World: r3wp

What else should be?

Well, now I restrain ... I am most probably not going to be a fluent 
Red/System user, so I don't want to influence such a decision, if 
other ppl feel differently about the topic ...

I'll have to refer you to Boolean algebra

Also, I didn't say "incidental" about R2 and R3, I said it about 
Red and REBOL

Well, I can describe more compatibility examples like Brian's backward 
compatibility library, but I think it is absolutely unnecessary having 
mentioned the common interpreter core tests. However, I am curious 
how is that related to Boolean algebra?

Your statement seems to be that R2 and R3 are compatible. Examples 
cannot prove that. Counterexample falsifying that: for my CMS of 
towards 4000 lines, although I wrote it in basic R2 to avoid porting 
issues, I had to make extensive changes throughout and a number of 
alternative functions and wrappers to get it to work on R3

The other way around, there are web sites I can build correctly on 
R3 but not correctly on R2

To put my two cents in: I disliked PRIN the first time I saw it, 
and I still dislike it everytime I have to use it, 10+ years later.

I also find myself using PRIN regularly (i.e. much more than 1% of 
the time I _don't_ want to print a newline right away).

Pekr: I fully agree with the 80% (or 99%) usage argument, the issue 
is not there, it is in finding a suitable name for "print-without-LF" 
if PRINT is used for printing with LF. Last time we tried that, we 
didn't find any good solution, so we ended up with a single PRINT 
command to cover all use cases.

the name has to be similar but suggest it is somewhat less .. so 
how about -print ?

or, print-

I suspect that is why the t is missing for 'prin ... but then you 
have to know what it stands for

I've renamed the print-string in the C library binding to make way 
for the Red/System one

Sorry, print-line

Thanks Kaj, I forgot it was defined also in C binding.

Is there any doc how to write a test case? Or should I just hack 
it somehow?

The testing framework done by Peter WA Wood is documented here: http://static.red-lang.org/red-system-quick-test.html

You can also browse inside %red/red-system/tests/source/ to see how 
tests are done. Basically:

- language unit tests go in %tests/source/units/<feature_name>-test.reds

- compilation passing tests and compilation error tests go in %tests/source/compiler/<feature_name>-test.reds

There are two additional assertions in the float-partial branch that 
you might find useful Oldes. They aren't in the documentation yet. 
They are:


--assertF~= which takes three float!s, it compares the first two 
to be equal within a tolerance supplied as a third argument. The 
tolerance is both an absolute and relative tolerance. So, a tolerance 
of 0.01 would return true if the two floats were equal + or - minus 
0.01 or within 1 percent.

I have used a tolerance of 1e-12 in testing.

The other one is --assertF32~= which is the equivalent for float32! 
datatypes. I have used a tolerance of 1E-6 with this assertion.

I would like to write tests for enumerations: https://github.com/dockimbel/Red/pull/201

(still not finished - I must add detection if values does not exists 
already)

Kaj, compatibility can't practically be measured as a whole with 
Boolean algebra. Try fuzzy logic math :)

BrianH: that's not only fuzzy logic. That's ALL kinds of logic, not 
only fuzzy IMNSHO.

Oldes: I've emailed a skeleton test file for enum to you.

what does backtrack ?

Also in the test doc is mentioned script run-test.r, which is not 
in repository, how to run just one test case?

(backtrack - to move back source code input cursor (PC) so it looks 
better in the NEAR field:)

'backtrack function just re-position the source code cursor to point 
to the right symbol or beginning of an expression when an error is 
raised.

run-test.r: it was there until recently, I guess Peter removed it 
as we don't use it often. I guess Peter could bring it back.

do we want to allow declarations like:
	integer!: func[a [integer!]][ a ]

Ah, old question :-) There's a ticket about that opened by Rudolf...let 
me find it...

I removed run-test.r as it was becoming difficult to get working 
once we needed to have tests not only in the red-system/tests dir. 
(I now use a script in another language which seems to be more flexible 
in it's file path handling.)

I've emailed a copy of the run-test.r to Oldes.


I'll take another look at getting run-test.r to run a test from any 
directory. If I can I'll restore it, if not I'll chane the docs.

Oldes: can't find the ticket anymore, but in a nutshell, datatypes 
should be unaffected by such declaration, they are not part of the 
same namespace as variables, so to answer your question, it is allowed 
but not recommended (there are warnings against that in the specification 
document).

I know it's allowed, the question is, if it's not better to disallow 
this, or at least throw warning

A warning could be added easily, the issue is that it would then 
make all datatypes names reserved words that would need to be added 
there: http://static.red-lang.org/red-system-specs.html#section-18

Sorry, I meant "An error..."

A warning might be a better solution.

And would avoid to add them in the "keywords list".

should this:
	#enum test! [foo: 3]
	f: func[foo [c-string!]][print foo]
	f "bar"
throw a declaration error or print "bar"?

Declaration error.

Any idea what else should be tested? https://github.com/Oldes/Red/blob/float-partial/red-system/tests/source/compiler/enum-test.r

Looking at it...

Hmm, you should have separated the compilation tests from the language 
features tests...but no problem, Peter could split that later.

A test with enum vs define?

Oh I see there's already one for that.

Multiple enums using the same name {#enum test! [...] #enum test! 
[...]}?

Setting an enumerated symbol to a value that is not an integer?

BTW, there's a `print-wide` function defined at runtime for printing 
with a blank space between values. :-)