Parslet: subtrees merged without warning?

addis_a · July 6, 2013, 5:29pm

Hi,

I miss subtrees which I wanted to be generated by certain “.as”'s and I
get no “Duplicate subtrees while merging result…”-warning.

When I add additional “.as” at higher level rules, I get these missed
subtrees, that’s OK.

Is there a way to get a warning or notice every time an “.as” is not
routed to the result?

I want to avoid loosing results without notice.

Or, in which cases results are lost/merged silently?

Kind regards,
Axel

Using:

ruby 1.9.3p327 (2012-11-10) [i386-mingw32]
parslet (1.5.0)

worldfox · July 6, 2013, 8:16pm

There was an error in my previous post. I corrected the message text
which I do not get. It is:

“Duplicate subtrees while merging result of…only the values of the
latter will be kept”

Sorry about this,
Axel

worldfox · July 8, 2013, 2:00pm

Or, in which cases results are lost/merged silently?

Only in one case: When merging a subtree containing .as(…) with
another subtree that doesn’t. The solution is in most cases to add more
annotations (.as).

regards,
kaspar

worldfox · July 8, 2013, 2:07pm

Hallo Kaspar,

thank you for your answer and thank you for Parslet!

Could I easily add some extension code to get a message in this case,
too? If yes, where is/might be the “start point”?

Kind regards,
Axel

Kaspar S. wrote in post #1114730:

Or, in which cases results are lost/merged silently?

Only in one case: When merging a subtree containing .as(…) with
another subtree that doesn’t. The solution is in most cases to add more
annotations (.as).

worldfox · July 10, 2013, 8:32am

Could I easily add some extension code to get a message in this case,
too? If yes, where is/might be the “start point”?

Probably

github.com

kschiess/parslet/blob/master/lib/parslet/atoms/can_flatten.rb#L87


      
            # Maybe classes are not equal, but both are stringlike?
            if l.respond_to?(:to_str) && r.respond_to?(:to_str)
              # if we're merging a String with a Slice, the slice wins. 
              return r if r.respond_to? :to_slice
              return l if l.respond_to? :to_slice
          
          
    fail "NOTREACHED: What other stringlike classes are there?"
            end
          
          
  # special case: If one of them is a string/slice, the other is more important 
            return l if r.respond_to? :to_str
            return r if l.respond_to? :to_str
          
          
  # otherwise just create an array for one of them to live in 
            return l + [r] if r.class == Hash
            return [l] + r if l.class == Hash
          
          
  fail "Unhandled case when foldr'ing sequence."
          end
          
          
# Flatten results from a repetition of a single parslet. named indicates

Such a feature would be quite the opposite of what result flattening is
about - and would probably not make it into parslet, unless done
cleverly. As is, for example, the case with error reporters that can be
swapped out.

cheers
kaspar

worldfox · July 11, 2013, 8:36am

I think, I haven’t understood the “philosophy” of Parslet yet. There are
one or two more issues, where I haven’t understood why Parslet is
“thinking” the way it does. But I must admit, Parslet is the first
Parser I use (and it is the first time, I use Parslet) and I’m not a
professional programmer. I think, I will have to work with it quite a
lot more.

The parslet mailing list has a bunch of friendly people that don’t mind
helping out. Welcome over there!

kaspar

worldfox · July 10, 2013, 10:05am

Probably

github.com

kschiess/parslet/blob/master/lib/parslet/atoms/can_flatten.rb#L87


      
            # Maybe classes are not equal, but both are stringlike?
            if l.respond_to?(:to_str) && r.respond_to?(:to_str)
              # if we're merging a String with a Slice, the slice wins. 
              return r if r.respond_to? :to_slice
              return l if l.respond_to? :to_slice
          
          
    fail "NOTREACHED: What other stringlike classes are there?"
            end
          
          
  # special case: If one of them is a string/slice, the other is more important 
            return l if r.respond_to? :to_str
            return r if l.respond_to? :to_str
          
          
  # otherwise just create an array for one of them to live in 
            return l + [r] if r.class == Hash
            return [l] + r if l.class == Hash
          
          
  fail "Unhandled case when foldr'ing sequence."
          end
          
          
# Flatten results from a repetition of a single parslet. named indicates

Thank you for pointing to this.

Such a feature would be quite the opposite of what result flattening is
about …

I think, I haven’t understood the “philosophy” of Parslet yet. There are
one or two more issues, where I haven’t understood why Parslet is
“thinking” the way it does. But I must admit, Parslet is the first
Parser I use (and it is the first time, I use Parslet) and I’m not a
professional programmer. I think, I will have to work with it quite a
lot more.

Regards,
Axel

worldfox · July 11, 2013, 2:09pm

I ask, because I see only 2 messages there for 2013.

Yeah, gmane has a problem lateley. Here:
http://kschiess.github.io/parslet/contribute.html

k

worldfox · July 11, 2013, 9:10am

The parslet mailing list has a bunch of friendly people that don’t mind
helping out. Welcome over there!

Is that right?:
http://news.gmane.org/gmane.comp.lang.ruby.parslet

I ask, because I see only 2 messages there for 2013.

Axel

worldfox · July 11, 2013, 6:38pm

Yeah, gmane has a problem lateley. Here:
parslet - Contribute

Ah, that’s the problem (gmane). On
http://librelist.com/browser/ruby.parslet/
I see a lot more discussions on Parslet. I’ll read through the archive
and see if I find helpful things.

Thank you
Axel