[REBOL] Bug! - maybe not?? Re:
From: joel::neely::fedex::com at: 3-Oct-2000 7:06
[Al--Bri--xtra--co--nz] wrote:
> Bug! - 'insert can't handle a binary value in a block as it's 'value
> argument
Note first that #{012345} is a three-byte sequence...
> >> head insert tail #{012345} #{0123}
> == #{0123450123} ; Correct!
because #{0123} is a two-byte sequence, the result is the expected
sequence of five bytes.
> >> head insert/only tail {012345} [{0123}]
> == "0123450123" ; Correct!
because {012345} is a six-byte string, and {0123} is a four-byte
string, the result is a ten-byte string. The content of the block
is treated as a string.
> >> head insert/only tail #{012345} [#{0123}]
> == #{012345237B303132337D} ; Wrong!
because #{012345} is a three-byte sequence, and the content of
the block is treated as the seven-byte string "#{0123}"
> >> head insert tail {012345} [{0123}]
> == "0123450123" ; Correct!
again, inserting a string into a string...
> >> head insert tail #{012345} [#{0123}]
> == #{012345237B303132337D} ; Wrong!
again, converting the block to a string, then inserting that
string as a sequence of bytes. Note that
>> any-string? #{1234}
== true
So, my CONJECTURE is that the second argument to join and insert
in these puzzling cases is simply being converted to a string,
which is then stuffed into the appropriate place in the first
argument. The result has the type of the first argument, and
is therefore binary! in the puzzlers.
It appears that join and insert like to have strings in some
not-totally-obvious cases:
>> probe foo: join #{2d} ["(minus)" '+ #{28706c757329}]
#{2D286D696E7573292B237B3238373036433735373332397D}
== #{2D286D696E7573292B237B3238373036433735373332397D}
>> to-string foo
== "-(minus)+#{28706C757329}"
>> probe foo: join none ["(minus)" '+ #{28706c757329}]
"none(minus)+#{28706C757329}"
== "none(minus)+#{28706C757329}"
>> probe foo: join 0 ["(minus)" '+ #{28706c757329}]
"0(minus)+#{28706C757329}"
== "0(minus)+#{28706C757329}"
>> probe foo: join #{2b} #{28706c757329}
#{2B28706C757329}
== #{2B28706C757329}
>> to-string foo
== "+(plus)"
However, this example highlights the urgency of having a complete
specification of the REBOL language. This would allow us to
a) read a specification ahead of time and avoid any unpleasant
surprises in the behavior of our code, and
b) (when surprised by our code) read the specification to learn
whether our surprise is because REBOL didn't behave according
to the specification (in which case we have discovered a bug in
the implementation of REBOL) or because we didn't read the spec
in advance (in which case we have discovered a bug in our mental
model of REBOL).
-jn-
