Re: [Jastadd] Beaver + fullCopy()

From: Jesper qvist <jesper.oqvist_at_cs.lth.se>
Date: Wed, 26 Sep 2012 13:42:16 +0200

Hi Jesper,

I have been following the Beaver development since I was working on the
Java 7 extension for JastAddJ. The main advantage of using the Beaver
parser generator is that it is supposedly fast, but the current stable
version uses table driven parsing and apparently the new version (1.0)
will move away from table driven parsing to something that is claimed to
be even faster.

I'm pretty excited about Beaver 1.0. Here is a quote from the author
(alder) that was posted on the SourceForge forum for Beaver:

"Another point to consider - the currently available implementation
(0.9), by today's standards, is not optimal in itself. Mostly because it
is still a table-driven implementation. The upcoming rewrite (1.0) has
moved to directly coded parsers and those would probably able to live up
to the original claim "as fast as LALR can get" :-) Additionally 1.0
also generates (also directly coded) scanners, which are parser driven
(i.e. it is able to return, for example, >> as >> or as 2 consecutive >
depending on the parser state). There are few other nice goodies in that
(1.0) bag - language specification is decoupled from the implementation
(there are no Java code in the spec anymore), there are some minor, but
useful enhancements to the grammar expressiveness (for example, it can
recognize and rewrite as a list generating non-terminal the following
expression: ... a ("," a)* ...), etc."

/Jesper

On 09/26/2012 11:19 AM, Jesper Mattsson wrote:
>
> Hi Jesper
>
> It's good that it will be fixed in Beaver. I could see that any
> workaround in JastAdd would have been awkward at best.
>
> I thought that it was a dead project, but I haven't looked at the forums.
>
> We have a workaround that will do until the new version is released,
> or perhaps we'll try your patch.
>
> Thank you.
>
> Jesper
>
> *Jesper MATTSSON, MSc*
>
> Software Developer & IT Administrator
>
> Phone direct: +46 73 324 5909
>
> Email: jesper.mattsson_at_modelon.com <mailto:jesper.mattsson_at_modelon.com>
>
>
> Description: Description: Description: Modelon_2011_Gradient_RGB_400
>
> ------------------------------------------------------------------------
>
> Modelon AB
> Ideon Science Park
> SE-223 70 Lund, Sweden
>
>
>
> Phone: +46 46 286 2200
> Fax: +46 46 286 2201
>
>
>
>
> Web: http://www.modelon.com <http://www.modelon.com/>
>
> *From:*jastadd-bounces_at_cs.lth.se [mailto:jastadd-bounces_at_cs.lth.se]
> *On Behalf Of *Jesper qvist
> *Sent:* den 20 september 2012 11:37
> *To:* jastadd_at_cs.lth.se
> *Subject:* Re: [Jastadd] Beaver + fullCopy()
>
> Hello again,
>
> We can not currently do anything to "fix" this problem with Beaver in
> JastAdd. The value field in beaver.Symbol is final. It is required to
> be set during the parsing as all the parser actions may use this
> field. For example:
>
> case 626: // resource_declarations = resource_declaration.a
> {
> final Symbol _symbol_a = _symbols[offset + 1];
> final ResourceDeclaration a =
> (ResourceDeclaration) _symbol_a.value;
> return new List().add(a);
> }
>
> The problem here lies with the Beaver API and it would be a bad idea
> to design around this flaw in JastAdd since a new version of Beaver
> (1.0) will be released in the near future that, from what I've seen in
> the pre-release version, does not use this final field. Hopefully it
> will be simple to move to the new Beaver version for existing JastAdd
> projects.
>
> Developers using Beaver 9.10 or previous together with JastAdd could
> build a custom version of Beaver. This approach only requires these
> changes in beaver.Symbol:
>
> --- Java1.4Frontend/beaver/Symbol.java (revision 9319)
> +++ Java1.4Frontend/beaver/Symbol.java (working copy)
> _at__at_ -11,7 +11,7 @@
> /**
> * Represents a symbol of a grammar.
> */
> -public class Symbol
> +public class Symbol implements Cloneable
> {
> static private final int COLUMN_FIELD_BITS = 12;
> static private final int COLUMN_FIELD_MASK = (1 <<
> COLUMN_FIELD_BITS) - 1;
> _at__at_ -43,7 +43,7 @@
> /**
> * Value assigned to this symbol.
> */
> - public final Object value;
> + public Object value;
>
> /**
> * Numeric symbol ID.
> _at__at_ -160,4 +160,10 @@
> {
> return end;
> }
> +
> + public Symbol clone() throws CloneNotSupportedException {
> + Symbol copy = (Symbol) super.clone();
> + copy.value = copy;
> + return copy;
> + }
> }
>
> /Jesper
>
> On 08/22/2012 05:16 PM, Jesper Mattsson wrote:
>
> Hi
>
> I have found a problem when using the beaver flag to JastAdd. The
> beaver.Symbol class has a final field /value/, that in the default
> constructor is set to /this/. When using copy() or fullCopy(),
> then the node is copied using clone(), and thus the /value/ field
> is set to a reference to the original node. This leads to strange
> memory leaks, where old nodes are retained because of this field.
>
> We have had huge problems in Eclipse with memory leaks, and
> changing ASTNode.clone() to set /value/ to /this/ solves those
> problems. Doing this requires using reflection (since /value/ is
> final). Here is one way to do that:
>
> *public*ASTNode<T> clone() *throws*CloneNotSupportedException {
>
> ASTNode node = (ASTNode)*super*.clone();
>
> node.in$Circle(*false*);
>
> node.is$Final(*false*);
>
> *try*{
>
> Field f = beaver.Symbol.*class*.getField("value");
>
> *new*ClearFinalFieldAction().perform(f, node);
>
> } *catch*(Exception e) {
>
> e.printStackTrace();
>
> }
>
> *return*node;
>
> }
>
> *private**static**class*ClearFinalFieldAction
> *implements*PrivilegedExceptionAction {
>
> *private*Field field;
>
> *public**void*perform(Field f, Object o) *throws*Exception {
>
> field= f;
>
> AccessController./doPrivileged/(*this*);
>
> f.set(o, o);
>
> }
>
> *public*Object run() *throws*Exception {
>
> field.setAccessible(*true*);
>
> *return**null*;
>
> }
>
> }
>
> Jesper
>
> *Jesper MATTSSON, MSc*
>
> Software Developer & IT Administrator
>
> Phone direct: +46 73 324 5909
>
> Email: jesper.mattsson_at_modelon.com
> <mailto:jesper.mattsson_at_modelon.com>
>
>
> Description: Description: Description: Modelon_2011_Gradient_RGB_400
>
> ------------------------------------------------------------------------
>
> Modelon AB
> Ideon Science Park
> SE-223 70 Lund, Sweden
>
>
>
> Phone: +46 46 286 2200
> Fax: +46 46 286 2201
>
>
>
>
> Web: http://www.modelon.com <http://www.modelon.com/>
>
> This email and any attachments are intended solely for the use of
> the individual or entity to whom it is addressed and may be
> confidential and/or privileged. If you are not one of the named
> recipients or have received this email in error, (i) you should
> not read, disclose, or copy it, (ii) please notify sender of your
> receipt by reply email and delete this email and all attachments,
> (iii) Modelon does not accept or assume any liability or
> responsibility for any use of or reliance on this email.
>
>
>
>
> _______________________________________________
>
> JastAdd mailing list
>
> JastAdd_at_cs.lth.se <mailto:JastAdd_at_cs.lth.se>
>
> https://mail1.cs.lth.se/cgi-bin/mailman/listinfo/jastadd
>
> This email and any attachments are intended solely for the use of the
> individual or entity to whom it is addressed and may be confidential
> and/or privileged. If you are not one of the named recipients or have
> received this email in error, (i) you should not read, disclose, or
> copy it, (ii) please notify sender of your receipt by reply email and
> delete this email and all attachments, (iii) Modelon does not accept
> or assume any liability or responsibility for any use of or reliance
> on this email.
Received on Wed Sep 26 2012 - 13:36:15 CEST

This archive was generated by hypermail 2.3.0 : Wed Apr 16 2014 - 17:19:06 CEST