Re: is shapeSlur broken?

[David Kastrup]

Jan-Peter Voigt <address@hidden> writes:

> On 14.05.2012 12:34, David Kastrup wrote:
>> Jan-Peter Voigt<address@hidden>  writes:
>>
>>> Hello  Harm and list,
>> Could you try _not_ posting in HTML?  This mail had both a plain text as
>> well as an HTML-specified part, and both rendered awfully here.  If you
>> send just as plain text, chances are that you see what you are actually
>> sending.
> ups, sorry! It was the last thunderbird update, or whatever ...
>>> \version "2.15.38"
>>>
>>> % predicate for a number or a list of numbers
>>> #(define (number-or-list? v)
>>>
>>> (or (number? v)
>>>
>>> (and
>>>
>>> (list? v)
>>>
>>> (eval `(and ,@(map (lambda (x)`(number? ,x)) v))(interaction-environment))
>>>
>>> )))
>> What is the crazy idea with using eval?  Why don't you use
>>
>> (every number? v) here?
> ... because I sometimes don't see the obvious ... ;-) thank you!

No, it is not just that.  Using eval is a recipe for trouble because it
runs at a time where the actual lexical environment you are evaluating
this in is _gone_.  It also adds a _spurious_ layer of evaluation, so
anything but self-evaluating expressions in the list will fail with the
weirdest of consequences.

eval is a _very_ special tool with a _lot_ of strange consequences.  On
par with self-modifying code in assembly language.  When you need its
special effects, namely creating a new interpretative environment and
_interpret_ something in this environment _completely_ detached from the
code around it, its use may be called for.

But it is not the first time I have seen you propose the use of eval
where it makes absolutely no sense at all calling a complete
encapsulated Scheme interpreter instead of using the
interpreter/compiler already dealing with the surrounding expression.

-- 
David Kastrup