Webware for Python

lloyd wrote:
> hello all -
> 
> 
> these questions are based on my recent experiences with webware 0.8.1
> 
> 1) is there a facility for running code on server start-up, and storing 
> application-wide data (eg, say, a database connection pool manager)?
> 
> 2) in a java servlet container like tomcat, each context is analagous to 
> a separate application.  in webware, this separation doesn't seem to 
> exist.  each context appears to be, basically, just a mapping of a URL 
> segment to a directory.  should i run a separate application server 
> instance for each application that needs "separation" from other 
> applications?
> 
> 3) are there examples or documentation for using webware, formkit, and 
> cheetah together in an optimal way?

Hi lloyd,

I'm still pretty green with webware, and python as well, so I can't say 
much about your first two questions.

As far as documenting the webware/formkit/cheetah convergence, I find 
bits and pieces here and there between manuals, tutorials, wikis and the 
mailing list archives, but no silver bullet. Bringing the tech together 
is a "creative" thing at this point ;-). That's not to say that it 
doesn't all fit well and work great together, just that there are so 
many ways to use these tools together, but no clear resounding 'right way'.

One 'issue' is that cheetah recommends that you inherit from a cheetah 
template. This is incompatible (I can't remember exactly how) with the 
normal webkit Page inheritance system. Inheriting from Page is one of my 
favorite things about webkit, so I don't use that method. Instead I call 
templates from different levels of my Webkit->Page->SitePage->etc. 
hierarchy. The thing is there are so many different ways you can use 
cheetah, it seems hard (i.e. takes a long time) to settle into a well 
rounded sensible system.

 > 4) if one is using cheetah, are the separate writeXXX() methods in a
 > servlet really useful, considering that most of that information may be
 > in the templates?

This is the trickiest thing for me: how to keep my apps sensible while 
mixing Page inheritance with cheetah templates. Here are some starting 
points from the mailing lists:

https://sourceforge.net/mailarchive/message.php?msg_id=5624068
http://sourceforge.net/mailarchive/message.php?msg_id=8736015
https://sourceforge.net/mailarchive/message.php?msg_id=7728701%3E

Personally, I'm still experimenting. So far I'm finding that having two 
parallel inheritance trees (one for webkit servlets, one for cheetah 
templates) is too confusing. I think for bigger projects, that I'll 
continue to control layout from Servlets, and build Controller objects 
to offload the 'business logic'.

I'll still use cheetah templates, but only for layout intensive 
sections. Cheetah provides a friendly fast clean and compact way to 
write text output of all kinds, which is very handy, but I prefer to 
keep my logic and control in webware. Cheetah is so flexible that it 
works great either way.

BTW, if you haven't used SQLObject (sqlobject.org), I highly recommend 
it. For me, it is an incredible framework, and it is only getting better.

Good luck,
- Sam

On Jan 29, 2005, at 3:29 PM, lloyd wrote:

> 1) is there a facility for running code on server start-up, and 
> storing application-wide data (eg, say, a database connection pool 
> manager)?

I don't know, but what I do is have my page base class create some 
singletons when first instantiated.  That is, all pages in my website 
inherit from SitePage.  The SitePage.__init__() creates my DB store on 
the first call only.

> 2) in a java servlet container like tomcat, each context is analagous 
> to a separate application.  in webware, this separation doesn't seem 
> to exist.  each context appears to be, basically, just a mapping of a 
> URL segment to a directory.  should i run a separate application 
> server instance for each application that needs "separation" from 
> other applications?

I think the answer is No.  But I'm curious what kind of separation 
would you like?  I haven't used Tomcat and I am thinking about if 
contexts should be removed from Webware since it doesn't do anything 
useful at the moment.  Is there something good about contexts in Tomcat 
that we should build up?

> 3) are there examples or documentation for using webware, formkit, and 
> cheetah together in an optimal way?

FormKit I would say is pretty straight forward.  Cheetah is a problem 
though, because as Sam Nilsson mentioned, Cheetah.Template does not 
work well with WebKit.Page.  The issue is that Cheetah.Template tries 
to inherit from WebKit.Servlet if it exists in the sys.path.  The 
inheritence of WebKit.Page is: WebKit.Page > WebKit.HTTPContent > 
WebKit.HTTPServlet > WebKit.Servlet.  So Cheetah.Template inherits from 
lower on the tree than WebKit.Page, but there are some functions like 
callMethodOfServlet() that are handy in WebKit.Page that you lose when 
your CheetahTemplate is a servlet.  Sam Nilsson's way around this is to 
have his Page object just use Cheetah.Template objects.  I.e. his page 
"has a" template instead of "is a" template.  My solution was to remove 
the hack in Cheetah.Template that conditionally inherits from 
WebKit.Servlet.  Yuck.

Another thing you will encounter using Cheetah is how to compile them.  
Since there is not a standard way to use them, I don't want to put my 
way into the Webware repository.  I have two utilities that I use.  One 
is a CheetahServletFactory that will load .tmpl files and compile them 
when requested.  That only compiles templates that are referenced as 
pages though.  If your template inherits from another template, then 
the second one is never called directly and is never recompiled.  So 
for that case, I have a second utility that that searches for new .tmpl 
files every 3 seconds, and compiles them.  I only start it when I am 
developing.

> 4) if one is using cheetah, are the separate writeXXX() methods in a 
> servlet really useful, considering that most of that information may 
> be in the templates?

Not really.  You can use WebKit.HTTPContent which is one level below 
WebKit.Page, and it doesn't have all those convenience methods that you 
won't use.

-winston

_________________________________________
winston wolff - (646) 827-2242 - http://www.stratolab.com - learning by 
creating

Winston Wolff wrote:
>> 2) in a java servlet container like tomcat, each context is analagous 
>> to a separate application.  in webware, this separation doesn't seem 
>> to exist.  each context appears to be, basically, just a mapping of a 
>> URL segment to a directory.  should i run a separate application 
>> server instance for each application that needs "separation" from 
>> other applications?
> 
> I think the answer is No.  But I'm curious what kind of separation would 
> you like?  I haven't used Tomcat and I am thinking about if contexts 
> should be removed from Webware since it doesn't do anything useful at 
> the moment.  Is there something good about contexts in Tomcat that we 
> should build up?

it might just be a difference in architecture.  i still don't quite get 
how the server and applications/contexts relate to each other in webware.

let's say i have two unrelated applications, "inventory" and 
"human-resources", to be hosted on the same physical machine.  would i 
run MakeAppWorkDir for each:

MakeAppWorkDir -c inventory /usr/local/webapps/inventory
MakeAppWorkDir -c human-resources /usr/local/webapps/human-resources

...and then create two init scripts, run two ThreadedAppServer instances 
on different ports, add entries for each in the apache config files, 
etc...?  if so, it seems it would become unwieldy once you start getting 
into dozens of applications on the same box.

or, use MakeAppWorkDir once, and just create additional context 
directories manually:

MakeAppWorkDir -c inventory /usr/local/webapps
mkdir /usr/local/webapps/human-resources

...and just put the human resources servlets etc in that directory?


while i have the opportunity, an unrelated question (which may stem from 
my conception of contexts as isolated applications): in servlets, is it 
possible (or even desirable) to have paths interpreted relative to the 
context root?

eg, using the example above, when i reference a cheetah template i have 
to do something like this:

   template = Template(file="inventory/stock-levels.tmpl")

this feels awkward to me - again, perhaps because i think that the 
context is the application root, and therefore the application shouldn't 
have to know the name of the directory that its files are in.


thanks

> it might just be a difference in architecture.  i still don't quite 
> get how the server and applications/contexts relate to each other in 
> webware.

I'm not sure of the original intention for contexts either.  Ian 
Bicking said they wanted to allow different configuration settings in 
different contexts but that was never implemented.  At the moment, a 
context is just a way to specify multiple folders to store your HTML 
files in.

As far as "web applications" go, with Webware, there is one Python 
interpreter and that runs the Application Server, and your servlets.  
The Application Server may create multiple instances of each servlet to 
handle heavy loads.  You can specify multiple places to store your 
"logic layer" python code just by adding to sys.path.  So contexts here 
don't really do much.

> let's say i have two unrelated applications, "inventory" and 
> "human-resources", to be hosted on the same physical machine.  would i 
> run MakeAppWorkDir for each:
>
> MakeAppWorkDir -c inventory /usr/local/webapps/inventory
> MakeAppWorkDir -c human-resources /usr/local/webapps/human-resources
>
> ...and then create two init scripts, run two ThreadedAppServer 
> instances on different ports, add entries for each in the apache 
> config files, etc...?  if so, it seems it would become unwieldy once 
> you start getting into dozens of applications on the same box.
>
> or, use MakeAppWorkDir once, and just create additional context 
> directories manually:
>
> MakeAppWorkDir -c inventory /usr/local/webapps
> mkdir /usr/local/webapps/human-resources
>
> ...and just put the human resources servlets etc in that directory?

What is your goal in making separate "applications"?  If your goal is 
to isolate a severe crash in one from crashing the other, then you 
would want to run separate instances of Webware.  The downside is more 
administration of multiple application servers and more ports to 
configure with Apache.  If your goal is to organize the files relevant 
to a particular application, I would say put it all in one webware 
instance, and just use different sub-folders for each "application".

I have had few problems with page crashes bringing the whole system 
down.  I would do it all with one AppServer first, and then if you 
start to feel the need for it, separate them into multiple instances.



> while i have the opportunity, an unrelated question (which may stem 
> from my conception of contexts as isolated applications): in servlets, 
> is it possible (or even desirable) to have paths interpreted relative 
> to the context root?
>
> eg, using the example above, when i reference a cheetah template i 
> have to do something like this:
>
>   template = Template(file="inventory/stock-levels.tmpl")
>
> this feels awkward to me - again, perhaps because i think that the 
> context is the application root, and therefore the application 
> shouldn't have to know the name of the directory that its files are 
> in.

I always have file paths either relative to the current page, or 
relative to a global HTTP_ROOT variable.  I check the hostname to 
determine what I set HTTP_ROOT to.

-winston



_________________________________________
winston wolff - (646) 827-2242 - http://www.stratolab.com - learning by 
creating

> -----Original Message-----
> From: webware-discuss-admin@...
> [mailto:webware-discuss-admin@...] On=20
> Behalf Of lloyd
> Sent: Saturday, January 29, 2005 12:30 PM
> To: webware-discuss@...
> Subject: [Webware-discuss] several questions about webware
>=20
>=20
> hello all -
>=20
>=20
> these questions are based on my recent experiences with webware 0.8.1
>=20
> 1) is there a facility for running code on server start-up,=20
> and storing=20
> application-wide data (eg, say, a database connection pool manager)?

There is a database connection manager, but I've not used it before, =
sorry
not much help on this point
>=20
> 2) in a java servlet container like tomcat, each context is=20
> analagous to=20
> a separate application.  in webware, this separation doesn't seem to=20
> exist.  each context appears to be, basically, just a mapping=20
> of a URL=20
> segment to a directory.  should i run a separate application server=20
> instance for each application that needs "separation" from other=20
> applications?

This is one of the big differences between a java servlet container and
webware (beside the fact that its python and not java).  Essentially if =
you
want to isolate your "applications" form each other, then yes you need =
to
run multiple App servers from different applications directories.  That =
is
essentially what Makeappdir is for.  Basically what I do is keep related
applications together in a given app server, this way if for some reason =
I
have to restart a particular server not all of my webkit servlets need =
to go
down, only those in that particular group.  It really is a very nice =
feature
once you start using it
>=20
> 3) are there examples or documentation for using webware,=20
> formkit, and=20
> cheetah together in an optimal way?

Sorry no single source of documentation, this list is probably the best
single place to look, I've gotten lots of very good advice just by =
trolling
this list.  As for Cheetah, the point it is, is that there is no single =
best
way to use it.  It is a very good template engine and can be integrated =
with
Webware very nicely in a number of ways
>=20
> 4) if one is using cheetah, are the separate writeXXX() methods in a=20
> servlet really useful, considering that most of that=20
> information may be=20
> in the templates?
>=20
>=20

I can only tell you what I did, and it is only one way to use cheetah =
with
webware.  What I did was write my own cheetahPage which inherits from =
Page.
To it I added all the "special" methods that I use to make cheetah work =
the
way I think it should.  I also added some things to the basic Page which
also make my life easier.  Then I have all my servlets inherit from my
cheetahPage (or I have a sitePage inherit from cheetahPage first, you =
get
the idea) and my servlet then calls the compiled template.  I try to put
most of my code in the servlets and then call what I need in the =
template,
as a result my compiled templates will not run by themselves most of the
time, because they would be missing many of the methods and variables =
that
the servlets expose to them.  I find that this keeps my presentation =
code
nicely separated from my program code, and keeps the templates very =
clean.
I would be happy to share my cheetahPage with anyone if there is =
interest.

To answer your question about the writexxx methods, I have writeTemplate
(basically just an overwrite of writeContent, I could have use =
writeContent,
but didn't) and then I have a method called
cheetahTemplate('templatename').render() which returns the rendered =
template
as a string.  I usually pass that to write in my servlet.  This way I =
never
use the template directly but indirectly though my servlets.  This way =
works
very well for me.  Like I said above there is no single best way to do =
it,
and lots of ways that work well.

Jose

> thanks!
>=20
>=20
>=20
> -------------------------------------------------------
> This SF.Net email is sponsored by: IntelliVIEW -- Interactive=20
> Reporting Tool for open source databases. Create drag-&-drop=20
> reports. Save time by over 75%! Publish reports on the web.=20
> Export to DOC, XLS, RTF, etc. Download a FREE copy at=20
> http://www.intelliview.com/go/osdn_nl
> _______________________________________________
> Webware-discuss mailing list Webware-discuss@...
> https://lists.sourceforge.net/lists/listinfo/webware-discuss
>=20
>=20

Winston Wolff wrote:
> I've been thinking about future directions for Webware and WSGIKit. What 
> do you think about shifting Webware into a WSGI container, with WSGIKit 
> as one possible containee? Advantages would be greater separation of 
> concerns, and easier testing. Perhaps more flexibility if lots of WSGI 
> clients become available.

That's definitely the idea behind WSGIKit.

> Here are the responsiblities of Webware that I can come up with. The 
> first five would be handled by the Webware container, and the last three 
> would be handled by the WSGI "client" if that's the proper term.
> 
> • Interface with Apache or provide built-in web server

This is kind of the quintessential WSGI thing.  Right now there are 
several pure HTTP servers, CGI servers, and FastCGI servers.  FastCGI 
still drives me nuts when I try to configure it, and the only server is 
multi-process.  So there needs to be a little something more.

> • Decode URIs and choose the servlet to call

wsgikit.urlparse -- it's pretty similar to 0.8's parsing, without the 
contexts.  Overloading or other manipulation is done by introducing 
other parsers.  It's also where ServletFactory-like functionality comes 
in, and where PSPFactory could come into play.

> • Contain broken pages so it doesn't break the rest of the site

That's just hard.  Very useful, but hard.

> • Produce error reports when pages break.

WSGIKit doesn't have this yet, but it's definitely on my list to do.

> • Speed development by auto-reloading when source has changed.

I can't figure out how to fit this one in.  It's important.  Subway has 
some reloading code, though I know little about it.  Auto-restart 
doesn't fit quite as well, as there's no standard way in WSGI to ask the 
server to restart.  We could add extensions to do that, at least for 
select servers.

> • Manage sessions

WSGIKit has a small, kind of lame session handler.  It's really part of 
the above stack more than the "client" (the webkit-simulating portion).

> • Build an HTTP response with headers, cookies, and so forth.

This is solidly part of the webkit portion of WSGIKit, along with 
parsing the request.  But even then it's pretty small -- parsing the 
request is done mostly with the cgi module, and the response is pretty 
simple.

> • Write HTML

This is WebKit.Page, which I was able to move over with little change.

-- 
Ian Bicking  /  ianb@...  / http://blog.ianbicking.org

I just installed python2.2, mx.DateTime and python2.2-mysqldb (it
tooks about 30 sec using apt-get, gotta love debian :) :) :)

and everything works fine now.

-jacob


On Thu, 27 Jan 2005 10:42:05 -0600, jacob martinson
<martinson.jacob@...> wrote:
> Here is the generated stub that raises the exception:
> 
>   def setLastPkt(self, value):
>     assert value is not None
>     # have DateTime
>     if value is not None:
>       if type(value) is type(''):
>         value = DateTime.DateTimeFrom(value)
>       if type(value) is not DateTime.DateTimeType:
>         raise TypeError, 'expecting datetime type, but got value %r of
> type %r instead' % (value, type(value))
> 
> 
> On Thu, 27 Jan 2005 10:39:12 -0600, jacob martinson
> <martinson.jacob@...> wrote:
> > Hi,
> >
> > I am having a problem with MiddleKit.  I think the generated code is
> > not handling datetime data correctly.  Can save datatime data ok (it
> > populates the table correctly) but when I try to retrieve objects, I
> > get an exception like this:
> >
> > Traceback (most recent call last):
> >   File "./NSC.py", line 116, in ?
> >     main()
> >   File "./NSC.py", line 113, in main
> >     test3()
> >   File "./NSC.py", line 110, in test3
> >     nsc.recycleSessions()
> >   File "./NSC.py", line 23, in recycleSessions
> >     sessions = self.store.fetchObjectsOfClass('NetSession',clauses=srch)
> >   File "MiddleKit/Run/SQLObjectStore.py", line 298, in fetchObjectsOfClass
> >   File "MiddleKit/Run/MiddleObject.py", line 90, in readStoreData
> >   File "/root/airlock/Middle/GeneratedPy/GenNetSession.py", line 231,
> > in setLastPkt
> >     raise TypeError, 'expecting datetime type, but got value %r of
> > type %r instead' % (value, type(value))
> > TypeError: expecting datetime type, but got value
> > datetime.datetime(2005, 1, 27, 9, 58, 38) of type <type
> > 'datetime.datetime'> instead
> >
> > I found another user with a similar problem from back in December:
> > http://www.mail-archive.com/webware-devel@...>
> >
> > I am using:
> >
> > webware 0.8.1
> > python 2.3.4
> > mx 2.0
> > mysql 4.0.23
> >
> > I'm guessing the temporary solution to this is to use Python 2.2?
> >
> > Thanks!
> >
> > -Jacob
> >
>

Here is the generated stub that raises the exception:

  def setLastPkt(self, value): 
    assert value is not None
    # have DateTime
    if value is not None:
      if type(value) is type(''):
        value = DateTime.DateTimeFrom(value)
      if type(value) is not DateTime.DateTimeType:
        raise TypeError, 'expecting datetime type, but got value %r of
type %r instead' % (value, type(value))


On Thu, 27 Jan 2005 10:39:12 -0600, jacob martinson
<martinson.jacob@...> wrote:
> Hi,
> 
> I am having a problem with MiddleKit.  I think the generated code is
> not handling datetime data correctly.  Can save datatime data ok (it
> populates the table correctly) but when I try to retrieve objects, I
> get an exception like this:
> 
> Traceback (most recent call last):
>   File "./NSC.py", line 116, in ?
>     main()
>   File "./NSC.py", line 113, in main
>     test3()
>   File "./NSC.py", line 110, in test3
>     nsc.recycleSessions()
>   File "./NSC.py", line 23, in recycleSessions
>     sessions = self.store.fetchObjectsOfClass('NetSession',clauses=srch)
>   File "MiddleKit/Run/SQLObjectStore.py", line 298, in fetchObjectsOfClass
>   File "MiddleKit/Run/MiddleObject.py", line 90, in readStoreData
>   File "/root/airlock/Middle/GeneratedPy/GenNetSession.py", line 231,
> in setLastPkt
>     raise TypeError, 'expecting datetime type, but got value %r of
> type %r instead' % (value, type(value))
> TypeError: expecting datetime type, but got value
> datetime.datetime(2005, 1, 27, 9, 58, 38) of type <type
> 'datetime.datetime'> instead
> 
> I found another user with a similar problem from back in December:
> http://www.mail-archive.com/webware-devel@...>
> 
> I am using:
> 
> webware 0.8.1
> python 2.3.4
> mx 2.0
> mysql 4.0.23
> 
> I'm guessing the temporary solution to this is to use Python 2.2?
> 
> Thanks!
> 
> -Jacob
>

That's good.  I thinking what the problem might be, and couldn't come  
up with anything.

-winston

On Jan 27, 2005, at 11:47 AM, jacob martinson wrote:

> I just installed python2.2, mx.DateTime and python2.2-mysqldb (it
> tooks about 30 sec using apt-get, gotta love debian :) :) :)
>
> and everything works fine now.
>
> -jacob
>
>
> On Thu, 27 Jan 2005 10:42:05 -0600, jacob martinson
> <martinson.jacob@...> wrote:
>> Here is the generated stub that raises the exception:
>>
>>   def setLastPkt(self, value):
>>     assert value is not None
>>     # have DateTime
>>     if value is not None:
>>       if type(value) is type(''):
>>         value = DateTime.DateTimeFrom(value)
>>       if type(value) is not DateTime.DateTimeType:
>>         raise TypeError, 'expecting datetime type, but got value %r of
>> type %r instead' % (value, type(value))
>>
>>
>> On Thu, 27 Jan 2005 10:39:12 -0600, jacob martinson
>> <martinson.jacob@...> wrote:
>>> Hi,
>>>
>>> I am having a problem with MiddleKit.  I think the generated code is
>>> not handling datetime data correctly.  Can save datatime data ok (it
>>> populates the table correctly) but when I try to retrieve objects, I
>>> get an exception like this:
>>>
>>> Traceback (most recent call last):
>>>   File "./NSC.py", line 116, in ?
>>>     main()
>>>   File "./NSC.py", line 113, in main
>>>     test3()
>>>   File "./NSC.py", line 110, in test3
>>>     nsc.recycleSessions()
>>>   File "./NSC.py", line 23, in recycleSessions
>>>     sessions =  
>>> self.store.fetchObjectsOfClass('NetSession',clauses=srch)
>>>   File "MiddleKit/Run/SQLObjectStore.py", line 298, in  
>>> fetchObjectsOfClass
>>>   File "MiddleKit/Run/MiddleObject.py", line 90, in readStoreData
>>>   File "/root/airlock/Middle/GeneratedPy/GenNetSession.py", line 231,
>>> in setLastPkt
>>>     raise TypeError, 'expecting datetime type, but got value %r of
>>> type %r instead' % (value, type(value))
>>> TypeError: expecting datetime type, but got value
>>> datetime.datetime(2005, 1, 27, 9, 58, 38) of type <type
>>> 'datetime.datetime'> instead
>>>
>>> I found another user with a similar problem from back in December:
>>> http://www.mail-archive.com/webware-devel@...> 
>>> msg02075.html
>>>
>>> I am using:
>>>
>>> webware 0.8.1
>>> python 2.3.4
>>> mx 2.0
>>> mysql 4.0.23
>>>
>>> I'm guessing the temporary solution to this is to use Python 2.2?
>>>
>>> Thanks!
>>>
>>> -Jacob
>>>
>>
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
> Tool for open source databases. Create drag-&-drop reports. Save time
> by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
> Download a FREE copy at http://www.intelliview.com/go/osdn_nl
> _______________________________________________
> Webware-discuss mailing list
> Webware-discuss@...
> https://lists.sourceforge.net/lists/listinfo/webware-discuss
>
>
_________________________________________
winston wolff - (646) 827-2242 - http://www.stratolab.com - learning by  
creating

Actually I think it was a change in the MySQLdb API. I believe that
insert_id() moved from the cursor object to the connection object in the
latest versions.

I've fixed it either by installing an older version of MySQLdb or by
editing MySQLObjectStore.py to:

def retrieveLastInsertId(self, conn, cur):
        conn, cur = self.executeSQL('select last_insert_id();', conn)
        return cur.fetchone()[0]

-Seth

On Tue, 2005-01-25 at 02:51 -0600, jacob martinson wrote: 
> Hi,
> 
> I was just trying out middlekit for the first time and ran into a
> problem.  I was following the quick start guide and when I tried
> running the code with "store.saveChanges()" in main, I get this error:
> 
> $ ./main.py 
> Traceback (most recent call last):
>   File "./main.py", line 31, in ?
>     main()
>   File "./main.py", line 27, in main
>     store.saveChanges()
>   File "MiddleKit/Run/ObjectStore.py", line 316, in saveChanges
>   File "MiddleKit/Run/SQLObjectStore.py", line 193, in commitInserts
>   File "MiddleKit/Run/SQLObjectStore.py", line 214, in _insertObject
>   File "MiddleKit/Run/MySQLObjectStore.py", line 36, in retrieveLastInsertId
> AttributeError: 'Cursor' object has no attribute 'insert_id'
> 
> I would guess it was related to the wrong version of Mysql, but I
> didn't see anything in the webware docs about which versions are
> supported.
> 
> System details:
> debian/sarge
> 
> python2.3  2.3.4-18
> python2.3-mysqldb   1.1.6-1
> mysql-client        4.0.23-1
> mysql-server        4.0.23-1
> webware 0.8.1
> 
> Has anyone ever seen this before?
> 
> Thanks!
> 
> -Jacob
>

I think this creates another problem.  A new connection/cursor is
created for every object that is stored when you call
"store.saveChanges()".

If my mysql max_connections variable is less than the number of
objects I have to commit, I'll get an exception like this:

Traceback (most recent call last):
  File "./nstest.py", line 27, in ?
    main()
  File "./nstest.py", line 24, in main
    store.saveChanges()
  File "MiddleKit/Run/ObjectStore.py", line 316, in saveChanges
  File "MiddleKit/Run/SQLObjectStore.py", line 193, in commitInserts
  File "MiddleKit/Run/SQLObjectStore.py", line 211, in _insertObject
  File "MiddleKit/Run/SQLObjectStore.py", line 339, in executeSQL
  File "MiddleKit/Run/SQLObjectStore.py", line 375, in connectionAndCursor
  File "/root/src/Webware-0.8.1/MiddleKit/Run/MySQLObjectStore.py",
line 33, in newConnection
    return self.dbapiModule().connect(**args)
  File "/usr/lib/python2.3/site-packages/MySQLdb/__init__.py", line
64, in Connect
    return Connection(*args, **kwargs)
  File "/usr/lib/python2.3/site-packages/MySQLdb/connections.py", line
100, in __init__
    super(Connection, self).__init__(*args, **kwargs2)
_mysql_exceptions.OperationalError: (1040, 'Too many connections')

If I increase max_connections beyond the number of objects I try save
at once it works fine.

I guess the solution is to downgrade my python/mysql library or just
up max_connections for now?

Thanks!

-Jacob

How many ObjectStore objects do you have living?  I believe you are 
only supposed to have one instance for your entire application.  I had 
a lot of trouble with running out of connections when I first started 
using MiddleKit, and as I recall I fixed it by making a global 
MySQLObjectStore instance.

-winston

On Jan 25, 2005, at 5:43 PM, jacob martinson wrote:

> I think this creates another problem.  A new connection/cursor is
> created for every object that is stored when you call
> "store.saveChanges()".
>
> If my mysql max_connections variable is less than the number of
> objects I have to commit, I'll get an exception like this:
>
> Traceback (most recent call last):
>   File "./nstest.py", line 27, in ?
>     main()
>   File "./nstest.py", line 24, in main
>     store.saveChanges()
>   File "MiddleKit/Run/ObjectStore.py", line 316, in saveChanges
>   File "MiddleKit/Run/SQLObjectStore.py", line 193, in commitInserts
>   File "MiddleKit/Run/SQLObjectStore.py", line 211, in _insertObject
>   File "MiddleKit/Run/SQLObjectStore.py", line 339, in executeSQL
>   File "MiddleKit/Run/SQLObjectStore.py", line 375, in 
> connectionAndCursor
>   File "/root/src/Webware-0.8.1/MiddleKit/Run/MySQLObjectStore.py",
> line 33, in newConnection
>     return self.dbapiModule().connect(**args)
>   File "/usr/lib/python2.3/site-packages/MySQLdb/__init__.py", line
> 64, in Connect
>     return Connection(*args, **kwargs)
>   File "/usr/lib/python2.3/site-packages/MySQLdb/connections.py", line
> 100, in __init__
>     super(Connection, self).__init__(*args, **kwargs2)
> _mysql_exceptions.OperationalError: (1040, 'Too many connections')
>
> If I increase max_connections beyond the number of objects I try save
> at once it works fine.
>
> I guess the solution is to downgrade my python/mysql library or just
> up max_connections for now?
>
> Thanks!
>
> -Jacob
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
> Tool for open source databases. Create drag-&-drop reports. Save time
> by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
> Download a FREE copy at http://www.intelliview.com/go/osdn_nl
> _______________________________________________
> Webware-discuss mailing list
> Webware-discuss@...
> https://lists.sourceforge.net/lists/listinfo/webware-discuss
>
>
_________________________________________
winston wolff - (646) 827-2242 - http://www.stratolab.com - learning by 
creating

I only have one ObjectStore object, but it will be handling
persistance for several hundred other objects.  Here's the code that
was causing the problem (after I made a change to the ObjectStore code
to make it work with the newer python/mysql lib per instructions
earlier in this thread):

import os, sys
sys.path.insert(1, os.path.abspath(os.pardir))

from mx.DateTime import DateFrom
from MiddleKit.Run.MySQLObjectStore import MySQLObjectStore
from NetSession import NetSession

def main():
        # Set up the store
        store = MySQLObjectStore(user='', passwd='')
        store.readModelFileNamed('proto1.mkmodel')

        for x in range(1,255):
                s = NetSession()
                s.setUserId(x)
                store.addObject(s)

        store.saveChanges()

if __name__=='__main__':
        main()

Thanks!
-jacob

On Tue, 25 Jan 2005 18:24:29 -0500, Winston Wolff
<winstonwolff@...> wrote:
> How many ObjectStore objects do you have living?  I believe you are
> only supposed to have one instance for your entire application.  I had
> a lot of trouble with running out of connections when I first started
> using MiddleKit, and as I recall I fixed it by making a global
> MySQLObjectStore instance.
> 
> -winston
>

It sounds like you are traveling down the exact same path I went down a
few weeks ago ;) I was getting a similar MySQL "max connections
exceeded" error as well. The fix for me was to configure MiddleKit to
use an SQL connection pool. In your model's Settings.config file add the
following:

'SQLConnectionPoolSize': 20

http://www.webwareforpython.org/Webware/MiddleKit/Docs/UsersGuide.html#Configuration

-Seth

On Tue, 2005-01-25 at 21:49 -0600, jacob martinson wrote:
> I only have one ObjectStore object, but it will be handling
> persistance for several hundred other objects.  Here's the code that
> was causing the problem (after I made a change to the ObjectStore code
> to make it work with the newer python/mysql lib per instructions
> earlier in this thread):
> 
> import os, sys
> sys.path.insert(1, os.path.abspath(os.pardir))
> 
> from mx.DateTime import DateFrom
> from MiddleKit.Run.MySQLObjectStore import MySQLObjectStore
> from NetSession import NetSession
> 
> def main():
>         # Set up the store
>         store = MySQLObjectStore(user='', passwd='')
>         store.readModelFileNamed('proto1.mkmodel')
> 
>         for x in range(1,255):
>                 s = NetSession()
>                 s.setUserId(x)
>                 store.addObject(s)
> 
>         store.saveChanges()
> 
> if __name__=='__main__':
>         main()
> 
> Thanks!
> -jacob

On Jan 26, 2005, at 3:46 AM, Pierre-Yves Delens wrote:

> Bonjour,
>
> I need to create an app based on a folder living outside of 
> webware\WebKit;
> my 1st (failed) test of MakeAppWorkDir was confusing :
>
> I'm running fine a couple of contexts under (Windows)
> c:\ServApps\Webware081\WebKit.
> I guess that these should keep alive and OK if I create another app via
> MakeAppWorkDir.
>>> Question : Shall there be any interference with the run of
> MakeAppWorkDir ?

You probably want to have one running app dir for the stuff you are 
writing.  While you are writing files in that folders, the admin and 
documentation contexts stay running.  If you later want to deactivate 
them, you can comment out the lines in the Application.config file.

> - But MakeAppWorkDir is creating a new (concurrent ?)  webkit.cgi file.
> - The doc  is saying to copy this cgi in the cgi-bin folder.
>     Which is not existing in my config
>     (standalone appserver, NOT Apache in this case).
> -  I read into the standard (already existing) webkit.cgi file, where 
> the
> doc says to alter  the AppWorkDir parameter. But this was not said in 
> the
> MakeAppWorkDir doc.
> - >> Question : Shall there be 2 webkit.cgi files coexisting ?
> - >> Question : Shall there be 2 AppServer.bat files coexisting ?
>     ( AppServer.bat http)

Each time you run MakeAppWorkDir, it will generate a new .cgi adapter-, 
or mod_kit file.  But you only need one.  If you are running with CGI, 
then you can just configure Apache to find your cgi-adapter, and the 
rest should work.

>
>>> Question : About the path to the new app folder (outside of WW4PY), 
>>> what
> is the syntax for an absolute path ?
>     (I had a couple of tests, with no success. But the error was 
> perhaps not
> in the path : see above)
>     - >>> Q : are there restrictions in the naming of contexts and 
> folders
> (name length ?)

I don't know of a specific limit.  If you are using reasonable names, 
it shouldn't be a problem.  I don't use windows so I don't know if 
there will be an issue with back slashes (\).  Maybe you need to double 
them. (\\)

>     - >>> Q : is a CMD line like this OK :
>         MakeAppWorkDir.py -l -c RestCollections
> 'C:\PythonZone\LTF\DocutilsTools\RestCollections'
>
>>> Question : More generally, after runningMakeAppWorkDir.py , what 
>>> should I
> do as to complete the config of both my new app and my already existing
> contexts ?

Are you trying to use a CGI adapter?  If so, the biggest issue is to 
see that Apache can run your cgi.  Try writing another simple cgi and 
see if apache will run it.  When that's working, go to your Webware cgi 
adapter and see if it works too.  That way you isolate getting apache 
configured with getting Webware configured.

-winston
_________________________________________
winston wolff - (646) 827-2242 - http://www.stratolab.com - learning by 
creating