World: r3wp

that is not a valid path, but this one is:

>> type? second load "a/1.2.3/b"
== tuple!

That's right it makes the tupple valid which is wrong

The tuple *is* valid, what is invalid is just the path in your example

no, it makes the PATH invalid.

what !!!!!!

I can't believe it
[1.2.3/b] = [1.2.3  /b]

it may be valid but it has nothing to do with a path anymore

Paths have to start with a word, don't they?

path-syntax shall be something like path-syntax: [word-syntax some 
[#"/" value-syntax]]

Otherwise, they load as a block.

yep.  so the above is not a path   :-)   its a tuple, followed by 
a refinement.

Argh sorry
>> [a/1.2.3/a]
== [a/1.2.3/a]
valid path

Ok you"re right again with tuple!

Same for other types too. e.g. paren!

My last try.
It doesn't work with decimals

the key here really is that AND  ;-)

>> [a/0.3/a]
** Syntax error: invalid "decimal" -- "0.3"
** Near: (line 1) [a/0.3/a]

>> [a/0.3/a]
** Syntax error: invalid "decimal" -- "0.3"
** Near: (line 1) [a/0.3/a]

Ahaaaaaaaaaaaaa

Hmm, this must be a LOAD bug, I think

>> type? second load "a/1.1"
== decimal!

its valid in R2

neither this in R3
>> [0.3/a]
** Syntax error: invalid "decimal" -- "0.3"
** Near: (line 1) [0.3/a]

That really deserves a CC ticket

Looks like a bug indeed.

>> a: [0.3 xxx]
== [0.3 xxx]

>> a/0.3
== xxx

So you got me with the "it's a bug !"
Okkkkkkkkk :)

Well, I do not know exactly what to do now, there is an option to 
reflect the state in the DECIMAL-SYNTAX definition

At least it should be noted as a comment in the source

yes

I still think there is a problem with that form [aaa/]

It will be checked like 2 separate valid words although it's an invalid 
path

btw, I was not able to use decimals in paths in R2.
>> a: [0.3 test]
== [0.3 test]
>> a/0.3
== none

system/version?

2.7.8.3.1

aha, in R2, OK

I'm thinking its a rounding/precision error.

no, it looks like an unimplemented feature

(it is not possible to obtain different rounding when using 0.3 twice)

A change committed. "aaa/" is not accepted now, however, there are 
differences we should discuss.

I don't see the new commit, weird !

Hallo !

corrected

Another thing...

a zero length word is valid with that rule

 
 -> word-syntax succeded

That way, word-syntax must have one character length à least
word-syntax: [
	[
		slash-word
		| more-less-word

  | and word-char opt sign [#"." | not #"'"] not digit any word-char
	]
	termination 
]

good catch

>> load "a<a>"
== [a <a>]
Looks good to me.

There were some good reasons for the complexity of the rule in

Thanks I forgot to whine about that one