Subj : Re: Recalculating scope chain (i.e., 'with' statement).
To   : netscape.public.mozilla.jseng
From : Christopher M. Balz
Date : Tue May 31 2005 11:12 pm

[replying to Brendan Eich; forgot to CC list on first
mail; full thread copied below]

Thank you.  I believe that I understand your
explanation.  In light of your explanation, the JS 1.5
entry on 'with' at DevMo
(http://developer-test.mozilla.org/docs/Core_JavaScript_1.5_Reference:Statements:with)
did not seem accurate.  So I wrote up the following
(below) and posted it there using the Wiki feature. 
 
On the other topic, it would be helpful for debugging
purposes to capture the activation object.  I am not
as experienced in Venkmann as I'd like to be, so maybe
that is already possible and I just don't know how. 
Most of my JS debugger experience is with Interdev. 
Specifcally, sometimes stepping into the right
function closure is not practical for time reasons or
the closure simply cannot be stepped into as the
debugger does not give the opportunity.  On Interdev,
I have definitely run into situations where I needed
to check local variables but a function closure
identifier presented a "black box" to me that I
couldn't get into.

Thanks tons for your help!  My entry for 'with' in the
"JS Reference" is below:

Description 
JavaScript looks up any unqualified names within the
set of statements to determine if the names are
properties of the default object specified by the
'with' statment. If an unqualified name matches a
property in this default object, then the property is
used in the statement. Otherwise, a local or global
variable is used. 

Peformance Pro & Con 

Pro: 'with' can help reduce file size by reducing the
need to repeat an object reference identifier without
performance penalty. The scope chain change required
by 'with' is not computationally expensive. Use of
'with' will relieve the interpreter of parsing
repeated object references. 

Con: 'with' forces the specified default object to be
searched first for all lookups. Therefore all
identifiers that match formal function argument and
declared local variable names will be found more
slowly in a 'with' block. Where performance is
important, 'with' would likely only be used to
encompass code blocks that do not use function
argument and declared local variable identifiers. 

  - Chris B.
Christopher M. Balz wrote:
> It is said in books of lore that using the 'with'
> statement is computationally expensive.  But why is
> this so, if the scope chain is merely a chain of
> object references?  Would not using the 'with'
> statement merely amount to changing about two
> references (that of the child scope, and that of the
> parent scope)?


Entering and leaving the statement governed by with is
not expensive. 
It's the name lookups inside the with.  They must be
made at runtime; 
you cannot optimize a name get or set within a with
statement to a 
stack 
frame slot for all identifiers that match formal
argument and declared 
local variable names.


> Separately, are there any plans to open up access to
a
> function closure via object references?


No.  Why do you want to capture the activation object
(SpiderMonkey's 
internal class for this is named Call), exactly?

/be


Christopher M Balz
http://treelogic-swe.com
160 Lincoln Avenue, Palo Alto, CA
94301-2437, USA
(h) 650.327.0367 / (w) 650.965.2900x207
cbalz@andrew.cmu.edu
christophermbalz@stanfordalumni.org
chris@treelogic-swe.com
 
". . . / This Cabinet is formd of Gold / And Pearl & Crystal shining bright And within it opens into a World / . . . 
Another England there I saw / Another London with its Tower Another Thames & other Hills / And another pleasant Surrey Bower . . ." 
   - from "The Crystal Cabinet", a poem by William Blake.

.