Re: Converting byte position to character position

On Fri, Aug 15, 2014 at 8:45 PM, Per Hedbor () @ Pike (-) importmöte
för mailinglistan <[email protected]> wrote:
> I am fairly sure there is a function somewhere in glib or gtk that
> will actually return the char position given a byte position in a
> utf-8 string.
>
> And, indeed, maping at least some of the g_utf8 family of functions
> would probably be helpful.
>
> Say, g_utf8_pointer_to_offset(utf8,utf8+byteoffset).

That quite probably would be best. However, I have to support the most
easily downloadable Windows version of Pike, so making use of this
feature would have to wait. (Though if it's significantly faster than
the double conversion, I might put a #if into the code to take
advantage of this if it's available.)

> It might be argued that the GTK module should do this conversion
> automatically.

I did toy with this possibility. Considering that I create the Pango
layout object by giving it a Unicode string (specifically,
some_object->create_pango_layout(txt)), it would make the most sense
to have xy_to_index return a character position, and in fact I wrote
the code assuming naively that this was the case. It's only just today
that I noticed exactly *why* cursor selection was doing the wrong
thing around non-ASCII characters.

However, everything seems to be in agreement that xy_to_index returns
a byte offset (looking at Python, PHP, C, and Vala bindings). Plus, of
course, there's backward compat. So I was thinking the best way would
be to create a new function, xy_to_char(), which would wrap
xy_to_index() and convert the byte position to character position
before returning it. That'd be extremely clean and tidy, and would
justify a #if guard in my code and consequent support for multiple
Pike versions. Arne was talking about me being given commit access to
take charge of the GTK2 module; this would probably be my first
feature addition.

ChrisA