Screen Reader NVDA Japanese

Fork

• howto
• nvdajp
• nvdajpmiscdep

Updated documentation.

dos console events now check the consoleProcessList to make sure that there are at least 2 processes on the current console, if it drops to less than 2, this means that NVDA is now the only process on the console and therefore the other app has closed. NVDA frees the console if this happens and this closes the console nicely with out hanging or crashing.

Dis console new text changed back to the diff code with a few tweeks. I will try not to touch this for a while.

No longer using any fancy diff code for new text reading in dos consoles. Now it all just uses event_console_update_region, and so far, it all seems to read fine. The diff code worked fine in certain situations, but had a lot of floors.

speakMessage, speakObjectProperties, speakSymbol, speakText, cancel, all from audio.py, now add synth functions to a queue, and a loop in audio.py processes the queue. KeyboardHandler now calls audio.cancel directly rather than putting it in its queue for core to grab. This all now means that speech is silenced always before anything else happens. Reasons for this being a problem was where dis console had new text events which sometimes were executed before the silenceSpeech command had been processed by core. speakRealtimeMessage has been added to audio.py and this is the only speech function that does not use the queue. core's messages for starting and exiting NVDA now use this function.

*Added getForegroundObject, getForegroundLocator and setForegroundObjectByLocator functions to api.py.
*When core receives foreground events, like focusEvents it now sets the current foreground object.
*ExecuteEvent has been reordered a little, and now it passes foreground events to the foreground object. All this is so that both focus and foreground events can now be checked to see if they are different from the last one, to stop duplicate events.
*Core compares foreground and focus events and only lets through different ones.
*NVDAObject_consoleWindowClass has been renamed to NVDAObject_consoleWindowClassClient and a new NVDAObject_consoleWindowClass has been added that handles consoleWindowClass role_system_window objects, and all it does it does a pass statement for event_nameChange events, stopping the reading of the window title when it changes, when running commands etc.
*A time.sleep(0.01) has been added to event_gainFocus in NVDAObject_consoleWindowClassClient just before it registers the console events to receive new text and cursor movement info. Hopefully this fixes up the problem with the events just not firing sometimes.

Internet explorer virtualBuffer now updates the text in edit fields after focus is moved off them. Documents can't seem to receive onchange events from form fields, form fields must have their own event listeners, which isn't going to be that pretty.

Added getAsyncKeyState to winUser.py. Was going to use it with keyboardHandler to try and fix the alt+tab bug, but that didn't work. Still, eventually I think this function will come in handy.

Internet explorer virtualBuffer now shows the currently selected item in combo boxes.

Updating the bugs file, second try.

appModuleHandler now imports win32com.client to make sure that it is available to appModules when it is an executable. Also updated the bugs file.

setFocusObjectByLocator now starts the virtualBuffer for the focus window after it sets the focusLocator and focusObject global variables. Previously virtaulBuffer was started before and therefore the virtualBuffer couldn't yet tell if it had focus.

registerVirtualBufferClass and unregisterVirtualBufferClass have been added to the virtualBuffer module so that appModules can register their own virtualBuffer classes if needed.

Event support for internet explorer virtualBuffer has been implemented. So far the only event is onReadyStateChange. Now when a page is fully loaded (after clicking a link, refreshing ... etc) NVDA re-loads the text etc in the virtualBuffer. If the virtualBuffer has focus, it also automatilcally reads the contents of the buffer.

Setup.py now looks for .py* files in comInterfaces to include, so that the .pyc files if generated are included which should give quite a big speed up when running on a usb key.

This is a very big update.
*a modified version of comtypes.client module has been added. This version allows comtypes com interfaces to be generated in the comInterfaces dir. The module also makes sure it uses __import__ rather than import to import the comInterfaces so that they don't get frozen with py2exe.
*Added the most useful generated com Interfaces to the comInterfaces dir so they are already generated for both the source and exe versions of NVDA.
*Changed setup.py so it also copies the comInterfaces dir in to the distribution.
*Changed MSAAHandler, NVDAObjects, virtualBuffer, all to properly use the modified comtypesClient module.
*Changed the virtualBuffer implementation so that it stores created virtualBuffers for later use. VirtualBuffers are deleted when that actual window has a hide object event (MSAA documentation states not to use destroyObject or createObject). This all means that for instance in internet explorer, you can move to and from the menu bar, or alt tab away and back again, and the previously created virtualBuffer will be used. It is up to that specific virtualBuffer to listen for events and update its contents whether or not it has focus. BTW, internet explorer virtualBuffer doesn't yet have document events written.
*NVDAObject_virtualBuffer now has a focusInteractionMode. Use insert+space to toggle on and off. With it on, all virtualBuffer reading key presses will be passed straight through to the object with focus.
*virtualBuffer_internetExplorerServer now shows the state of checkboxes and radioButton input fields.
*A bug with the dos console support (NVDAObject_consoleWindowClass) has been fixed. This bug was that new text wasn't read until the consoleBuffer had changed about 5 times or so. Reason was that the lists of newLines and oldLines were checked for differences with a simple != before the actual line by line / character by character diff was created. Seems that first comparison just doesn't work like expected.

Internet explorer virtual buffer now removes the last new line character if the next start tag being inserted starts with a new line character.

When NVDA starts, if there is no foreground window, it grabs the desktop window instead. This stops it from exiting when starting as a swervice at windows log in. However, seems still that no events can be listened for at this time anyway.

Internet Explorer virtual Buffer now reads that stuats bar as it changes while loading a document. Also definition lists and ordered lists are recognised (though numbering is not in place yet). TD and TH cells in talbes now cause a line break before them.

Limited objectEventCallback in MSAAEventHandler to only work with caret, client, menu, sysMenu, titleBar and window object IDs. Receiving events from vertical and horizontal scrollbars etc at the moment seems to cause havic and can tend to freeze NVDA when starting the run dialog box etc.

Internet Explorer virtual Buffer now refreshes when a page is loaded or reloaded. Block quotes are recognised. Divds now cause line breaks. BR tags now cause line breaks.

*Excel now uses win32com again... seems now that Internet Explorer will probably always depend on it so may as well use it in other places.
*Re enabled virtualBuffer and created some basic, but promicing support for internet explorer. The document is rendered in the virtual Buffer, and then there is an NVDAObject_virtualBuffer which interacts with it. So far, you can use the arrows to navigate the document, talbes, lists, headings, links, some form fileds, all read. You can press enter on links etc to click on them, the position in the buffer also tracks the focus so you are always at the right bit. No support for dynamically updating the buffer when the page changes yet though. -- a work in progress, please try out!!!

Re-thought documentFormatting changes code. Now NVDAObject_edit has a presentationTable whcih is a list of lists, each sublist containing a messageFunction that takes position as its parameter, a list which pints to a config option to check, the last value that was at the caret position, and the last value that was at the review cursor position. Each item in the presentationTable can be used to manage things like fontName, size, attributes, alignment, table, page and so forth. reportPresentation and reportReviewPresentation now use the presentationTable. ITextDocument and MS Word documents have formatting rules now. Also reviwing by character, line and word in MS Word documents has been fixed up.

Ok, now we have an event_looseFocus and event_gainFocus for appModules and NVDAObjects. At least for console windows, all console preparation is done in event_gainFocus and cleaning up is down in event_looseFocus. Also fixed up some bugs with word navigation, it seems now that moving by character, line and word with the numpad in edit, ITextDocument and console windows works exactly as expected, no errors. I am yet to check MS Word though. Finally also removed a "return True" from executeEvent in api.py which was put there for debugging purposes and was forgotten about. And, aadded a return statement in executeScript to make sure that when in keyboard training mode the actuall script wouldn't get executed after its name was spoken.

Added review scripts to NVDAObject_edit (by character, word and line, using the numpad). For now have completely removed virtualBuffer support, have to re-think it for internet explorer etc. NVDAObject_edit, and by inheritence ITextDocument and consoleWindowClass, now have their own review scripts. getLineStart, getLineLength and getLine in NVDAObject_edit and friends all take pos as their argument not lineNum. In some circomstances such as in MS Word, line numbers are not completely individual (line numbers restart at each page break).

Re-wrote the consoleWindowClass NVDAObject to now use console events rather than a thread. New text is now checked on update region and update scroll events. It is also checked on update simple events (typing or deleting a character) but it is not spoken anymore.

Added insert+F12 to key commands.txt

Oops, move to parent and move to next and previous broke because of typing errors in MSAAHandler.py ... all fixed now. accParent was always returning the object it started with.

Fixed up getValue in NVDAObject (it actually wasn't returning a value, hence no value ever being reported). Also created an NVDAObject_outlineItem which simply makes its getValue report "level %s" where %s is the value.

getNVDAObjectClass now selects a class also by (None,objectRole) if (windowClass,objectRole) and (windowClass,None) have both failed. Also what used to be ("button",ROLE_SYSTEM_CHECKBUTTON) in staticMap is now (None,ROLE_SYSTEM_CHECKBUTTON). This all makes MSAA roles less dependent on window classes.