local_import

[Jose]

I have problems with local_import in the latest version of the trunk.

Jose

[NetAdmin]

What type of errors or messages do you get?

[Jose]

Sorry, I did not put what the problem. I saw in the log of changes
that had something new with Imports and thought I had to do with it.

The problem is that no matter the form directly. This started to
happen with the changes to the trunk between yesterday and today.

Jose

[mdipierro]

Please tell us more. I do not understand.

[Jose]

In the modules directory of my application I have mymodule.py

When run from the controller:
mymodule = local_import ('mymodule', reload = True)

this produces no error, but when I access the functions or classes of
the module,

myobject = mymodule.MyClass(...)

it fails:

AttributeError: 'module' object has no attribute 'MyClass'

You understand?

This worked well until yesterday, after upgrading from the trunk,
started the error.

Jose

[mdipierro]

I think I fixed it. Please try again. Also check whether reload works
or not.

[Jose]

Now works!

Thanks

[Iceberg]

Sorry gentlemen, I don't get it.

First of all, the new implementation, even with the latest bugfix,
does not use the "force" parameter at all, neither python reload(). I
reckon that Jose did not test the reload feature when he said "Now
works!". Actually now the reload feature is gone.

Second, I would recommend Massimo to test the new code before
committing it. Even we don't have to follow TDD (test-driven
development), but a TAD (test-after-development) should be a minimal
rule. Otherwise you actually put those enthusiastic web2py users on a
really "bleeding" edge, with unnecessary "blood", and therefore
potentially discourage them from checking out the latest trunk often.

The last, well, and the least, I would like to suggest "If It Works,
Don't Fix It". I remember local_import(...) was not an easy job. More
than 7 months ago, it experienced some back and forth adjustment,
until Massimo commented out the "old implementation" and code
stabilized since then. So, why the sophisticated local_import(...) is
being reimplemented 2 days ago, introducing all this hustle?

[Jose]

On 3 jul, 08:05, Iceberg <iceb...@21cn.com> wrote:
> Sorry gentlemen, I don't get it.
>
> First of all, the new implementation, even with the latest bugfix,
> does not use the "force" parameter at all, neither python reload(). I
> reckon that Jose did not test the reload feature when he said "Now
> works!". Actually now the reload feature is gone.

I tried changing the code of the module, and without stopping and
restarting web2py, the change is reflected in the application. From
this, understand that the reload = True, is working well.

[mdipierro]

It is true that force is ignored. That is because __import__ always
reloads. This may seem slower but actually it may be faster than the
previous solution using exec with force=False.

he reason for the change is that the new implementation is supposed to
be equivalent, cleaner and faster.
Does it work for you?

Massimo

[Iceberg]

Are you so sure __import__() always reloads? I don't believe so, and
that is also why there is a built-in reload(). At least __import__()
does not reload in my following simple test. Jose, would you please
also try this one?

# myapp/modules/foo.py
def bar():
return 'hello world'

# myapp/controllers/test.py
def index():
return local_import('foo', reload=True)

Visit http://localhost:8000/myapp/test and you will see "hello world".
Fine. Don't restart web2py, now change bar() as return 'something
else', refresh your browser. So far I still see "hello world" but not
"something else". Besides, the myapp/modules/foo.pyc is NOT rebuilt
according to its file create timestamp, thus another evidence that the
reload did not happen. I am using latest trunk on Windows XP, Python
2.5.4

By the way, an equivalent, cleaner and faster refactoring is
legitimate, as long as it works. I will agree on a __import__() feels
better than an exec(...), so let's first get things right.

On Jul3, 7:49pm, mdipierro <mdipie...@cs.depaul.edu> wrote:
> It is true that force is ignored. That is because __import__ always
> reloads. This may seem slower but actually it may be faster than the
> previous solution using exec with force=False.
>

> The reason for the change is that the new implementation is supposed to

[mdipierro]

You are right. I fixed it in trunk and it works for me now.

Massimo

On 4 Lug, 01:22, Iceberg <iceb...@21cn.com> wrote:
> Are you so sure __import__() always reloads? I don't believe so, and
> that is also why there is a built-in reload(). At least __import__()
> does not reload in my following simple test. Jose, would you please
> also try this one?
>
> # myapp/modules/foo.py
> def bar():
>     return 'hello world'
>
> # myapp/controllers/test.py
> def index():
>     return local_import('foo', reload=True)
>

> Visithttp://localhost:8000/myapp/testand you will see "hello world".

[Iceberg]

Thanks for quick action! I also confirm latest trunk has reload
feature now. So we have a happy ending, for the Independent Day. :-)

On Jul4, 3:36pm, mdipierro <mdipie...@cs.depaul.edu> wrote:
> You are right. I fixed it in trunk and it works for me now.
>
> Massimo
>
> On 4 Lug, 01:22, Iceberg <iceb...@21cn.com> wrote:
>
> > Are you so sure __import__() always reloads? I don't believe so, and
> > that is also why there is a built-in reload(). At least __import__()
> > does not reload in my following simple test. Jose, would you please
> > also try this one?
>
> > # myapp/modules/foo.py
> > def bar():
> >     return 'hello world'
>
> > # myapp/controllers/test.py
> > def index():
> >     return local_import('foo', reload=True)
>

> > Visithttp://localhost:8000/myapp/testandyou will see "hello world".

[Jose]

The Version 1.80.1 (12/07/2010 16:32:46) still remains the problem, if
the application is served by Cherokee. If it runs directly from the
server web2py, works well.

The error it shows is:

File "/usr/home/web2py/applications/sueldos/controllers/informes.py",
line 27, in informe_ipam
reportes = local_import('reportes', reload=True)
File "gluon/compileapp.py", line 228, in <lambda>
local_import_aux(name,reload,app)
File "gluon/compileapp.py", line 164, in local_import_aux
module = __import__(name)
ImportError: No module named applications.sueldos.modules.reportes

The trunk version works well with both Cherokee and the server itself.

Regards
Jose

[mdipierro]

1.80.1 is supposed to be identical to trunk. (!?)

Do you have this file: applications/sueldos/modules/reportes.py

[Jose]

On 12 jul, 16:32, mdipierro <mdipie...@cs.depaul.edu> wrote:
> 1.80.1 is supposed to be identical to trunk. (!?)
>
> Do you have this file: applications/sueldos/modules/reportes.py
>

No, the error does not occur due to missing module.

Now I know how to avoid it, but do not know why it happens.

The steps are:
1) Download the zipped source code (web2py_src.zip)
2) I unzip.
3) Copy the application.
4) Build the config.xml file (for Cherokee)
5) Change the owner: chown-Rh www: www web2py
6) Start cherokee

Under these conditions, the error mentioned in the above message
happens.

Well, If I run web2py (python web2py.py) and stop and try again from
the page served by Cherokee, works well.

Why is that?

[mdipierro]

No clue