. Things seem to be good, even a bit more responsive, though it seems to freeze up after a random interval, or when you switch applications.

Excel now uses win32com.client.dynamic.Dpspatch to access its object model. Seems when I tested last night things weren't very normal... after a restart, the excel object model works fine again. This means now that everything in the binary version of NVDA works just as well as the source version.

Trying to move back to using win32com as much as possible since comtypes seems to always insist on generating full type library objects for everything evne if it can just use IDispatch dynamically, and this is not a good idea when running from a stand alone executable <sits and waits up to 30 seconds for MS Word support to generate>. MS Word now uses win32com.client.dynamic.Dispatch to access its object model, and ITextDocument in NVDAObjects.py still uses comtypes but comtypes.client.dynamic.Dispatch, which at least for ITextDocument doesn't seem to take very long to generate. There is some issues with Excel and win32com ... excel.ActiveCell or excel.Selection for instance always return None, where as when using comtypes they work properly.

Excel appModule now has insert+f for format info (so far font name, font size, bold, italic and underline) when in a spreadsheet. Also fixed up the formatInfo script for ITextDocument in NVDAObjects.py which was accidentily trying to report style (which only exists in MS Word). Got rid of that and also created a formatInfo script in winWord appModule that reports style, and then calls the ITextDocument version of the script.

The Excel appModule now can anounce the current selection as you select cells. For the moment this only works with contiguous ranges, not groups of ranges, but this could be easily done.

Updated setup.py to copy *.txt from source's parent dir (all the documentation so far) and place it in a documentation dir.

Updated setup.py to contain version, author, maintainer etc, plus also classifiers... seems a bit bloted, but at least then that is more metadata that can be used at any time, like for registering with various software lists.

Added entry in staticMap in NVDAObjects.py for RichEdit20W,ROLE_SYSTEM_TEXT using NVDAObject_ITextDocument, which is used in outlook express when viewing messages in text. This makes outlook express usable at least for reading messages now, though it seems there is no way in outlook express to compose messages in text... or I could be wrong...

Started on an appModule for MS Excel. So far you can arrow around and hear the coordinates content of the cells. No selecting yet. It also tells you if there is a formula or not. No editing cells yet, this seems to be a bit of a problem because the control excel6 client does not take edit field window messages and is also not an ITextDocument. Some more playing around needs to be done.

Fixed typo in winword.py: in __del__ of the appModule class it called _MSOfficeHelper.__del__(), should have been _MSOffice.__del__()o

Fixed textProcessing so that it not only breaks up words that are delimited by a capital letter, but handles words where one of these imbedded words is entirly capitalized (e.g. NVDAObjects) it splits between the A and the O. It also now breaks up words that end in numbers, at least viavoice does funny things with the pronunciation if this isn't done.

Again made dos console windows a little more stable, hpefully this time really really got rid of the double speaking, this time now the consoleUUpdater thread loop only runs while the object has focus. Previously __del__ was setting a variable to false, but of course, this would not work because the object would never be deleted anyway because of the active thread. Also when we see there is no new text, it does a further 9 or so more loops making sure that no text arrives to really make sure that the screen has stablized.

*Added a keyboard help mode (insert+1). This is built in to the api functions hasScript and executeScript directly. When on if you press a key, it reports the script name, the palce it is from (either an NVDAObject class, or an appModule) and a description ot ehs cript, which is the doc string of the script itself.
*Added a getParagraph method to NVDAObject_ITextDocument
*Added a moveByParagraph script to the winword NVDAObject_wordDocument and bound it to control+ExtendedUp and control+extendedDown.

Added script_speech_toggleMute (insert+s) to default appModule, toggles audio.allowSpeech ... each speak function in audio now checks allowspeech and if false, returns straight away.

Hopefully fixed bug where consoleWindowClasses were sometimes reading new text twice.

Changed api.py to better handle the new appModule class, as in use hasattr and getattr to access methods dynamically, rather than __dict__. Also removed a whole bunch of really old debug statements out of NVDAObjects.py which were there actually from before even that.

Moved __init__.py of appModules to appModuleHandler.py and __init__.py of synthDrivers to synthDriverHandler.py. It seems that doing the __init__ thing actually fixed a major import bug in NVDA, but now doing it this way again, lots of fun has been had to fix it ... __path__ is needed now in both of these files.

Updated setup.py to include synthDrivers, and not to include waves (which no longer exist).

Deleting synth_viavoice.py, that last commit was all over the shop, but hopefully the end result is ok

all synth drivers have been moved in to the synthDrivers directory, and an __init__.py created so that now synthDrivers is a module. Each synth driver file contains a class called synthDriver. Hopefully now the viavoice driver has been added to svn. com_sapi4 and com_sapi5 are now just sapi4 and sapi5 respectivly. the pyTTS sapi5 driver has now been removed for good. The waves directory has also been removed for now.

Got rid of appModules/default.py was just added as appModules/_default.py.

appModules.py is now appModules/__init__.py and all appModules are a class called appModule in their own respective .py files. This is the first step to having all modular things written in this fassion. Synths, will also soon be classes in a directory, rather than separate modules in the base dir.

*Viavoice synth driver safely encodes unicode strings in to ascii before sending to viavoice.
*ITextDocument NVDAObject is now structured like the MS Word document NVDAObject as much as possible, giving it the ability to reportChanges of font name, size, font attributes and alignment.
*MS word NVDAObject has been greatly prooned back and now inherits from ITextDocument NVDAObject. No features have been lost but now code is much smaller and less duplicated.