On Tue 21 Jun 2016 14:38, Zefram <zef...@fysh.org> writes:

> Andy Wingo wrote:
>>Thoughts?
>
> How was this managed in Guile 1.8?

The printers and the backtrace handling was quite different, but it used
"print states".

> It seems that you need the truncated-print mechanism to be always
> available internally, but this doesn't require that it be always visible
> to the user.  You can still require the full libraries to be loaded for
> the user to get access.
>
> Lazy loading sounds like a bad idea.  Error handling is a bad place to
> attempt something that complex and failure-prone.

That's what we do for backtrace printing in 2.2, for what it's worth.
It's a tradeoff between memory size, good errors, maintainability,
startup time, safety, duplication... the global optimimum corresponds to
no per-axis optimimum :/

Andy



Reply via email to