Comments on: Gazelle is (at least briefly) the most powerful top-down parser generator there is http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/ parsing, performance, minimalism with C99 Mon, 06 Feb 2012 23:44:51 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: The importance of being earnest » Josh the Outspoken http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/comment-page-1/#comment-1143 The importance of being earnest » Josh the Outspoken Mon, 12 Jan 2009 01:36:55 +0000 http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/#comment-1143 [...] too frequently made claims that are unsupported or premature. The worst example of this was when I claimed to have found an algorithm that computes LL(k) for all k (this is undecidable — my algorithm turned out to be a heuristic). In other cases, what [...] [...] too frequently made claims that are unsupported or premature. The worst example of this was when I claimed to have found an algorithm that computes LL(k) for all k (this is undecidable — my algorithm turned out to be a heuristic). In other cases, what [...]

]]> By: Sylvain http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/comment-page-1/#comment-1124 Sylvain Tue, 07 Oct 2008 16:32:56 +0000 http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/#comment-1124 It is unsolvable whether or not a context-free grammar is LL(k) for some k>=0: see Rosenkrantz and Stearns, Properties of deterministic top-down grammars, Information and Control 17:226--256, 1970 http://www.dx.doi.org/10.1016/S0019-9958(70)90446-8 It is unsolvable whether or not a context-free grammar is LL(k) for some k>=0: see Rosenkrantz and Stearns, Properties of deterministic top-down grammars, Information and Control 17:226–256, 1970 http://www.dx.doi.org/10.1016/S0019-9958(70)90446-8

]]> By: Adrian Thurston http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/comment-page-1/#comment-1011 Adrian Thurston Wed, 30 Jul 2008 17:01:35 +0000 http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/#comment-1011 Hi Josh, you should have a look at the generalized top-down parsing algorithms. These algorithms handle any grammar except those with left recursion. Some of them also include additional hacks for dealing with left recursion. Algorithms/technologies to look at: GRDP, Unger's Method, Combinatorial Parsing, Packrat Parsing, DCGs and TXL, Not all of these are practical, but since they are general (barring left recursion) and LL(*) is not, any tool that implements these parsing methods must be more powerful than Gazelle. Hi Josh, you should have a look at the generalized top-down parsing algorithms. These algorithms handle any grammar except those with left recursion. Some of them also include additional hacks for dealing with left recursion. Algorithms/technologies to look at: GRDP, Unger’s Method, Combinatorial Parsing, Packrat Parsing, DCGs and TXL, Not all of these are practical, but since they are general (barring left recursion) and LL(*) is not, any tool that implements these parsing methods must be more powerful than Gazelle.

]]> By: josh http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/comment-page-1/#comment-1002 josh Mon, 28 Jul 2008 17:12:13 +0000 http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/#comment-1002 @maetl: thanks! It's funny, when I started writing Gazelle, I believed pretty strongly in that mantra ("parsing and grammars are amongst the most studied and well understood areas of computer science"). I didn't expect to invent any new algorithms, I mainly just wanted to take existing algorithms and give them different packaging. But before long I was trying to figure out if I could improve on the properties of those algorithms that made them less that ideal from a user perspective. I studied the documentation from existing tools and looked at the cases they said they couldn't handle, and tried to figure out if I could handle them. And before I knew it, I had thought of a new algorithm! @Kay: thanks! I agree that having the tool be able to tell you definitively that it is not LL(k) or LL(*) will make things much more user-friendly. I could still use a bit of work on error messages that give you every bit of information about *why* it was not LL(k) or LL(*). In the future I should be able to draw pretty graphs that make the problem obvious. That will take a bit of work, but it's totally doable, and I think will make for a compelling user experience. @maetl: thanks! It’s funny, when I started writing Gazelle, I believed pretty strongly in that mantra (“parsing and grammars are amongst the most studied and well understood areas of computer science”). I didn’t expect to invent any new algorithms, I mainly just wanted to take existing algorithms and give them different packaging.

But before long I was trying to figure out if I could improve on the properties of those algorithms that made them less that ideal from a user perspective. I studied the documentation from existing tools and looked at the cases they said they couldn’t handle, and tried to figure out if I could handle them. And before I knew it, I had thought of a new algorithm!

@Kay: thanks! I agree that having the tool be able to tell you definitively that it is not LL(k) or LL(*) will make things much more user-friendly. I could still use a bit of work on error messages that give you every bit of information about *why* it was not LL(k) or LL(*). In the future I should be able to draw pretty graphs that make the problem obvious. That will take a bit of work, but it’s totally doable, and I think will make for a compelling user experience.

]]> By: Kay Roepke http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/comment-page-1/#comment-1001 Kay Roepke Mon, 28 Jul 2008 04:23:03 +0000 http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/#comment-1001 That's great news! I just love "tell you definitively when a grammar is not LL(k) or LL(*) (others make you tweak search depth constants)" which can be a hassle currently, especially if the cause is not immediately obvious. All power to innovators! :) That’s great news!

I just love “tell you definitively when a grammar is not LL(k) or LL(*) (others make you tweak search depth constants)” which can be a hassle currently, especially if the cause is not immediately obvious.

All power to innovators! :)

]]> By: maetl http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/comment-page-1/#comment-1000 maetl Mon, 28 Jul 2008 02:18:51 +0000 http://blog.reverberate.org/2008/07/27/gazelle-is-at-least-briefly-the-most-powerful-top-down-parser-there-is/#comment-1000 Well done Josh... Great to see that you're not content to let the standard dogma that "parsing and grammars are amongst the most studied and well understood areas of computer science" get in the way of making new implementations in this field... Well done Josh…

Great to see that you’re not content to let the standard dogma that “parsing and grammars are amongst the most studied and well understood areas of computer science” get in the way of making new implementations in this field…

]]>