Hi Jay,

I see, thanks. I believe both methods (keeping the current value or using a default value)
are reasonable approaches. My primary goal was to have some reasonable value
for all ⎕-variables, so that I can avoid tests all over the place (in particular for ⎕IO and ⎕CT).

I would also say that a well-written APL program should not make use of this "feature".

To avoid incompatibilities with existing GNU APL programs, I believe I should not
touch the current behavior.

/// Jürgen


On 08/21/2015 12:40 PM, Jay Foad wrote:
FYI this is different from NARS2000 and Dyalog, which leave
⎕-variables set to their current value, rather than setting them to
some default value.

Jay.

On 21 August 2015 at 11:04, Juergen Sauermann
<juergen.sauerm...@t-online.de> wrote:
this is a GNU APL feature. Instead of leaving ⎕-variables undefined
and raising the according error when referenced, GNU APL sets them
to their default values when they are being localized.

    

Reply via email to