UIA Word Navigation

[carlos-zamora]

Description of the new feature/enhancement

UIATextRanges don't know how to move by "word" because a terminal doesn't really know what a "word" is. This is a place to discuss what that should look like.

Proposed technical implementation details (optional)

For Windows Terminal, the user defined "wordDelimiters" and the space character could be used for this.

For ConHost, the user can't define a "wordDelimiter". So, should we just use the space character?

[carlos-zamora]

Related to #2160. Might even be the same??

I don't think its the same as #2160 since the text navigation helpers explicitly decide that requests to move by word should just do line navigation instead.

[zadjii-msft]

As an aisde, you technically can set wordDelimiters for conhost. We'll read that setting from the registry, but never exposed a UI for it ¯\_(ツ)_/¯

terminal/doc/ConsoleHostSettings.md

Line 27 in f8f0798

|`WordDelimiters` |REG_SZ |A list of characters that are considered as delimiting words (e.g. `' .-/\=|,()[]{}'`) |

terminal/src/host/registry.cpp

Lines 78 to 116 in 9b92986

	auto& delimiters = ServiceLocator::LocateGlobals().WordDelimiters;
	delimiters.clear();
	Status = RegistrySerialization::s_QueryValue(hConsoleKey,
	CONSOLE_REGISTRY_WORD_DELIM,
	sizeof(dwValue),
	REG_DWORD,
	reinterpret_cast<BYTE*>(&dwValue),
	nullptr);
	if (!NT_SUCCESS(Status))
	{
	// the key isn't a REG_DWORD, try to read it as a REG_SZ
	const size_t bufferSize = 64;
	WCHAR awchBuffer[bufferSize];
	DWORD cbWritten = 0;
	Status = RegistrySerialization::s_QueryValue(hConsoleKey,
	CONSOLE_REGISTRY_WORD_DELIM,
	bufferSize * sizeof(WCHAR),
	REG_SZ,
	reinterpret_cast<BYTE*>(awchBuffer),
	&cbWritten);
	if (NT_SUCCESS(Status))
	{
	// we read something, set it as the word delimiters
	const std::wstring regWordDelimiters{ awchBuffer, cbWritten / sizeof(wchar_t) };
	for (const wchar_t wch : regWordDelimiters)
	{
	if (wch == '\0')
	{
	break;
	}
	delimiters.push_back(wch);
	}
	}
	else
	{
	// the key isn't a REG_DWORD or a REG_SZ, fall back to our default word delimiters
	delimiters = { '\\', '+', '!', ':', '=', '/', '.', '<', '>', ';', '|', '&' };
	}
	}

[carlos-zamora]

oooh. If that's the case, then we can make the UIA Provider read what the wordDelimiters are. Then, just have it read them differently according to the project. We already have a set of UIA Providers for ConHost and another for Terminal (both with the same base class).

Why not use the definition of a word boundary as defined by unicode? http://www.unicode.org/reports/tr29/#Word_Boundaries