Re: CiderPress v2.3.1 released

[pausch@saaf.se (Paul Schlyter)]

In article <-P2dnWPtcZ-MXzTenZ2dnUVZ_vydnZ2d@comcast.com>,
Michael J. Mahon <mjmahon@aol.com> wrote:

> Paul Schlyter wrote:
> 
> <snip>
> 
>> Applesoft lax syntax checking also allowed some oddities, like:
>> 
>>  20 IF A <=> B THEN .....
>> 
>> Of course "A <=> B" always evaluated as "true" since every number is
>> always less than, equal to, or greater than, every other number.
> 
> Ha!  That's exactly what I would have expected, since that's
> actually a semantic check.  Each additional relational just
> ORs with the previous ones, if any.
> 
> -michael

Afair the Applesoft interpreter implemented the relational checks like
this:

When interpreting the relational operators, a small bitmap of 3 bits
was built, and each bit was set or cleared depending on whether each
of the < = > operators was present or absent: when a relational
operation was encountered, the corresponding bit in the bitmap was
set.  This also offered a syntax check: if a relational operation was
encountered and the corresponding bit already was set, this meant that
the operation was encountered a second time, and this was treated as a
syntax error.  Thus you cannot do things like "IF A <=<= B" in Applesoft.
(it would have been easy to also add a check if all the three bits of
the bitmap was set, and to also treat that case as a syntax error)

Next, the actual operations were compared.  First a type check was made:
both operators (where each operator could be a constant, a variable, or
an arbitrary expression) had to be both strings or both numbers; trying
to compare a number to a string yielded a syntax error.  Then the actual
comparison was done, and a similar 3-bit bit map was built, but with only
one bit set - which one depended on whether the first variable was less
than, equal to, or greater than, the second variable.

Finally, the two 3-bit bit maps were AND'ed with one another: if the AND
was non-zero, the result of the comparison was "true"; if it was zero,
the result was "false".

-- 
----------------------------------------------------------------
Paul Schlyter,  Grev Turegatan 40,  SE-114 38 Stockholm,  SWEDEN
e-mail:  pausch at stockholm dot bostream dot se
WWW:     http://stjarnhimlen.se/