Closed rubydesign closed 10 years ago
So i poked around a bit more and it was my recent addition of compound types (thought i'd checked that)
By leaving out ComoundTypes (just uncommenting) the test will parse. Compound types "should" not be parsed in the example, and are in fact not referenced anywhere in the grammar (i just coded and tested them separately)
Seems strange, and my only guess would be that the names :array and :hash cause something ?
Nothing to do with the bug, but...
Copyright (c) 2010-2014 Kaspar Schiess
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
Please take this seriously, if we OSS developers don't, how can we expect anyone else to?
Now working on that bug ;)
This guy here is to blame ;)
+ rule(:hash) { left_brace >> ((hash_pair.as(:hash_pair) >>
+ (comma >> space? >> hash_pair.as(:hash_pair)).repeat(0)).repeat(0,1)).as(:hash)>>
+ space? >> right_brace }
This essentially redefines the '#hash' method on the parser - which makes the parser impossible to treat as a normal Ruby object. This is the bad news.
The good news is that changing the context.rb around to
def lookup(obj, pos)
@cache[pos][obj.object_id]
end
def set(obj, pos, val)
@cache[pos][obj.object_id] = val
end
yields a 10% speed increase and also 'fixes' the bug ('#hash' is not called anymore). I'll credit you in the commit.
My tip is to maybe consider the fact that rules define methods and be more cautious with that ;)
Thanks for looking at this Kasper.
And rest assured the vendoring was meant to be a temporary solution (so i have only one load directory, though i forget why that was important). If not I shall plaster your name in big letters all over the parser part. It really is great stuff.
I hope you find time to look at the project a little. Admittedly docs are sparse, but i would be happy to write more if you have questions.
And i would be very happy if you find it interesting enough to help a little :-)
Parsing ruby is not a simple thing even with a good tool.
Hey that was lightning quick.
And good tip. That fact had not really gelled in my mind, but certainly will now!
The funny thing is the test causing this used to work, but the sad one is that i don't remember all the changes i made.
Some poking around reveals that the "set" in question is a Atom::Entity with the value "def", though the cache seems to be a hash and i don't see why any conversion should be necessary. The project is on github (https://github.com/ruby-in-ruby/crystal), has no dependencies, and the bug in this commit fdb5dd4f671bdf4eaa76b4d44561fdfe4ffcddf5
caused by running: ruby test/parser/test_root.rb
TypeError: no implicit conversion of Parslet::Atoms::Entity into Integer /Users/raisa/pi/home/crystal/lib/parslet/atoms/context.rb:88:in
[]=' /Users/raisa/pi/home/crystal/lib/parslet/atoms/context.rb:88:in
set' /Users/raisa/pi/home/crystal/lib/parslet/atoms/context.rb:34:intry_with_cache' /Users/raisa/pi/home/crystal/lib/parslet/atoms/base.rb:83:in
apply' /Users/raisa/pi/home/crystal/lib/parslet/atoms/base.rb:70:insetup_and_apply' /Users/raisa/pi/home/crystal/lib/parslet/atoms/base.rb:30:in
parse' /Users/raisa/pi/home/crystal/lib/parslet/convenience.rb:28:inparse_with_debug' /Users/raisa/pi/home/crystal/test/parser/helper.rb:26:in
check_parse' /Users/raisa/pi/home/crystal/test/parser/helper.rb:58:in `block (3 levels) in runnable_methods'