Screen Reader NVDA Japanese

Fork

• howto
• nvdajp
• nvdajpmiscdep

Changed executeEvent in api.py to take parameters: name, window, objectID and childID, rather than name, and locator (tuple of window, objectID, childID). Most of this was changed in last commit, but one bit was missed.

*Implemented all voice, pitch and volume get/set functions for com_sapi5 synth driver.
*Added a relativeUppercasePitch setting to the individual synth section of the config., Default is 20.
*audio.initialize now sets the voice, rate, pitch and volume of a synth.
*Changed audio.py and all synth drivers to now have synth.SpeakText only taking text, and wait=value as parameters (got rid of the unused markup parameter).
*Added a viavoice synth driver which uses IBM viavoice directly (ibmenu50.dll) assuming you have a version with this.
*Moved all audio.speakMessage statements to do with exceptions below their respective debug.writeException so that if there is a bug in a synth it does not cause the exception to not be written to the log.

Added a documentFormat section to the config, and added settings: reportFontChanges, reportFontSizeChanges, reportFontAttributeChanges, reportAlignmentChanges, reportStyleChanges, reportTables, reportPageChanges. Added checks in winword.py to honour these settings.

Microsoft Word: Paragraph alignment, bold, italic and underline is reported also with insert+f, and is also reported if changed when moving around the document.

Microsoft Word support: automatically reads format changes when moving around a document. Includes style, font name, font size. Also announces entering and exiting tables, plus row and column numbers. NVDAObject_edit: added script_moveByLine for control+extendedHome and control+extendedEnd so it reads the line when it moves there, and script_changeSelection for control+shift+extendedHome and control+shift+extendedEnd.

Basic support for Microsoft Word. Based in NVDAObject_edit, but uses word's object model to get selection and ranges etc, plus insert+f for font name/size.

Added register and unregister functions for appModules to add and remove custom NVDAObjects. Changed executeEvent in api.py to move the focus to the client of a window if the caret is in another window other than the focus, this fixes the rror where focus would move literally to the caret object. getNVDAObjectClass, getNVDAObjectByLocator, getNVDAObjectByPoint and getNVDAObjectByAccessibleObject to NVDAObjects.py, now all other modules use these functions from NVDAObjects directly, rather than using api.py functions. Cleaned up api.py ... deleting many old window functions not used now we have NVDAObjects.

Added bugs.txt and todo.txt

Updated documentation: added development notes.txt which gives an overview of the NVDA structure, updated key commands.txt, and updated readme.txt

Again more major changes:
*Renamed a lot of methods in NVDAObjects (getCaretIndex is now getCaretPosition, things like getNextCharacterIndex are now just nextCharacter and so forth).
*Changed the representation of the caret position with in NVDAObjects to just the character index, rather than a line,character-within-line tuple.
*Greatly relaxed checking of getNext, getPrevious etc in NVDAObjects since we no longer are generating a userBuffer of a foreground window. The only thing that is checked is whether the objects are not equal, fore if they are, then this is bad MSAA implementation.
*Added support for rich edit and MS word controls via the ITextDocument com interface. Moving around with the cursor is accessible, moving around with the userBuffer is accessible. Also can find out current font name and size with insert+f. Implementation only works with text so far, have not tested tables and objects etc.
*Moved event_menuStart in to the default NVDAObject, can still overwride this in an appModule if an application needs to.
*Fixed various little bugs with checkBoxes and state changes for other controls etc.
*default NVDAObject event_focusObject now makes sure it is only executed once for the lifetime of a focusObject by setting a self.doneFocus boolean. This fixes things like tab page tabs being spoken twice.

consoleWindowclass new text reading much improved. Waits till a cycle where no change has been made and then generates the difference of the screen. Now no letters randomly go missing.

Improved detection of entering context menus. Limited the NVDAObject_consoleWindowClass to consoleWindowClass_10 (client) so it doesn't try reading in the menus. Edited all speak functions in the audio module so they only send to the synth when there is actual text, previously sometimes empty strings and spaces were being sent, and this causes a delay of a second or there abouts.

Fixed up handling of menus, like the old update, but now using NVDAObjects. menuItems are not spoken on focus unless menuMode is true, menuMode is set by menuStart event. all focus objects that are not a menu set menuMode to false. There is still a focus issue with standard menu bars when leaving, focus is not put back on the real object with focus, though in edit fields, when the caret is moved focus is restored now.

Fixed bug in script_backspace for edit fields where it would fail when on a blank line.

Re-implemented tooltip reading with NVDAObject_tooltip. Now executeEvent apparet from passing MSAAEvents to the appModule, or the object with focus, it finally also tries sending the event to the actual object the event is fore directly. MSAAEventHandler now listens for specifical caret location change events and sends them as caret events.

Replaced getText and getTextRange back in to NVDAObject_edit which were accidentily taken out when designing the new virtual buffer. virtaulBuffer_consoleWindowClass has been renamed to virtualBuffer_cursorBufferWindow and now edit fields use this virtualBuffer also. The default virtualBuffer is no longer used to view the current foreground window, but only the window with focus, too many issues the otherway.

Fixed bug where lines that contained either a % or a ~ only read from that particular symbol. Also now consoleWindowclass blank lines now say blank.

consoleWindowClasses now read new lines, or only new chars if only characters one a certain line changed. Now this means that consoleWindowClasses are pretty much completely accessible, though a few characters get dropped from time to time ... (note: I am commiting using NVDA).

consoleWindowClasses now read new lines of text... however this means that it reads command prompt lines each time you type a character since it changes... eventaully ned to re-diff each new line to check the characters that have really changed.

Added NVDAObject and virtualBuffer object for consoleWindowClass. So far we can navigate around the text, arrows and backspace etc speak well enough, but no way yet to check for new text to speak, or for automatically tracking the cursor to update the position in the virtualBuffer. For now just always press numpad subtract to move to the caret.

and textSymbols, forgot.

Added right double angle bracket to textSymbols and characterSymbols ... eventaully may think of some how looking up the coorect unicode names if they don't exist.

Added config setting under virtualBuffer: includeTableStructure, boolean default false, this sets whether table, tbody, thead and cell objects are rendered in the virtualBuffer, if they arn't the objects with in these object are still rendered. For now set to false because table reading is rather long and verbose. Also added doc strings to all NVDAObject classes to give a small description as to how they differ from the base class. Added an updateVirtualBuffer method to NVDAObject which handles whether or not it should be updated, and if so, does it. Now event_focusObject doesn't have to be so mucked up in all the classes based on this one.

virtaulBuffer for mozillaUIWindowClasses now is generated properly, has to skip past silly tooltips and menu popups that don't allow to get Next. Also taken out hasFocus check in event_focusObject for mozillaUIWindowClasses which now allows things like the all folders button in thunderbird to speak properly. When recursivly generating object lines for the virtualBuffer in a mozillaUIWindowClass, we don't recurse past document objects, put focus to them to really read the document instead.

Yet another major change, sorry.
Now all the virtual buffer code is split out in to its own file (virtualBuffer.py).
The line and character and word navigation on the numpad now always navigates the virtual buffer, and usually, the buffer contains the current foreground window. But with mozillaContentWindowClass, it contains the current document.
Of course later it will be able to contain things like internetExplorer_server documents etc.
The buffer is very dynamic: it receives all events for any objects it contains, and refreshes the parts of the buffer that change due to events.
Focus changes move the navigator cursor to the correct line in the buffer, and you can also use numpad subtract to move to the line with focus if you have moved away.
Numpad return activates the current line, as in, clicks a link etc.
document loading in mozillaContentWindows has been improved muchly.

Added navigator commands to read current, move to previous and move to next word, assigned to numpad 5, numpad 4 and numpad 6 respectively.

Added a script_navigator_character_activate (numpad+return), so far in the virtual buffer it activates the current line. getBufferText is renamed to getVirtualBuffer and updateBuffer is renamed to updateVirtualBuffer. A line of the virtual buffer now is a tuple of the line text and an activate function ... usually the object's doDefaultAction.

Now the end tag of an object's text buffer now uses getTypeString rather than just getRole.

Hmmm ok, now calling keyHasScript from keyEventHandler, though not importing the api functions directly in to its global namespace.

Fixed up code in getLineLength in NVDAObject_Edit which was causing script_backspace and script_changeSelection to not work when dealing with the end of a line where the next line was blank. Also fixed script_delete which still had some old code in it.