World: r3wp

If it is I wouldn't call that a parse problem, ... remember you have 
to divide and conquer the problems , and that looks like a compound 
problem which parsing is not the big part of.. you just need regular 
"split" method

some are ranges & some are single ports

what does 2 mean in 2/x-y ?

the 2 is a module number on a Cisco catos switch

I qould 1) split on comma to get list of entities  2) for each entity 
if it's a single num append it to block, else split again and append 
 whole range to the block

the port information is stored in a form like
set port disable 2/2,2/4,2/6-8
set port disable 3/1-5,3/7
Then the same sort of thing for the names given to the ports.

I want to extract it all into a standard format for analysis & comparison 
of IOS based information (done the IOS stuff ok)

a: [...]
some-a-with-comma: [a any ["," a]]

CIsco operating systems are IOS or CATOS on switches and output formats 
are very different.

mhinson: the above {copy result #","} does not look intended to me

ah, do I need a copy result to perhaps?

if {copy result to ...} is what you want then I understand why you 
are having trouble with the END

not-comma: complement charset ","
copy result any [not-comma | skip]

sorry, just copy result not-comma

I think the advice to break the problem down and use more than one 
parse statement is good advice. Let's see how we can do that.


I'll go through how I would approach the problem step by step. Sorry 
that I'll be squashing things on to one line but AltME paste doesn't 
work well on Mac.

:-)

I usually try to take a few small steps rather than try to get the 
answer straight away.

The test data:

>> inp:  {random 2/2,2/4-6,2/33-37}

== "random 2/2,2/4-6,2/33-37"

random is sometimes a word & sometimes a number between 1 & 1000

We'll need to look for digits so :

> digit: charset [#"0" - #"9"]                   
  
== make bitset! #{

000000000000FF03000000000000000000000000000000000000000000000000

}

Does random ever have the sequence 2/number in it?

no

specific words (not many of them) on numeric alone

So what we need to look for are the patterns 2/number and 2/number-number.

on=or

or 3/number...etc or 4/ or 12/

but not on the same occasion

Let's start with 2/ and see how we get on 


>> parse/all inp [ any [copy range "2/" some digit  (print range) 
| skip ]]
2/

2/

2/

So I've found all the sequences of 2/ but didn't capture the following 
digits.

Can you see my mistake?

I see why it has done that but I dont know how to extend the copy 
to the comma except with 
to #","

Let's change the parse rule so that I capture more than just the 
2/:


>> parse/all inp [ any [copy range ["2/" some digit]  (print range) 
| skip ]]
2/2
2/4
2/33

== true

You don't need to worry about the commas. skip will take care of 
them.

so copy range has captured everything up to the point it no longer 
matches the expression in the []  a bit like a regular expression 
match.

So now I'm capturing the first number after the 2/ but not the range. 
So let's add a little more to the parse rule to capture that:


>> parse/all inp [ any [copy range ["2/" some [digit | "-"]]  (print 
range) | skip ]]                                                 
                         
2/2
2/4-6
2/33-37

== true

Yes copy captures everything that it matches.

a bit like copy thru?

You only need to use thru when you can't specify exactly what you 
want to copy.

I think of it this way:

copy variable pattern-to-copy 


copy variable pattern-to-start-copying-from thru pattern-to-stop-copying-at

So essential in cases where the middle bit is truly unpredictable

Yes

Now we can change the parse rule to work with 3/. 4 and  /12:


> parse/all inp [ any [copy range [["2/" | "3/" | "4/" | "12/"]  
some digit]  (print range) | skip ]]                             
              
2/2
2/4
2/33
== true

the 2/ part can be any number from 1 to 13

Do you want to capture all of them?

absolutely

one-to-thirteen: ["1" digit | non-zero-digit]

Then we can come up with a pattern for them:

digit opt digit "/"

could we use 1 2 digit #"/"   ?

at-most-three: [#"0" - #"3"]
non-zero-digit: [#"1" - #"9"]

one-to-thirteen: ["1" at-most-three | non-zero-digit]

I'm presuming that the only possible values are 1 to 13.

they are, but there will never be a 0/  or 22/ that we want to avoid 
in this case

Ladislav's will select values 1 to 13 from any numbers and so is 
more correct.