Dino Viehland wrote:
Makes sense - and having the same libraries loaded will certainly reduce any 
confusion...  can you just have sys.path be empty or different for the 2ndary 
hosted engines?  Also if you're using the pre-compiled modules you could 
actually just remove the compiled loader from sys.meta_path.


We allow user's spreadsheets to execute arbitrary code, including importing modules from the filesystem - so I think we're fine leaving the import hooks in place.

By the way - I'm very much enjoying porting Resolver One to IronPython 2. Particularly nice is being able to remove workarounds for bugs in IronPython 1 that are now fixed!

Many thanks and congratulations.

All the best,

Michael

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Michael Foord
Sent: Friday, October 31, 2008 9:53 AM
To: Discussion of IronPython
Subject: Re: [IronPython] IronPython 2: Oddity with Hosting API from within 
IronPython

Dino Viehland wrote:
If you also have those modules living on disk in sys.path - yes.  But honestly 
it is a little scary that you're using modules across ScriptRuntime's and I 
can't say that I'd advise that as a best practice.  Of course I don't know that 
anything will go wrong either but it might get really confusing...


The basic scenario is that our spreadsheets use a host of our
spreadsheet model libraries.

If every new document has to import these from scratch then it
drastically increases the time (and memory use) associated with creating
a new document and doing the first calculation.

Of course if importing large amounts of Python code was quicker and used
less memory it wouldn't be such an issue. ;-)

If a single runtime could have multiple Python engines associated then
we could do that, but as far as I can tell creating a new engine from a
runtime with an existing engine will always return the existing engine?

Though as they are all using the same libraries it seems sensible for
them to be shared rather than each document holding an identical copy.
If at some point we move to isolate documents using AppDomains then this
will obviously change - but the time cost of a new document having to
import everything from scratch is a factor in being able to do that.

Michael
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Michael Foord
Sent: Friday, October 31, 2008 9:40 AM
To: Discussion of IronPython
Subject: Re: [IronPython] IronPython 2: Oddity with Hosting API from within 
IronPython

Dino Viehland wrote:

We try to import from the file system before we attempt to import from the DLR 
(which includes both globals & .NET namespaces).  So in this case we'll pick up 
foobar from disk because presumably these 2 engines both share the entry in 
sys.path where foobar lives.

I think long term this logic is going to move into an importer hook because by 
CPython 3.1 the import logic may be written entirely in Python.  In that case 
you'd have the ability to re-order the import hooks so you could control the 
precedence.  But for now I think it's by design - we don't want to block 
potentially valid imports that would work in CPython (e.g. import System :) ).


So if we want to pre-populate an engine with modules we *ought* to be
hooking directly into 'sys.modules' of the hosted engines rather than
using runtime.Globals ?

Michael


-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Michael Foord
Sent: Friday, October 31, 2008 6:54 AM
To: Discussion of IronPython
Subject: [IronPython] IronPython 2: Oddity with Hosting API from within 
IronPython

Hello guys,

In Resolver One we use the IronPython hosting API from inside IronPython
code. I've noticed an oddity that is not how I would expect the hosting
API to behave if I was using it from C#.

My understanding is that the correct way to publish a module (make it
available for a ScriptEngine to import) is to set it in
'engine.Runtime.Globals'.

If I do this from within IronPython code with a module I have already
imported and then execute an import statement in the engine, the module
is re-imported (code executed) rather than using the one I have
published to the runtime globals.

If I have a 'foobar' module that prints when importing, the following
code prints twice instead of the once I would expect:

import sys
import clr
clr.AddReference('IronPython')
clr.AddReference('Microsoft.Scripting')

from IronPython.Hosting import Python
from Microsoft.Scripting import SourceCodeKind

import foobar

engine = Python.CreateEngine()
engine.Runtime.Globals.SetVariable('foobar', sys.modules['foobar'])

source = engine.CreateScriptSourceFromString('import foobar\r\n',
SourceCodeKind.Statements)
scope = engine.CreateScope()
source.Compile().Execute(scope)


*However*, if I change the code to not use Runtime.Globals, but instead
do the following, then the module is only imported once and I get one
print as expected:

hostedSys = Python.GetSysModule(engine)
hostedSys.modules['foobar'] = sys.modules['foobar']

Is there something I have overlooked here?

As a minor supplementary question, how do I get a reference to the
default ScriptScope on an engine? Is there any performance advantage in
using the default one, can I replace it, and does replacing it remove
any performance benefits we might have got? (OK, so strictly speaking
that wasn't just one question...)

All the best,

Michael Foord

--
Michael Foord
Senior Software Engineer, Resolver Systems Ltd.
[EMAIL PROTECTED]
+44 (0) 20 7253 6372

Try out Resolver One! <http://www.resolversystems.com/get-it/>

17a Clerkenwell Road, London EC1M 5RD, UK
VAT No.: GB 893 5643 79 Registered in England and Wales as company number 
5467329.
Registered address: 843 Finchley Road, London NW11 8NA, UK

_______________________________________________
Users mailing list
Users@lists.ironpython.com
http://lists.ironpython.com/listinfo.cgi/users-ironpython.com
_______________________________________________
Users mailing list
Users@lists.ironpython.com
http://lists.ironpython.com/listinfo.cgi/users-ironpython.com


--
http://www.ironpythoninaction.com/

_______________________________________________
Users mailing list
Users@lists.ironpython.com
http://lists.ironpython.com/listinfo.cgi/users-ironpython.com
_______________________________________________
Users mailing list
Users@lists.ironpython.com
http://lists.ironpython.com/listinfo.cgi/users-ironpython.com



--
http://www.ironpythoninaction.com/

_______________________________________________
Users mailing list
Users@lists.ironpython.com
http://lists.ironpython.com/listinfo.cgi/users-ironpython.com
_______________________________________________
Users mailing list
Users@lists.ironpython.com
http://lists.ironpython.com/listinfo.cgi/users-ironpython.com


--
http://www.ironpythoninaction.com/

_______________________________________________
Users mailing list
Users@lists.ironpython.com
http://lists.ironpython.com/listinfo.cgi/users-ironpython.com

Reply via email to