Webware for Python

Not positive, but I think if you were to switch to "Dynamic" session store,
then it won't load the sessions from disk at startup, but instead will leave
them in files and load them into memory only as needed.  So that might speed
things up.

- Geoff

> -----Original Message-----
> From: Aaron Held [mailto:aaron@...]
> Sent: Wednesday, March 27, 2002 9:50 AM
> To: webware-discuss@...;
> webware-devel@...
> Subject: [Webware-discuss] Webware takes a long time to load
> 
> 
> I have a production webware site that I shut down once a month
> for some maintainence on the database site.
> 
> The startups are taking a very long time.
> 
> Turns out that the delay is the system taking time to reload the
> session data that it previously saved to disk.  I have anywhere
> from 1-20 sessions and they can be holding 120 k each.  (During
> runtime I use memory sessions and its pretty quick)
> 
> This was a really frustrating bug and I spent a lot of time
> looking at the __init__.py of my context since the last message
> on the console was "loading context"
> 
> It would be great to have a another console message saying
> "restoring sessions...."
> 
> Thanks,
> -Aaron Held
> 
> On a side note I'm doing another project using postnuke, a PHP
> CMS "slashdot" system.  After working with webware I forgot how
> much fun it was to look through dozens of php files all over the
> filesystem to find the one variable or function I need.
> 
> 
> _______________________________________________
> Webware-discuss mailing list
> Webware-discuss@...
> https://lists.sourceforge.net/lists/listinfo/webware-discuss
>

I have a production webware site that I shut down once a month
for some maintainence on the database site.

The startups are taking a very long time.

Turns out that the delay is the system taking time to reload the
session data that it previously saved to disk.  I have anywhere
from 1-20 sessions and they can be holding 120 k each.  (During
runtime I use memory sessions and its pretty quick)

This was a really frustrating bug and I spent a lot of time
looking at the __init__.py of my context since the last message
on the console was "loading context"

It would be great to have a another console message saying
"restoring sessions...."

Thanks,
-Aaron Held

On a side note I'm doing another project using postnuke, a PHP
CMS "slashdot" system.  After working with webware I forgot how
much fun it was to look through dozens of php files all over the
filesystem to find the one variable or function I need.

I want to cut 0.7b3 on Monday, so if anyone has any more fixes to put in, 
please do it this weekend.

Then I'll cut 0.7 final by the end of the week if no showstopper bugs are 
found.

- Geoff

Fixed a recently introduced bug in PSP servlet generation which would cause
errors on Windows (not sure about Linux) when a PSP file already existed in
the cache.

Fixed PickleRPCServlet and XMLRPCServlet so that they properly handle
old-style string exceptions.

- Geoff

James wrote:

>>I am evaluating to use MiddleKit for our new project
development and would like to clarify how MiddleKit
with MySQL store data without losing data integrity.
For my case, I have an object (order) with couple of
sub-objects - order line, if one of the order line
update was fail, the whole order including any order
lines should not be stored. Did MiddleKit have
"transaction processing" in place? Anyone has the same
concern?<<

MiddleKit does not support transactions, and MySQL is not ACID compliant.
Nobody that I'm aware of has plans to add transactions to MK, and the
transaction support in MySQL is best described as "unproven."

If you're trying to develop an app to handle (presumably) mission critical
stuff like orders, you really ought to use a more robust DB. And, if you're
ever intending to support ad-hoc queries of your data (as most order
processing systems do), you're better off using a standard relational model
(IMO).

Check the Webware Wiki pages on Database Integration for more discussions
and links on DB related matters.

...Edmund.

Hi,

I am evaluating to use MiddleKit for our new project
development and would like to clarify how MiddleKit
with MySQL store data without losing data integrity.
For my case, I have an object (order) with couple of
sub-objects - order line, if one of the order line
update was fail, the whole order including any order
lines should not be stored. Did MiddleKit have
"transaction processing" in place? Anyone has the same
concern?

Thank you.

James



__________________________________________________
Do You Yahoo!?
Yahoo! Sports - live college hoops coverage
http://sports.yahoo.com/

On Friday 15 March 2002 07:15 pm, Tripp Lilley wrote:
> I'm planning on converting wholesale to Python 2.2. While I'm doing
> so, I'm going to spend some head-time getting to know iterators,
> generators, the new attributes stuff, and so on.
>
> What are you feelings on making MK require 2.2? One of the things I
> don't like about it right now is the impedance mismatch between how
> MK objects behave and how pure Python objects behave. I -think-
> (though I haven't explored it fully) that some of 2.2's features
> could lessen, if not eliminate, this mismatch.
>
> Since MK's thoroughly advertised as being alpha right now anyway,
> and, to the best of my knowledge, you, me, and Geoff are the only
> people with actual serious production apps running on top of it, I
> think now would be the most appropriate time for me to take on such a
> project.

I've considered the same thing, but mostly regarding properties. I'm 
not touching generators any time soon and iterators are pure frosting 
that we can get around to whenever. I don't think iterators will 
actually shorten your code. eg, you'll still have pretty much the same 
for loops.

BUT Python 2.2 has more caveats about its new features than I can fit 
in my head. It also has some bugs that have been mentioned on their 
site, but I don't remember them off the top of my head. And Guido has 
admitted that the syntaxes for things like invoking super and 
specifying properties are less than ideal.

For the first time, I'm actually skipping a Python version. I prefer to 
wait for the next one hoping that bugs, caveats and syntactic ugliness 
get reduced.

I would make the same recommendation to you, but it's your call. 
There's no reason we couldn't enhance MK to support 2.0 through 2.2 as 
Python seems flexible enough for that sort of thing.

But frankly, 2.2 isn't exactly a pot of honey. Just a step in the right 
direction.

BTW There are more MK users coming out of the woodwork on the mailing 
list, and had I thought of it, I would have marked it beta for the last 
release.

Also, I plan that a future version of Webware would use the 
attributes/properties style rather than the current all-methods style. 
Such a change would come with a script to update existing sites. We 
have, of course, big juicy projects to develop and test such a script 
with.  :-)

	self.store.model.klass('foo')
	self.session['foo']

I CCed webware-devel as these seem like thoughts to be shared.


-Chuck

I've experimented some with this using the built in server on my
branch.  I've just reloaded HTML, PSP and image files.

Mozilla (linux & WIndows),as far as I can tell, never sends an
If-Modified_since.
Netscape 4.x(Windows) never sends it.
IE 5.5 (Windows) never sent it.

Opera sends it for images and *.css on a refresh.  

So, I only got Opera to send it.  Everything else sent GET.

Anyone else have any thoughts here?

Jay





On Thu, 2002-03-14 at 22:18, Chuck Esterbrook wrote:
> On Thursday 14 March 2002 04:16 pm, Chuck Esterbrook wrote:
> > On Wednesday 13 March 2002 10:43 am, Chuck Esterbrook wrote:
> > > I just looked at my apache access_log and all the requests for the
> > > style sheet are GET and POST. No HEAD:
> > >
> > > 68.5.28.172 - - [13/Mar/2002:13:23:12 -0500] "GET /StyleSheet.css
> > > HTTP/1.1" 200 10928
> > >
> > >
> > > I was expecting HEAD for StyleSheet.css since it is just a static
> > > file being served by UnknownFileTypeServlet.
> > >
> > > A grep for HEAD reveals no more HEAD requests after March 5th. And
> > > the few HEADs that occurred were for anything but the StyleSheet.
> > > This adds 10K per request and will slow down my app server.
> >
> > Okay, the deal is that the GET from modern browsers is accompanied by
> > an If-Modified-Since header (so they don't bother with HEAD). I
> > checked this out using Mozilla, Apache and a static .html file served
> > directly. The requests are always GET, but the subsequent responses
> > return 0 bytes of document.
> >
> > Having dumped the attrs of HTTPRequest, I don't see that it has
> > access to that information, or even the request headers in general. I
> > also don't find any source code in webware referencing "modified" and
> > "since" in the same line.
> >
> > I presume that we can somehow get this out of mod_webkit, but I don't
> > know about CGI.
> >
> > I'll investigate further tonight.
> 
> Bottom line: Most browsers use a "conditional GET" such as a GET with 
> an If-Modified-Since: header as opposed to the old "HEAD and GET" 
> technique. However on Apache, the If-Modified-Since HTTP header does 
> not get passed to WebKit (via mod_webkit or CGI), so for static files 
> such as a style sheet, WebKit sends the full content every time.
> 
> One workaround would be to put your static files outside your webkit 
> location. For my special case of a style sheet, I may just may use a 
> symbolic link that points to its current location.
> 
> 
> Here is some of the research I did (mostly via Google):
> 
> - I found that some HTTP servers will offer an HTTP_IF_MODIFIED_SINCE 
> to their CGI scripts, but Apache does not.
> 
> - Searching for HTTP_IF_MODIFIED_SINCE hit a lot of PHP sites that were 
> dumping globals, but when I visited the pages (and reloaded) 
> HTTP_IF_MODIFIED_SINCE was missing. Perhaps it was available in an 
> earlier version of either Apache or PHP.
> 
> - The only examples of being able to hook into If-Modified-Since were 
> things (such as mod_perl) that lived in Apache. But the hook I saw 
> wasn't to get the value, but call a function with higher level 
> semantics.
> 
> - It might be possible to tweak Apache to include 
> HTTP_IF_MODIFIED_SINCE, but I'm suspicious that it's not already done 
> or at least available as a patch. There might be some special treatment 
> of this header by Apache that make it infeasible.
> 
> - Apache 2.0?
> 
> 
> 
> -Chuck
> 
> _______________________________________________
> Webware-discuss mailing list
> Webware-discuss@...
> https://lists.sourceforge.net/lists/listinfo/webware-discuss

Errrr...

Are we sure about this?  How do we indicate that UserKit is only half done
(or whatever)?

Maybe we should add a status page to the Webware documentation that lists
the status of each kit in general terms, and then links to a detailed status
page in each Kit?

Jay

> -----Original Message-----
> From: Geoffrey Talvola [mailto:gtalvola@...]
> Sent: Thursday, March 14, 2002 4:07 PM
> To: 'webware-devel@...'
> Subject: [Webware-devel] Plan for 0.7b2 version numbering
> 
> 
> It sounds like nobody objected to giving all subpackages 
> inside of Webware
> the same version number.  So here's my plan:
> 
> - Add Docs/RelNotes-0.7.html to all subpackages that don't 
> already have one.
> The only item in the release note is a note on the new 
> version numbering
> policy.
> - Try to find all version numbers embedded in docs, source 
> files, etc. and
> replace with 0.7b2.  Each subpackage will still have its own 
> Properties.py
> with a "status" rating that indicates the level of maturity.
> - Update the overall release notes to reflect the new version 
> numbering
> policy.
> - Cut the 0.7b2 release.
> 
> I'm hoping this makes future releases easier.
> 
> If you have objections, let me know now because I want to 
> make these changes
> soon.
> 
> 
> - Geoff
> 
> _______________________________________________
> Webware-devel mailing list
> Webware-devel@...
> https://lists.sourceforge.net/lists/listinfo/webware-devel
> 


----------------------------------------------------------------------------

This e-mail and any attachments may be confidential or legally privileged.
If you received this message in error or are not the intended recipient, you
should destroy the e-mail message and any attachments or copies, and you are
prohibited from retaining, distributing, disclosing or using any information
contained herein.  Please inform us of the erroneous delivery by return
e-mail. 

Thank you for your cooperation.

----------------------------------------------------------------------------

It sounds like nobody objected to giving all subpackages inside of Webware
the same version number.  So here's my plan:

- Add Docs/RelNotes-0.7.html to all subpackages that don't already have one.
The only item in the release note is a note on the new version numbering
policy.
- Try to find all version numbers embedded in docs, source files, etc. and
replace with 0.7b2.  Each subpackage will still have its own Properties.py
with a "status" rating that indicates the level of maturity.
- Update the overall release notes to reflect the new version numbering
policy.
- Cut the 0.7b2 release.

I'm hoping this makes future releases easier.

If you have objections, let me know now because I want to make these changes
soon.


- Geoff

PSP currently converts expression blocks like <%=1234%> into servlet code
like this:

	res.write(str(1234))

It would be convenient to be able to use a custom function instead of "str".
For example, if I wanted all integers to be formatted like "1,234" I could
do that in a custom function.

There are a few ways to do this.  One way is to have PSP use some predefined
method call instead of str():

	res.write(self.htForValue(1234))

and in the base class provide:

	def htForValue(self, value):
		return str(value)

as a default implementation.  You could override it in the PSP itself using
a method directive, or in your own base class.  (Any other suggestions for
the name of the method?)

Another thought is a special PSP directive that lets you specify the string
conversion function or method (which would still default to str()):

	<%@ page formatter="self.htForValue" %>

And a third thought is that you don't actually need to do _anything_ to PSP
to get this to work.  All you need to do is assign to str at the top of your
PSP:

	<%str = self.htForValue%>

but that seems a bit too sneaky and could break if PSP's implementation
changed to not use str() in the future.

Thoughts?


- Geoff

I like this idea in general.

I don't like reassigning the str function.

So I guess the best approach is one of the other two.  The key here is not
to slow things down.  This function is called ALOT, and can have a dramatic
effect on performance.  I want to avoid adding any extra overhead.  I guess
calling str() invloves one dictionary lookup.  Keep this in mind as you
ponder the best way to do this.  We don't want to go too far in this
direction.  The default should add no (or almost no) additional overhead.
The user should know that if they choose to customize this, it might slow
things down.

As far as which approach is best, out of the two options you proposed, I
think they would both work the same way.  If we add a directive, it would
just do some assignment in the __init__ method anyway.

Jay

> -----Original Message-----
> From: Geoffrey Talvola [mailto:gtalvola@...]
> Sent: Wednesday, March 13, 2002 11:12 AM
> To: Webware_Devel (E-mail)
> Subject: [Webware-devel] Hook in PSP for converting to string?
> 
> 
> PSP currently converts expression blocks like <%=1234%> into 
> servlet code
> like this:
> 
> 	res.write(str(1234))
> 
> It would be convenient to be able to use a custom function 
> instead of "str".
> For example, if I wanted all integers to be formatted like 
> "1,234" I could
> do that in a custom function.
> 
> There are a few ways to do this.  One way is to have PSP use 
> some predefined
> method call instead of str():
> 
> 	res.write(self.htForValue(1234))
> 
> and in the base class provide:
> 
> 	def htForValue(self, value):
> 		return str(value)
> 
> as a default implementation.  You could override it in the 
> PSP itself using
> a method directive, or in your own base class.  (Any other 
> suggestions for
> the name of the method?)
> 
> Another thought is a special PSP directive that lets you 
> specify the string
> conversion function or method (which would still default to str()):
> 
> 	<%@ page formatter="self.htForValue" %>
> 
> And a third thought is that you don't actually need to do 
> _anything_ to PSP
> to get this to work.  All you need to do is assign to str at 
> the top of your
> PSP:
> 
> 	<%str = self.htForValue%>
> 
> but that seems a bit too sneaky and could break if PSP's 
> implementation
> changed to not use str() in the future.
> 
> Thoughts?
> 
> 
> - Geoff
> 
> _______________________________________________
> Webware-devel mailing list
> Webware-devel@...
> https://lists.sourceforge.net/lists/listinfo/webware-devel
> 


----------------------------------------------------------------------------

This e-mail and any attachments may be confidential or legally privileged.
If you received this message in error or are not the intended recipient, you
should destroy the e-mail message and any attachments or copies, and you are
prohibited from retaining, distributing, disclosing or using any information
contained herein.  Please inform us of the erroneous delivery by return
e-mail. 

Thank you for your cooperation.

----------------------------------------------------------------------------

On Friday 08 March 2002 02:00 pm, Geoffrey Talvola wrote:
> Opinions sought on the following:
>
> - SessionMemoryStore checks if the appserver is persistent, and if
> not, it doesn't pickle the session to/from disk.  This seems to serve
> no purpose other than to make OneShot break when you try to use it
> with SessionMemoryStore.  Should I just remove the check in
> SessionMemoryStore for a persistent appserver, and always have it
> pickle to disk?

Up to you, but if you make the change, please test OneShot with 
DynamicSessionStore as well (since I believe it also uses Memory).


> - People in the past have asked for a built-in HTTP server.  Assuming
> that he agrees to it, should we put Ng Pheng Siong's HttpAppServer
> into an Experimental directory in the Webware 0.7 release, with the
> appropriate disclaimers (it's not well tested and it's not guaranteed
> to stay there in future releases)?  Has anyone tested this?

I tried using it for an app and had the notorious problem of not being 
able to connect to it from outside my LAN. I couldn't figure it out and 
ended up just installing another instance of Apache. I think we had 
similar problems with the original all-in-one server. I'll be curious 
to see if Jay has managed to avoid this.


> - Any other _SMALL_ changes or fixes you'd like to see in Webware
> 0.7?  I want to cut an 0.7 beta 2 release next week incorporating a
> few more small fixes.

Nope.  :-)



-Chuck

On Tuesday 12 March 2002 07:02 am, Geoffrey Talvola wrote:
> What _I_ find confusing is that every component of Webware has its
> own independent release number, while Webware as a whole _also_ has a
> release number.  It makes cutting a release much more difficult (and
> obviously despite my efforts I didn't catch some of the release
> numbers when I cut 0.7b1), and just confuses people.  I honestly
> can't remember any time when someone on the mailing list actually
> referred to, for example, the MiddleKit release number.
>
> Why not axe all of the separate release numbers and just give the
> entirety of Webware one release number?
>
> Chuck?

Yeah, I find that to be a headache as well. But one concern is that the 
quality level between components can vary (alpha, beta, production), so 
I'd still like to retain some kind of quality indicator that would show 
up in the component index.

Also, I'd still like to have release notes for the components. How do 
we match those up? Would a component in its second revision just get a 
RelNotes-0.8 for example? Would the version of the component package be 
0.8 and 0.2?

Perhaps most of the problems would go away if we either didn't refer to 
the version numbers in the non-release notes docs (such as User's 
Guides), or those were updated upon installation. That's actually where 
I have the most trouble.

-Chuck