Re: [Caml-list] New release of Menhir (20141215)

[Gerd Stolpmann <info@gerd-stolpmann.de>]

[-- Attachment #1: Type: text/plain, Size: 9050 bytes --]

Hi,

thanks for pointing me to your project. Maybe I'm just going to use it
(well, I'd need it to make it compatible with Ocamlnet, and I also need
TLS and SASL). Regarding the question of whether Menhir is usable: 

Part of the difficulty are the BLOBs that can occur anywhere. Here is an
example from the RFC (C = sent by client, S = sent by server, i.e. what
we need to parse):

C:    a004 fetch 12 body[header]
S:    * 12 FETCH (BODY[HEADER] {342}
S:    Date: Wed, 17 Jul 1996 02:23:25 -0700 (PDT)
S:    From: Terry Gray <gray@cac.washington.edu>
S:    Subject: IMAP4rev1 WG mtg summary and minutes
S:    To: imap@cac.washington.edu
S:    cc: minutes@CNRI.Reston.VA.US, John Klensin <KLENSIN@MIT.EDU>
S:    Message-Id: <B27397-0100000@cac.washington.edu>
S:    MIME-Version: 1.0
S:    Content-Type: TEXT/PLAIN; CHARSET=US-ASCII
S:
S:    )
S:    a004 OK FETCH completed

The {342} means that a BLOB with 342 bytes follows. This could be
completely harmless if we could recognize such tokens in the lexer, and
switch the lexer to a BLOB lexer before the tokens reach the parser.
However, this is actually not as easy, as the grammar also allows such
strings at other places where they do not prefix BLOBs.

My idea was that the lexer marks such tokens as "maybe-prefixes", and we
initially only parse to that point. These maybe-prefixes are always
followed by a line ending, so either the response ends and the parser
finishes (meaning that the maybe-prefix isn't a BLOB prefix), or the
parser needs more tokens, in which case it is either a BLOB prefix
followed by a BLOB or completely invalid. I haven't checked yet: could
we inspect Menhir's state and see whether the only possible
interpretation is a BLOB prefix? I.e. whether it is in the middle of a
production

blob := blob-prefix blob-data

and there is no other way of interpreting the tokens? This would make
the decision somewhat safer how such context-dependent tokens are
handled.

Gerd

Am Donnerstag, den 18.12.2014, 11:19 -0300 schrieb Nicolas Ojeda Bar:
> Hi Gerd,
> 
> I wrote an IMAP client library in pure OCaml
> (https://github.com/nojb/ocaml-imap).  It can handle the full RFC 3501
> grammar.  This library itself is not yet ready for prime-time, but I
> spent quite a bit of time thinking about the question of parsing since
> it is such a big part of the protocol.
> 
> While the incremental parsing in the new Menhir is great I do not
> think it will help to parse IMAP because the token types of IMAP
> grammar are highly context dependent.  By this I mean that there are
> many overlapping token definitions that are valid in different parts
> of the grammar.  While this could be hacked around (by changing the
> lexer used in dummy non-terminals) the complexity of doing so quickly
> gets out of hand.  In my view this almost requires using a scannerless
> parser, which rules out ocamlyacc/menhir-type parsers.
> 
> A parser generator that could handle the IMAP grammar nicely is Dypgen
> (which generates GLR parsers).  You can almost copy the IMAP grammar
> verbatim from RFC 3501.  However it can not be made incremental.  This
> means that you need to have all the input available before starting
> the parse.  Since the amount of data is potentially very big, this
> rules using Dypgen for anything else than toy applications.
> 
> In my library I use a monadic combinator parser
> (https://github.com/nojb/ocaml-imap/blob/master/imap/imapParser.ml).
> When programming monadically you are reifying the continuation at
> every `bind` point so you get the incremental bit for free (I think
> this goes by the fancy name of `iteratees` nowadays).  On the other
> hand it suffers from poor time/space performance common to this type
> of parser.  Also, while combinator parsers can handle arbitrary
> backtracking (and you end up paying for this), IMAP itself requires
> very little, so it would seem that the full flexibility of combinators
> are not needed.
> 
> It seems that the "scannerless, incremental" space in parser
> technology is not very developed (in OCaml; maybe other languages is
> different ?).  One intriguing possibility that I am exploring (but
> this is a much longer term project) is to use PEG parsers for this.
> It would be easy to write an IMAP parser with PEG. The main problem is
> that the usual implementation of PEG parsers uses memoization to keep
> parsing time from blowing up exponentially on the length of input (
> (in OCaml the Aurochs parser generator is of this type).  This
> requires storage proportional to input length, which is a problem when
> parsing large amounts of data.  But for grammars that don't need a lot
> of backtracking (like IMAP) the memoization is not actually required,
> so I think there is an interesting space to explore between all these
> constraints.
> 
> In any case, I would love to hear about your experience if you do try
> to parse IMAP using Menhir.
> 
> Best wishes,
> Nicolas
> 
> On Thu, Dec 18, 2014 at 9:45 AM, Gerd Stolpmann <info@gerd-stolpmann.de> wrote:
> > Thanks for this, it is great news. A couple of months ago I tried to
> > develop a monadic parser for the IMAP protocol, which has a weird
> > grammar and needs some strange interactions between lexing and parsing.
> > Lacking a parser generator I started to write the parser manually, but
> > it never made good progress. It looks like Menhir can now be used for
> > this case, and I'll try it.
> >
> > Gerd
> >
> > Am Mittwoch, den 17.12.2014, 21:14 +0100 schrieb Francois Pottier:
> >> Dear OCaml users,
> >>
> >> I have recently started making a series of changes in Menhir, inspired by
> >> the work of Frédéric Bour on Merlin (a smart emacs-mode for OCaml, which
> >> uses a modified version of Menhir). Thanks to Frédéric for his stimulating
> >> ideas, and thanks to Gabriel Scherer for not letting me sleep until I
> >> promised I would do something about them! :-)
> >>
> >> I have made a first release a couple days ago. The relevant chunk of the
> >> CHANGES file is appended below. In summary, a new incremental API is
> >> available; Menhir now requires ocaml 4.02; and a couple of obscure features
> >> (--error-recovery and $previouserror) have been removed in the interest of
> >> speed and simplicity.
> >>
> >> The incremental API means that you can take a snapshot of the parser's state,
> >> essentially at no cost, at any time. It also means that the parser no longer
> >> drives the lexer; you drive the lexer, and you provide tokens to the parser
> >> when it requests them. This can be convenient if the lexer is in a monad (the
> >> Lwt monad, for instance).
> >>
> >> More changes are planned. The type "env" exposed by the new incremental API is
> >> currently opaque. We are planning to offer a range of functions that allow
> >> inspecting and building values of type "env". This should allow the user to
> >> implement new error handling and error recovery strategies outside of Menhir.
> >>
> >> The new release is available now as a .tar.gz archive:
> >>
> >>   http://gallium.inria.fr/~fpottier/menhir/menhir-20141215.tar.gz
> >>
> >> It is also available via opam ("opam install menhir").
> >>
> >> Cheers,
> >>
> >> --
> >> François Pottier
> >> Francois.Pottier@inria.fr
> >> http://gallium.inria.fr/~fpottier/
> >>
> >> 2014/12/15:
> >> New incremental API (in --table mode only), inspired by Frédéric Bour.
> >>
> >> 2014/12/11:
> >> Menhir now reports an error if one of the start symbols produces
> >> either the empty language or the singleton language {epsilon}.
> >>
> >> Although some people out there actually define a start symbol that recognizes
> >> {epsilon} (and use it as a way of initializing or re-initializing some global
> >> state), this is considered bad style. Furthermore, by ruling out this case, we
> >> are able to simplify the table back-end a little bit.
> >>
> >> 2014/12/12:
> >> A speed improvement in the code back-end.
> >>
> >> 2014/12/08:
> >> Menhir now requires OCaml 4.02 (instead of 3.09).
> >>
> >> 2014/12/02:
> >> Removed support for the $previouserror keyword.
> >> Removed support for --error-recovery mode.
> >>
> >
> > --
> > ------------------------------------------------------------
> > Gerd Stolpmann, Darmstadt, Germany    gerd@gerd-stolpmann.de
> > My OCaml site:          http://www.camlcity.org
> > Contact details:        http://www.camlcity.org/contact.html
> > Company homepage:       http://www.gerd-stolpmann.de
> > ------------------------------------------------------------
> >

-- 
------------------------------------------------------------
Gerd Stolpmann, Darmstadt, Germany    gerd@gerd-stolpmann.de
My OCaml site:          http://www.camlcity.org
Contact details:        http://www.camlcity.org/contact.html
Company homepage:       http://www.gerd-stolpmann.de
------------------------------------------------------------


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 473 bytes --]