This was definitely helpful at prodding me to go back and try again. My
current parser looks like:
class IpParser < Parslet::Parser
root(:firewall)
rule(:firewall) { table.repeat(1).as(:firewall) >> eol }
rule(:table) { (name.as(:name) >> (chain.repeat(1)).as(:chains)
>> rule.repeat(1).as(:rules)).as(:table) >> commit }
rule(:name) { star >> line >> eol }
rule(:chain) { (colon >> word.as(:name) >> space >>
word.as(:policy) >> space >>
left_bracket >> integers.as(:packet_counter) >>
colon >> integers.as(:byte_counter) >>
right_bracket >> eol).as(:chain) }
rule(:rule) { rule_piece.repeat(1) >> eol }
rule(:rule_piece) { argument >> space? >> negation? >> space? >> word?
>> space? >> negation? }
rule(:commit) { str('COMMIT') >> eol }
rule(:star) { str('*') }
rule(:line) { match['^\n'].repeat(1) }
rule(:eol) { match["\n"] }
rule(:colon) { str(':') }
rule(:word) { match['\S'].repeat(1) }
rule(:word?) { word.maybe }
rule(:space) { match('\s').repeat(1) }
rule(:space?) { space.maybe }
rule(:dash) { str('-') }
rule(:left_bracket) { str('[') }
rule(:right_bracket) { str(']') }
rule(:integers) { match['0-9'].repeat(1) }
rule(:negation) { str('!') }
rule(:negation?) { negation.maybe }
rule(:argument) { str('-') >> word }
end
Along with:
stuff = File.read('./example').gsub(/^#.*\n/, '')
pp parse(stuff)
This fails with:
Failed to match sequence (firewall:(TABLE{1, }) EOL) at line 1 char 1.
`- Expected at least 1 of TABLE at line 1 char 1.
`- Failed to match sequence (table:(name:NAME chains:(CHAIN{1, })
rules:(RULE{1, })) COMMIT) at line 1 char 1.
`- Failed to match sequence (name:NAME chains:(CHAIN{1, })
rules:(RULE{1, })) at line 7 char 1.
`- Expected at least 1 of RULE at line 7 char 1.
`- Failed to match sequence (RULE_PIECE{1, } EOL) at line 8
char 4.
`- Failed to match [\n] at line 8 char 4.
I'm struggling with the rule(:rule_piece) rule. The trouble is there's a
number of valid rule pieces, and there's no easy way I can understand to
deliminate it. Given something like:
-a blah ! thing ! thing -b
My understanding is if I did something like (str('-') >>
word).as(:argument) >> something >> str('-') I would then have the problem
that it would begin parsing past the -. I think what I need is a way to
"rewind" a character. I've tried:
rule(:rule_piece) { argument >> nondash >> dash.present? }
rule(:nondash) { match['^-'].repeat(1) }
But this causes other weird problems where it seems to match too far and
fail on line 20 of the `example` file in my repo. Is there a better way to
handle this problem of "argument >> stuff >> until you see a dash"?
On Sun, May 11, 2014 at 12:43 PM, Torsten Ruger <[email protected]>wrote:
> Hi Ashley,
> i am just new too, but i was online, so 2 pennies:
>
> Your file seems to have a clear structure, so i would definitely go for
> the "all in one" approach, not line by line as you seem to.
> You talk about breaking up, and sub-parsers, and breaking up is good, but
> they are all just rules (not parsers). You just need one parser, with one
> root (which will probably be a repeat of the sections you have)
> And like you sort of suggested in 1) you just create rules that parse the
> sub-content. And they are made up of rules that parse the sub-content,
> until you are at strings.
>
> The "gobbling up" problem comes from not having clear ends to the rules.
> Rules are greedy and will read as much as they can, so if you write
> any.repeat, that's that.
> Good news is that your content has clear delimiters, so you just parse
> until then. So for example you'd have a rule for a line that parses those
> option strings and is delimited by newline.
> And you could have a rule that parses lines, until "commit"
>
> Hope that helps
>
> Torsten
>
>
> Ashley Penney <[email protected]>
> 11. toukokuuta 2014 18.18
> --
> Ashley Penney
> [email protected]
> Module Engineer
>
> *Join us at PuppetConf 2014**, September 23-24 in San Francisco
> - http://puppetconf.com <http://puppetconf.com/>*
>
>
--
Ashley Penney
[email protected]
Module Engineer
*Join us at PuppetConf 2014**, September 23-24 in San Francisco
- http://puppetconf.com <http://puppetconf.com/>*
