Re: [Bug-apl] Incorrect encoding in ATF files.

[Juergen Sauermann]

Hi,

yes, as I said - ATF files are problematic.

The '⍬' problem occurs because ⍬ is not in IBMs character sets.
Also, they are using ⎕AF instead of ⎕UCS; ⎕AF is platform dependent,
so the results can differ between platforms.

I am currently looking into the matter of a really platform-independent
file format for storing, exchanging, and archiving APL workspaces (see 
10⎕CR 'VAR' for a start).
But this is subject to change because I am not yet fully satisfied with 
it. It sort
of works for GNU APL, but it uses computed left hand side of assignments
(xxx⊂B)← which is not a standard APL feature.

/// Jürgen



On 01/20/2014 03:50 AM, Kacper Gutowski wrote:
> Actually, some characters can be lost even in simple character array,
> not only inside functions, when they aren't recognized as not suitable
> for being transported literally.  Aforementioned zilde for example:
>
>        x←'⍬'  ⍝ written as quoted string to ATF file
>        )out x x
>        )in x
>        x
> ░
>        x←'ä⍬'  ⍝ written to ATF as ⎕UCS 228 9068
>        )out x x
>        )in x
>        x
> ä⍬
>
>
> -k