June 1999 - Zope - Zope lists

Re: [Zope] Sessions needed
by michel＠digicool.com 28 Jun '99

28 Jun '99

Frank Buhr <buhr(a)kogs1.informatik.uni-hamburg.de> writes: > Hi, > > after doing some experiments with Zope I'm still missing an important feature: > > How can I easily keep state information over several HTTP-Requests? > > The abstraction that provides this is called "session" for instance in the Java Servlet API > http://java.sun.com/products/servlet/2.1/html/api-reference.fm.html#945 > JS-sessions provide an abstraction from cookie-mechanisms or URL encoding. > > Netscapes server side javascript provides similar things. There are some > choices for the implementation of sessions: keeping state information > on the server (the client only references an ID) or on the client > (all values are stored on the client), using cookies, URL encoding or > hidden form elements. > > I found some Classes in the Zope sources where there is already state information > maintained by cookies (TreeTag and ZRDB.DA) > This seems to be equivalent to the combination "data on the client, use cookies" from above. > > I very much like the idea of separating the service of "state keeping" from > the concrete implementation (which might even be chosen dynamically according > to browser capabilities or user preferences ("no cookies -> use URL encoding") > > Did I overlook something or is there really no session _abstraction_ in Zope? You are correct, there is no low level session abstraction in Zope. This problem is much thornier than it sounds. Attempting to have a both useable and flexable session toolkit has show to us to be a difficult concept that requires a lot of thought. What are generally called 'sessions' is actually a smaller piece of a much larger puzzle we call 'membership'. More than just a private namespace for a series of HTTP transactions, membership is more of a concept of 'meta-sessions', keeping state between a series of sessions. There are, actually, only a small number of scenarios where construction and destruction of 'sessions' is useful unless the information on the user of that sessions is kept for later sessions. Everyone and their brother has a gazzillion custom requirements they need satified in this very vague (and ill-defined by the JS standard) area we call membership. The actualy 'session' part (which we use a cookie to solve) is one of the easiest pieces of the puzzle. Although many other toolkits have the ability to create an abstraction they call 'session' it is hard to find one that handles membership in a usable and flexable fasion (rumor has it WebObjects has a rather nice model in this regard, of course you have to drop some TALL cash to get it). We want our system to be able to satisfy most requirements, and be able to be molded without breaking any of it to fit all the rest. For example, membership should be flexible enough to store membership information in an SQL database, but it should not be required. So, that aside, I can tell you that this is a deep problem we are thinking about actively here at DC in various top secret projects, and we plan on solving in the usual ultra cool Zope Zen way. > > (At first sight the feature now called "versions" seemed to do something similar > but this was only the first sight.) > > A Session (in my definition) should provide a private namespace for exactly one user > (or web client). It should have an optional timeout in order to destroy it if the > user is inactive. > Ah there lies the rub. 'exactly one user' is not a concept currently in Zope. It is possible for me to create users called 'michel' at various and disjoined areas of Zope. This is a wildly usable feature of Zope. Creating 'members' is a higher level concept than this, members should be singular, there should be no two 'Michel' members in the same 'membership'. Members should have an area that is 'theirs' (like the UNIX concept of a home directory). etc. etc. It gets sticky and easy to paint oneself in a corner. > Has anyone build a nice Zope product which does this? > I believe there is a product somewhere out there that lets you manipulate cookies in the browser to store some small amount of state (something like 4K) on the client. I believe it was called CookieCutter. This could be useful for your short term goals. Perhaps the author could chime in? > May be sessions could be integrated into the object database > (by declaring objects to have a special scope and extent)? We generally avoid putting HTTP specific (or even Zope specific) feature in the database, this grossly violates our componentized model. The ZODB should (and is) be useful for a variety of applications completely unrelated to Zope or the web. In either case, the database is pretty transparent and low level at Zope's python API layer, so the problem could be easily solved without bringing any thought about the database into it. -Michel > > > > Frank > > _______________________________________________ > Zope maillist - Zope(a)zope.org > http://www.zope.org/mailman/listinfo/zope > > (For developer-specific issues, use the companion list, > zope-dev(a)zope.org - http://www.zope.org/mailman/listinfo/zope-dev )

1 0

Sessions needed
by Frank Buhr 28 Jun '99

28 Jun '99

Hi, after doing some experiments with Zope I'm still missing an important feature: How can I easily keep state information over several HTTP-Requests? The abstraction that provides this is called "session" for instance in the Java Servlet API http://java.sun.com/products/servlet/2.1/html/api-reference.fm.html#945 JS-sessions provide an abstraction from cookie-mechanisms or URL encoding. Netscapes server side javascript provides similar things. There are some choices for the implementation of sessions: keeping state information on the server (the client only references an ID) or on the client (all values are stored on the client), using cookies, URL encoding or hidden form elements. I found some Classes in the Zope sources where there is already state information maintained by cookies (TreeTag and ZRDB.DA) This seems to be equivalent to the combination "data on the client, use cookies" from above. I very much like the idea of separating the service of "state keeping" from the concrete implementation (which might even be chosen dynamically according to browser capabilities or user preferences ("no cookies -> use URL encoding") Did I overlook something or is there really no session _abstraction_ in Zope? (At first sight the feature now called "versions" seemed to do something similar but this was only the first sight.) A Session (in my definition) should provide a private namespace for exactly one user (or web client). It should have an optional timeout in order to destroy it if the user is inactive. Has anyone build a nice Zope product which does this? May be sessions could be integrated into the object database (by declaring objects to have a special scope and extent)? Frank

1 0

Re: [Zope] ANN: ZCatalog 1.0.0 alpha
by michel＠digicool.com 28 Jun '99

28 Jun '99

Paul Prescod <paul(a)prescod.net> writes: > Michel Pelletier wrote: > > > > Greetings Zopeheads. > > > > After much sweat and toil ..... > > > ... We would, however, like the tired to be well > > kicked. > > Well then get over here! How am I going to kick you over the > Internet? You just did. ;) > > > We recommend you read the ZCatalog Tutorial, written by Mike Pelletier > > (Note that 'Mike' Pelletier is a completely different person than > > 'Michel' Pelletier, which is me) which is available on the web at the > > above location. > > Maybe one of you could change your name. I've heard that the name "Bobo" > is available. > Nope. We allready have a Bobo 'Conner in the community. -Michel

1 0

Re: [Zope] Re: ZCatalog alpha
by michel＠digicool.com 28 Jun '99

28 Jun '99

Alex Rice <alrice(a)swcp.com> writes: > On 26 Jun 1999 15:20:57 -0400, > michel(a)digicool.com said: > > michel> The several MB of text would probably translate into 2 or 3 > michel> times that in memory growth. However, the Persistence machinery > michel> will quickly start to deactivate unused portions of the index. > michel> Try tweaking your Cache settings to see if you can get Zope to > michel> be more aggresive about deactivating objects. > > OK, here's some more data points, just in case it's useful: > > My z2.py process starts at about 5-6MB memory usage. After indexing > about 2MB of text, it goes up to 40 MB memory usage. Manually running > the GC, it goes down to 38MB. (manage_pack and manage_cache_size are not > implemented, yet, I think) Stopping, then starting Zope it goes back > down to 5-6MB. After searching via the catalog it's at about 8MB and > stays right around there. > Ok, this is pretty unusual, good clue that it only grows your process size by 2MB after a restart, I suspect then that there is some leak in the 'find and catalog' code. > Thanks for ZCatalog. It's going to be immensely useful. Yes, it fills an aching gap in Zope's Open Source capabilities. -Michel > > Alex Rice | alrice(a)swcp.com | http://www.swcp.com/~alrice > Current Location: N. Rio Grande Bioregion, Southwestern USA > > > > > > _______________________________________________ > Zope maillist - Zope(a)zope.org > http://www.zope.org/mailman/listinfo/zope > > (For developer-specific issues, use the companion list, > zope-dev(a)zope.org - http://www.zope.org/mailman/listinfo/zope-dev )

1 0

Re: [Zope] Startup error Zope 2.0.0a3: Error 139
by michel＠digicool.com 28 Jun '99

28 Jun '99

Alexander Staubo <alex(a)mop.no> writes: > I really, really hate to be a pest, but could somebody at DC please look > this way and at least make a small attempt at helping me out? > > Python 1.5.2 and Zope both compiled with gcc 2.8.1; relevant? > Were there any errors during the compile? A core dump usualy means that something went wrong in one of the C Extensions that caused a low level exception, like accessing a NULL pointer. It would be nice to find out exactly where in the code this is happening. First, I'd like to eliminate ZServer from the equation, could you go to the lib/python directory and run python1.5.2, then: >>> import Zope, ZPublisher >>> ZPublisher.Main('/') This will simulate a request for '/' (the root folder's index_html method). If this doesn't fail, then the problem is in ZServer (which I find unlikely, there being no C in ZServer). In the case of failure, it will either happen after the import or after the call to Main(). If the former, let us know that, if the latter, load python and do the import and then try: >>> ZPublisher.Main('/', d=1) This will put you in the debugger (which you might want to read about a bit on python.org if you don't know how to use it) using combinations of (s)tep and (n)ext, try and narrow down right to the line of code where the core dump is happening. -Michel > -- > Alexander Staubo http://www.mop.no/~alex/ > "`This must be Thursday,' said Arthur to himself, sinking low over > his beer, `I never could get the hang of Thursdays.'" > --Douglas Adams, _The Hitchhiker's Guide to the Galaxy_ > > >-----Original Message----- > >From: Alexander Staubo [mailto:alex@mop.no] > >Sent: 25. juni 1999 16:07 > >To: Zope Mailing List (E-mail) > >Subject: RE: [Zope] Startup error Zope 2.0.0a3: Error 139 > > > > > >I hear rumours that error code 139 indicates memory corruption. Also I > >just noticed the crash below has Python producing a core dump, anybody > >want to look at it? > > > >-- > >Alexander Staubo http://www.mop.no/~alex/ > >"What the hell, he thought, you're only young once, and threw > >himself out of the window. That would at least keep the element of > >surprise on his side." > >--Douglas Adams, _The Hitchhiker's Guide to the Galaxy_ > > > >>-----Original Message----- > >>From: Alexander Staubo [mailto:alex@mop.no] > >>Sent: 25. juni 1999 15:43 > >>To: Zope Mailing List (E-mail) > >>Subject: [Zope] Startup error Zope 2.0.0a3: Error 139 > >> > >> > >>I just installed Zope 2.0.0a3 (no CVS patches) on a Slackware > >Linux 2.2 > >>(libc6, Python 1.5.2) box. Of course, it refuses to start: > >> > >>firewall:/usr/local/zope-2.0.0a3# python z2.py -d 192.168.1.1 > >>-w 8080 -f > >>2121 > >>zdeamon: Fri Jun 25 15:32:33 1999: Houston, we have forked > >>zdeamon: Fri Jun 25 15:32:33 1999: Houston, we have forked > >>zdeamon: Fri Jun 25 15:32:33 1999: Hi, I just forked off a kid: 24905 > >>zdeamon: Fri Jun 25 15:32:35 1999: Aiieee! 24905 exited with > >>error code: > >>139 > >> > >>Curiously, no tracebacks? > >> > >>-- > >>Alexander Staubo http://www.mop.no/~alex/ > >>"What the hell, he thought, you're only young once, and threw > >>himself out of the window. That would at least keep the element of > >>surprise on his side." > >>--Douglas Adams, _The Hitchhiker's Guide to the Galaxy_ > >> > >> > >>_______________________________________________ > >>Zope maillist - Zope(a)zope.org > >>http://www.zope.org/mailman/listinfo/zope > >> > >>(For developer-specific issues, use the companion list, > >>zope-dev(a)zope.org - http://www.zope.org/mailman/listinfo/zope-dev ) > >> > > > > > >_______________________________________________ > >Zope maillist - Zope(a)zope.org > >http://www.zope.org/mailman/listinfo/zope > > > >(For developer-specific issues, use the companion list, > >zope-dev(a)zope.org - http://www.zope.org/mailman/listinfo/zope-dev ) > > > > > _______________________________________________ > Zope maillist - Zope(a)zope.org > http://www.zope.org/mailman/listinfo/zope > > (For developer-specific issues, use the companion list, > zope-dev(a)zope.org - http://www.zope.org/mailman/listinfo/zope-dev )

1 0

RE: [Zope-dev] Z SQl Method (Newbie)
by Alexander Staubo 28 Jun '99

28 Jun '99

But does Gadfly work in 2.0? I know for a fact it doesn't work in 1.11.0pr1, and the error you cite is the exact same error I get using the latter version. I asked about this earlier, but I'm still waiting for an answer. -- Alexander Staubo http://www.mop.no/~alex/ "`This must be Thursday,' said Arthur to himself, sinking low over his beer, `I never could get the hang of Thursdays.'" --Douglas Adams, _The Hitchhiker's Guide to the Galaxy_ >-----Original Message----- >From: zope-dev-admin(a)zope.org [mailto:zope-dev-admin@zope.org]On Behalf >Of Paul Chung Chee Soong >Sent: 28. juni 1999 02:09 >To: mj(a)antraciet.nl; zope-dev(a)zope.org >Subject: RE: [Zope-dev] Z SQl Method (Newbie) > > >Thanks pal. Anyway, I still failed to create a table. My SQL stmt is >"CREATE TABLE policy (id varchar, status varchar)" and the >error I get is > > >Zope Error >Zope has encountered an error while publishing this resource. >exceptions.AttributeError >Sorry, a Zope error occurred. >Traceback (innermost last): > File >/usr/local/src/Zope-2.0.0a1-src/lib/python/ZPublisher/Publish.py, >line 255, in publish_module > File >/usr/local/src/Zope-2.0.0a1-src/lib/python/ZPublisher/Publish.py, >line 165, in publish > File >/usr/local/src/Zope-2.0.0a1-src/lib/python/BoboPOS/TJar.py, line >281, in commit > File >/usr/local/src/Zope-2.0.0a1-src/lib/python/BoboPOS/TJar.py, line >200, in maybe_commit_objects >AttributeError: _p_changed > >Pls help. T.Q. > > > -----Original Message----- >From: Martijn Pieters [SMTP:mj@antraciet.nl] >Sent: Friday, June 25, 1999 9:53 AM >To: Paul Chung Chee Soong; zope-dev >Subject: Re: [Zope-dev] Z SQl Method (Newbie) > >At 11:53 25/06/99 , Paul Chung Chee Soong wrote: >>Create a Z Gadfly Database Connection. I created a directory 'test' in >>Zope-2.0.0a1-src/var/gadfly and it appeared in the Data >Source field. >How >>can I access it thru SQL Method since it isn't a database at >all (just a >>plain dir !!). Can I create a database or table via SQL Method? >> >>Do I need to link the data source with a gadfly database? If >yes, how?? >> >>By default, there's a demo data source. How can I use it????? How do I >>know what are the tables in it anyway?? Thus, I can't use the SQL >Method. >> >>I would appreciate very much if you can reply as soon as >possible. I >need >>to use SQL Method desperately now. > >No tables have been defined yet. You can create tables using a SQL >method, >using standard Data Definition Language syntax: > >CREATE TABLE <tablename> (<column definition> [, <column definition>]*) > >x>"' > > > >_______________________________________________ >Zope-Dev maillist - Zope-Dev(a)zope.org >http://www.zope.org/mailman/listinfo/zope-dev > >(For non-developer, user-level issues, use the companion list, >zope(a)zope.org, http://www.zope.org/mailman/listinfo/zope ) >

1 0

RE: [Zope] RFC: Multiple databases
by Alexander Staubo 28 Jun '99

28 Jun '99

Unless you invent some special magic, you'd lose transactions, too, wouldn't you? Or is that built into the ODB engine? (I don't know exactly how this thing is built.) Under BFS (BeOS) and NTFS 5 (Windows 2000) you don't even have to store properties externally, as these file systems support file attributes. On NTFS 4 (NT 4.0 and earlier) you could store the metadata as separate streams inside each file; this is a barely documented, largely overlooked feature of NT that is rarely used, and is in fact there for Mac resource-fork compatibility. But it would certainly work. I think it's a good idea, if only because FTP and WebDAV (at least as far OS integration is concerned) suck. -- Alexander Staubo http://www.mop.no/~alex/ "`This must be Thursday,' said Arthur to himself, sinking low over his beer, `I never could get the hang of Thursdays.'" --Douglas Adams, _The Hitchhiker's Guide to the Galaxy_ >-----Original Message----- >From: zope-admin(a)zope.org [mailto:zope-admin@zope.org]On Behalf Of >Anthony Pfrunder >Sent: 28. juni 1999 02:29 >To: zope(a)zope.org >Subject: [Zope] RFC: Multiple databases > > >[The following is also in the Collecter under features. I'm >posting this >as it relates to the above messages.] > >My thoughts on extensions to Zope: > >I'm <will be> working on constructing an additional Data.fs plugin >(dropover FileStorage.py) which may be of use. The idea is to >store stuff >on the native filesystem and put properties etc in prop.x (ie the >properties for file1.txt are in prop.file1.txt). The >advantage of this is >that it allows ppl to have a smb connection to a shared directory hence >allowing them to more easily update the files. Of course, >this means that >you lose the ability to undo changes. > >This database could be enhanced by allowing it to reconize Microsoft >Office documents and store properties directly in the document itself. >This means that the properties now travel with the document. > >The second part of this proposal is to produce a Product which >allows you >to "mount" a database onto a subpath. This is very similiar >to importing >the database except that they are kept separate. These means that >toplevel stuff goes into one Data.fs and the department >Data.fs contains >/Department/... stuff. This way, you can have a consistent >style across >various webservers just by running multiple copies of Zserver. In >addition, you could make toplevel.fs read-only to protect it. This >Product should be able to handle different filesystems (ie the current >one, the above one and others) > > >Cheers, > >Anthony Pfrunder > > >_______________________________________________ >Zope maillist - Zope(a)zope.org >http://www.zope.org/mailman/listinfo/zope > >(For developer-specific issues, use the companion list, >zope-dev(a)zope.org - http://www.zope.org/mailman/listinfo/zope-dev ) >

1 0

RFC: Multiple databases
by Anthony Pfrunder 28 Jun '99

28 Jun '99

[The following is also in the Collecter under features. I'm posting this as it relates to the above messages.] My thoughts on extensions to Zope: I'm <will be> working on constructing an additional Data.fs plugin (dropover FileStorage.py) which may be of use. The idea is to store stuff on the native filesystem and put properties etc in prop.x (ie the properties for file1.txt are in prop.file1.txt). The advantage of this is that it allows ppl to have a smb connection to a shared directory hence allowing them to more easily update the files. Of course, this means that you lose the ability to undo changes. This database could be enhanced by allowing it to reconize Microsoft Office documents and store properties directly in the document itself. This means that the properties now travel with the document. The second part of this proposal is to produce a Product which allows you to "mount" a database onto a subpath. This is very similiar to importing the database except that they are kept separate. These means that toplevel stuff goes into one Data.fs and the department Data.fs contains /Department/... stuff. This way, you can have a consistent style across various webservers just by running multiple copies of Zserver. In addition, you could make toplevel.fs read-only to protect it. This Product should be able to handle different filesystems (ie the current one, the above one and others) Cheers, Anthony Pfrunder

1 0

ZClasses problem
by Alex Rice 28 Jun '99

28 Jun '99

I'm trying to create some ZClasses via an external method, but I'm running into a problem: I can't getattr() the product's methods. For instance, I have a Product called Etexts, containing a zclass etext a dtml method etext_add (etext constructor) a dtml method etext_addForm (etext constructor input form) a factory etext_factory (etext factory) I can go anywhere in my Zope heirarchy, and in an external method, I can getattr() to manage_addProduct/Etext, but when I try to getattr() to manage_addProduct/Etext/etext, etext_add, or etext_addForm, I get an AttributeError. I don't have this problem within Zope-- for instance in my browser I can successfully call and create a ZClass with http://some_long_path/manage_addProduct/Etexts/etext?id=spam What am I doing wrong to not be able to get to the same method via an external method? Is this the correct way to programatically add ZClasses? As usual, I'm kind of emulating what goes on in the Zope user-interface. TIA Alex Rice | alrice(a)swcp.com | http://www.swcp.com/~alrice Current Location: N. Rio Grande Bioregion, Southwestern USA

1 0

RE: [Zope] Tree tag and SQL methods?
by Alexander Staubo 27 Jun '99

27 Jun '99

Here you go:      -- Alexander Staubo http://www.mop.no/~alex/ "`This must be Thursday,' said Arthur to himself, sinking low over his beer, `I never could get the hang of Thursdays.'" --Douglas Adams, _The Hitchhiker's Guide to the Galaxy_ >-----Original Message----- >From: Phillip J. Eby [mailto:pje@telecommunity.com] >Sent: 28. juni 1999 00:37 >To: Alexander Staubo; Zope Mailing List (E-mail) >Subject: RE: [Zope] Tree tag and SQL methods? > > >Can you post the source of your tree tag call as well? Thanks. > >At 07:21 PM 6/27/99 +0200, Alexander Staubo wrote: >>No offense taken -- yes, of course I have tested it. :-) >> >>I have not tested using the #tree tag with the second output >though, but >>that would meaningless (the second output should be rendered as leaf >>nodes). I even tried having the second query output _all_ data in the >>table, to no avail. Here's my query: >> >> >>select distinct publication, count(publication) as pub_cnt >>from articles >>where publication != '' >>group by publication >>order by publication >> >>select title >>from articles >>where publication =  >> >> >>Here, pb is the publication argument. I assure you, the second part of >>the query does execute correctly. >> >>-- >>Alexander Staubo http://www.mop.no/~alex/ >>"`This must be Thursday,' said Arthur to himself, sinking low over >>his beer, `I never could get the hang of Thursdays.'" >>--Douglas Adams, _The Hitchhiker's Guide to the Galaxy_ >> >>>-----Original Message----- >>>From: Phillip J. Eby [mailto:pje@telecommunity.com] >>>Sent: 27. juni 1999 18:07 >>>To: Alexander Staubo; Zope Mailing List (E-mail) >>>Subject: RE: [Zope] Tree tag and SQL methods? >>> >>> >>>Don't be offended by this suggestion, but, have you tried >>>having the tree >>>print out all the variables that would affect the 'second' >>>query, and then >>>plugging them into manage_test of the query? My guess is >>>you'll find that >>>there's something you've overlooked as far as how output of >>>the first query >>>gets fed into the second, or else the second is just plain wrong. >>> >>>At 03:36 PM 6/26/99 +0200, Alexander Staubo wrote: >>>>Doesn't improve anything: While the "second" part of the >query -- the >>>>one that's supposed to collect the leaf articles -- clearly >>>executes, no >>>>data is displayed. Only the first level -- the publications >list -- is >>>>rendered, with "plus" icons. Clicking on the icon runs the >>>query but no >>>>data is there. >>>> >>> >> >> >>_______________________________________________ >>Zope maillist - Zope(a)zope.org >>http://www.zope.org/mailman/listinfo/zope >> >>(For developer-specific issues, use the companion list, >>zope-dev(a)zope.org - http://www.zope.org/mailman/listinfo/zope-dev ) >> >

2 1