http://www.perlmonks.org?node_id=674301


in reply to Reversible parsing (with Parse::RecDescent?)

I'm only vaguely familiar with parse::recdescent, but I'll try to give some general hints.
* Is it necessarily possible to reverse the data through the grammar to get the command line?
No. At least, for most grammars there are multiple variations of code that will end up as the same syntax tree / data structure. IOW if the original input is unambiguous, you should be able to reverse it into something equivalent but not necessarily character-for-character the same. If the input is ambiguous you'll have more problems.

* or is this only possible for special grammars?
I think in theory it would be possible to re-generate equivalent source data for any parse tree, given unambiguous grammar and input, but I'm not sure that parse::recdescent makes it easy (or even possible) to do that given only the grammar and the output data.

Looking at your code it doesn't seem *that* difficult to come up with something to handle your cases, though you may have to deal with some duplication of grammar rules.

"What should it profit a man, if he should win a flame war, yet lose his cool?"
  • Comment on Re: Reversible parsing (with Parse::RecDescent?)

Replies are listed 'Best First'.
Re^2: Reversible parsing (with Parse::RecDescent?)
by blokhead (Monsignor) on Mar 14, 2008 at 23:54 UTC
    No. At least, for most grammars there are multiple variations of code that will end up as the same syntax tree / data structure. IOW if the original input is unambiguous, you should be able to reverse it into something equivalent but not necessarily character-for-character the same. If the input is ambiguous you'll have more problems.

    You would still be able to do the direction that the OP wants, even in an ambiguous grammar. Say the OP uses a parser to convert string to parse tree. There may be multiple parse trees for that input, but the parser will find one. That parse tree is unambiguous and refers to just one string, so you will be able to go back. Formally, the deparsing operation is always a left inverse of the (set of) parsing operation(s), but only a right inverse if the grammar is unambiguous.

    Of course, the above discussion is all in the world of theoretical context-free languages, where the parse tree contains every production that was applied. In real life, we don't parse strings, we parse a stream of tokens, and anything lost in tokenization doesn't make it into the parse tree. We also flatten/simplify parse trees on the fly, resolve syntactic sugar, and do various other shortcuts.. not to mention non-context-free things that P::RD can do. Anyway, if your parse tree contains the important syntactic structure, as ikegami says, you can always deparse it back to obtain at least a syntactically equivalent string to the original.

    blokhead

[reply]
[reply]
        The OP asks to go from string to parse tree back to the original string, not parse tree to string to original parse tree.

        blokhead

[reply]

In Section Seekers of Perl Wisdom