Bug? 'func not really 'func

func: func [ "Defines a user function with given spec and body." [catch] spec [block!] {Help string (opt) followed by arg words (and opt type and string)} body [block!] "The body block of the function" ][ throw-on-error [make function! spec body] ]

== [spec body]

== [ throw-on-error [make function! spec body] ]

== [ "Defines a user function with given spec and body." [catch] spec [block!] {Help string (opt) followed by arg wor... ----^^^^^^^^^^^^^ Take special notice of "spec [block!]" here.

-------------^^^^^^^^^^^^^^^ That should make a copy, shouldn't it? Let's check:

== [ "Defines a user function with given spec and body." [catch] spec [datatype!] {Help string (opt) followed by arg ... ----^^^^^^^^^^^^^^^^ Hmmmm... That seems a little odd?!

zot: func [ "Defines a user function with given spec and body." [catch] spec [datatype!] {Help string (opt) followed by arg words (and opt type and string)} body [datatype!] "The body block of the function"][ throw-on-error [make function! spec body]]

** Script Error: zot expected spec argument of type: datatype. ** Where: test: zot [arg] [none]

f: func [arg][none] Where did 'throw-on-error go to?! After all it's present here:

func: func [ "Defines a user function with given spec and body." [catch] spec [block!] {Help string (opt) followed by arg words (and opt type and string)} body [block!] "The body block of the function" ][ throw-on-error [make function! spec body] ] Looks like there's _two_ 'func in Rebol. Andrew Martin ICQ: 26227169 http://members.nbci.com/AndrewMartin/

Silly Andrew! It didn't go anywhere. But the rest of the bug is still there!! :-( Andrew Martin I blame my diabetic hypo... ICQ: 26227169 http://members.nbci.com/AndrewMartin/

I realised that shortly after I sent my email. Thanks!

why) I think I found out why:

== datatype! I am confused as to why 'copy seems to be evaluating the words in the block it's given. And also why it doesn't show up until after the function is put together. BTW, here's a patch for 'function, to add 'throw-on-error around it, like 'does and 'func have: if not equal? second third :function [catch] [ function: func head insert/only at load mold third :function 2 [catch] compose/deep [ throw-on-error [ (copy/deep second :function) ] ] ] Andrew Martin Falling prey to hypos... ICQ: 26227169 http://members.nbci.com/AndrewMartin/

block

put

1) Not Copy, but Func evaluates its spec block to get the datatypes supplied, instead of words. That is why you normally won't see the difference, but look: spec: [a [block!]] type? probe first second spec block! == word! f: func spec [] type? probe first second third :f block! == datatype! Load Mold converts the datatype back to word. 2) Your patch is not advisable in general. The reason: x: 1 y: 0 z: 4 u: 2 f1: does [x / y] f2: does [z / u] trial: func [blk [block!]] [ probe do blk ] trial [f1 f2] Normally yields ** Math Error: Attempt to divide by zero. ** Where: x / y , while transformed-trial: func [[catch] blk [block!]] [ probe throw-on-error blk ] transformed-trial [f1 f2] yields: ** Math Error: Attempt to divide by zero. ** Where: transformed-trial [f1 f2] , which may be useless for debugging. That is why error throwing is advisable sometimes, but sometimes it isn't. Have a look at For source as an example of balanced throwing, although the series bug I discovered still remains unrepaired. Regards Ladislav

COPY isn't evaluating the words in the block. The values in the "opt type" subblocks of THIRD :FUNC are datatypes:

block! == datatype! whereas FUNC usually needs words in those subblocks. If you don't have a word there, the contents of those subblocks are treated as samples of the datatype:

cube: func [x [integer! decimal!]][x ** 3] LOAD MOLD, Ladislav's solution, converts the datatypes back to words. Hope this makes some sense ... Eric

This is an old issue. Look:

== word!

== [ "Defines a user function with given spec and body." [catch] spec [block!] {Help string (opt) followed by arg wor...

== [block!]

== datatype! When MAKE FUNCTION! processes the spec, it does not leave the type specifiers as words. Words are replaced by their values, while other values are replaced with their datatype. For example:

== [a [string!]] So, your BLOCK! above is not the word 'BLOCK! but the DATATYPE! BLOCK!; since its type is DATATYPE!, it is replaced by DATATYPE!. I think this should be considered as a bug, and DATATYPE! values should be left as they are. HTH, Gabriele. -- Gabriele Santilli <[giesse--writeme--com]> - Amigan - REBOL programmer Amiga Group Italia sez. L'Aquila -- http://www.amyresource.it/AGI/

[...actually useful content snipped...]

Beats falling prey to hippos! -jn-