[e-lang] Causeway JSON format doc?

[David-Sarah Hopwood]

Kevin Reid wrote:
> On Aug 1, 2008, at 19:46, Tyler Close wrote:
>> So, the dominant IDEs are: Visual Studio, Eclipse, IntelliJ IDEA and
>> NetBeans. I think the last three, being implemented in Java, probably
>> use the Java char definition of a column. I'd expect Visual Studio to
>> use the definition of wchar_t, which is also a 16 bit unit that I
>> think is the same as a Java char. Barring new information, I think we
>> should decree that a column is a UTF-16 code unit.
> 
> I really doubt that they'd all do this. Could you please check how  
> many columns these count as in those applications?
> 
>   k̈ -- U+006B U+0308 k-diaeresis, using a combining accent (two code  
> points, two scalar values, one grapheme cluster)

Oddly, Visual Studio (7.1.3088) counts this as two columns, even though
it correctly counts é (U+0065 U+0301) as one column as I reported earlier.
Perhaps the column counting only works properly for sequences that have a
precomposed form (U+00E9 in this case). If so, that would be just a bug,
not the intended behaviour. There seem to be some other strange bugs in
which inserting a combining character causes a miscount earlier in the
line.

So, scratch my earlier comment about VS doing the Right Thing.
Right idea, but lousy implementation.

>   𐐀 -- U+10400 DESERET CAPITAL LETTER LONG I, a scalar value  
> represented as two UTF-16 code points

VS (incorrectly) counts this as two characters and two columns.

-- 
David-Sarah Hopwood