Subj : Different ways to embed SpiderMonkey
To   : netscape.public.mozilla.jseng
From : block111
Date : Sun Sep 12 2004 05:05 pm

Hi,
I've been reading about SpiderMonkey and wanted to better understand
different techniquest for interfacing it in c++ apps.
consider a real-life example:
server side scripting using javascript (mutlithreaded server, thread
per client or using thread pool)
aside of StandardClasses will have two properties always available to
script: header (representing http header) and session (for storing
data acros requests).

Implementing header is quite easy - it adds http headers in the two
ways:
header["Location"] = "http://www.example.com/";
header.add("Location: http://www.example.com/");

and it has a method:

header.show() returning string containing generated header; /* used
for development for example*/

Implementing session is not as easy as header. Internally, all
sessions on server could be  stored in a global_sessions map<string,
JSOBJECT>, where string indicates SESSION_ID which is managed
transparently from javascript, and all we get is that global property
session would represent a JSOBJECT (which is
global_sessions["SEESION_ID"]...) from our map. My troubles mostly
came a this point: what should be this JSOBJECT stored in the map? it
could be another map<string,string>, but I would like to be able to
store javascript functions as session data as well: (they could be
stored as javascript source, but it's not what I mean)

session.letmein = function(){ print('you are not welcome here,
bye...'); }	/* print(...) was also defined */
or
session.letmein = function(){ -- ask for a password -- }
then at any other request I would do session.letmein(); 

When a script finishes the session property is stored back into
global_sessions map by its SESSION_ID. session property also could
have a few methods - ex.: delete, empty, etc.

This requirement to store native javascript objects confuses me alot -
should I store JSOBJECT as (JSObject*)?

And here goes rainfall of questions - if the server is mutlithreaded
should there be a new JSContext allocated for every client being
processed? (or in case with pool of threads there would be separate
context for each thread). Will the session property (wich is a js
object) be garbage collected after the script finishes? Would
JS_AddRoot help in this case? Suppose thread A_t processed a request
and created in context A_c a js object that is stored in the
global_sessions map. Then thread B_t processes next request and it
retrives pointer to that JSObject corresponding to the SESSION_ID
which was created by A_t thread in context A_c. Will it be possible to
access that object from another context (B_c) ???


What would you suggest as better design alternative for the two cases
(header and session) considering better practices using SpiderMonkey c
api?


PS. I do not try to reinvent the wheel - i'm not trying to write a
server. These are question to know how to implement different stuff
using the engine.


Thakns


Pavel.

.