406069 - nsIAccessibleText::GetTextAtOffset() not practical for getting the line at the caret

Status: RESOLVED FIXED

(Core :: Disability Access APIs, defect)

Description

[Aaron Leventhal]

For a line that ends in text that wraps to a new line, a caret position at the end of that line has the same character offset as when the caret is at the start of the next line.

This causes a problem when a screen reader wants to speak the line a user has moved to, or the current line. If the caret is at the end of the line, the screen reader ends up getting the offset for the next line and speaking the next line instead of the correct one.

However, Mozilla knows internally which line the caret is on.

The proposed solution for IA2 and ATK/AT-SPI is to allow a new magic value for char offsets.
Currently -1 can be passed to indicate end of text
Now, -2 would mean current caret position

This would allow AT's to save on calls as well, because they usually are interested in what's at the caret position.

Comment 1

[Aaron Leventhal]

Attachment: WIP -- need efficient to know when caret is at the end of a line

If anyone knows what the Gecko test for caret @ end of line would be, let me know, otherwise I'll experiment.

Comment 2

[Aaron Leventhal]

Attachment: WIP again -- still need efficient test for end-of-line-caret

Comment 3

[William Smith]

Am I correct that this would mean that if we query for the characterExtents of -2 and compare that with the characterExtents of two characters adjacent to the the caretOffset to decide which line is current?

If this is the right understanding, aren't there boundary cases where this will fail, such as drop caps where there isn't text on the rest of the line of the drop cap? 

Comment 4

[Aaron Leventhal]

You won't have to do that.

If you just ask for the line at the caret using get_textAtOffset(-2,...) we'll give you the actual line the caret is on, instead of using the caret's offset into the hypertext.

Comment 5

[Aaron Leventhal]

Attachment: Support new caret offset

Comment 6

[Ginn Chen]

Aaron, you missed nsIAccessibleText.idl in your patch.

Comment 7

[Aaron Leventhal]

Okay, that part is the same as from the previous obsolete patch. Do you need me to post a new one?

Comment 8

[Aaron Leventhal]

Attachment: With IDL

Comment 9

[Evan Yan]

patch checked in.