- Zope-Dev - Zope lists

Trapping Interbase Error Codes
by Beverly Dolor 09 Mar '02

09 Mar '02

Hi! I'm currently doing a program on Zope 2.5 using gvibDa as my database. I would like my program to trap Interbase error codes such as (isc_deadlock, isc_lock_conflict), the problem is I'm still at a lost on how to do this. I saw some progam written in C that trap these kinds of Interbase error using isc_print_sqlerror and isc_sql_interprete, in fact I saw these functions being used on gvibBase.c :) But since I don't know C, the programs didn't help much in finding out how the errors are trapped. I know gvibDa has its own exception handler (gvibExceptions?), but this handler converts the original database error into its own error message (DatabaseError,InternalError, etc). what I want to do is to trap the "raw" database error before it gets converted into one of the customized error messages. And another question, is there a way to commit/rollback a database transaction(using a gvib object) and zope transaction without using an external method? Any answer, opinion, view regarding these issues will be very much appreciated. Thanks :) __________________________________________________ Do You Yahoo!? Try FREE Yahoo! Mail - the world's greatest free email! http://mail.yahoo.com/

3 3

Manual object-to-relational persistency framework
by Bjorn Stabell 09 Mar '02

09 Mar '02

Hello, I'm trying to use Zope without storing everything in ZODB while still having the perception of an object-oriented database. I want to control the object-to-relational mapping layer, but would like the loading and saving of objects to be automatic. Is there some way to do that using Zope? Where would I, e.g., hook into to get automatic loading/cacheing/saving of objects that weren't stored in ZODB, and where I could actually create the methods to load/save objects myself (I want to control the data structure). Regards, -- Bjorn

5 4

zope transaction bug?
by Ross J. Reedstrom 09 Mar '02

09 Mar '02

Hey all - Boy, it's been a long time since I've hung around here. Anyway, I think we've found a bug somewhere in the zope transaction machinery. Every now and then, a backend will get into 'idle in transaction' state. After that, no work on that backend ever commits - the 'END' query just never gets sent. Careful analysis of the log files indicated that the ZpsycopgDA instance started 'losing' proper transactional control after an admin who was logged in via an exUserFolder that _also_ used that DA for authentication used the management interface to undo a Zope transaction. The transaction in question was merely an edit on a DTML file. No db interaction beyond the authentication. Either a successful or an unsuccessful attempt to rollback causes the problem. Seems the Zope transactional machinery has 'forgotten' that our connection needs to be told about transaction boundries. We can make this happen at will, with Zope 2.4.3 or 2.5.0, exUserFolder 0.10.7, and ZpsycopgDA 1.0.5. It does require using the usAuth (User Supplied Auth source) part of exUserFolder. Our developer who's somewhat familiar with the guts of exUserFolder thinks that this is probably still a bug in Zope, since all usAuth does is use customized ZSQL methods to change where various strings are stored in the DB. We're seeing similar problems with ZpopyDA, but haven't tested it thoroughly to see if it's the same trigger. I've constructed the test case for this: it's available at: http://wallace.ece.rice.edu/xact-bug.tar.gz That tarball contains an SQL schema script, zexp, and minimal instructions for replicating the problem. Anyone have any clues as to where to look to fix this? We're losing commits to our database, whenever a developer tries to Undo something: not good! Ross -- Ross Reedstrom, Ph.D. reedstrm(a)rice.edu Executive Director phone: 713-348-6166 Gulf Coast Consortium for Bioinformatics fax: 713-348-6182 Rice University MS-39 Houston, TX 77005

2 3

Shallow copies/links/references
by Mario Valente 08 Mar '02

08 Mar '02

Hi: OK everyone, I've managed to get object links to work. As discussed previously by others (re: object references) and asked by myself: I had the need to be able to refer to objects from different points of the folder hierarchy without duplicating those objects. Yes I know that can be done *programatically*. That is not the idea. The idea is to have a folder with differente widgets/components and to be able to copy and paste references to those widgets throughout the folder hierarchy. Of course that the usual copy/paste is not OK, since that *duplicates* the objects and as such further changes to the original object wont be reflected in the copies. With something like this + Ordered Folders you are able to have an index_html that just renders the objects within a folder and you can then drop objects/widgets into the folder to produce a page. What I did: - edited CopySupport.py - copied the manage_pasteObjects method to a manage_pasteMonikers method - commented the #ob=ob._getCopy(self) line (the duplicate object part) - added a "Paste Ref" button to the lib/python/OFS/dtml/main.dtml file that calls the pasteMonikers method """ <input class="form-element" type="submit" name="manage_pasteMonikers:method" value="Paste Ref." /> """ It works. It does what I want/need. When you change a reference, the original also changes. When you change the original, the references change. You can delete either the reference or the original and all is OK. Issues/TODO - find a way to distinguish references from the original (tried to change the meta_type at paste time but no luck). At least the meta_type/icon should change to provide a visual cue. - provide tool(s) to find impact of changing a reference. Provide a link from a reference object to the original. Provide a link from an object to its several references. Comments and experiences are welcome. C U! -- Mario Valente

2 2

TF & ZMI, was: RE: [Zope-dev] Zope 2.6 and OrderedFolders
by sean.upton＠uniontrib.com 08 Mar '02

08 Mar '02

You know, the funny thing is that TF really accomplishes the goals of a ZMI/UI-improvement hack (unclutter a view of a folder to VISUALLY separate the software/content/presentation space) implemented, IMHO, in the wrong place. I would think there is another way that a Folder UI could be uncluttered without the magic of TF, and that is add "translucency" (items are there, but obscured from view) to the UI layer - that is, the ability to VISUALLY unclutter different users' views of objects in a folder via ZMI; views could be stored per user. For example, you could select ZSQL Methods and RDB connections in a folder to be hidden from everybody but the Database programmer; you could keep these and Python Scripts obscured from the view of a content person or presentation designer, etc. Well, they wouldn't be hidden, but visually 'compacted' and relegated to a small drop-down box at the bottom of the page to unclutter it... so the standard tabled object list wouldn't iterate over objectIds() or objectValues(), but instead a higher-level filter function. A complimentary filter function would render the 'hidden' objects as a sequence as well. __ZMI manage_main Now:______________________ ________________________ ===\===\===\===\======== <--- ZMI tabs % :::::::::::::::::::=== TP.NAME..... .SZ. ..LM.. []---------- ---- ------ []---------- ---- ------ []---------- ---- ------ []---------- ---- ------ []---------- ---- ------ [RN][CT][CP][DL][I/E][SA] <--- Buttons for checked __ZMI With Translucency:___________________ ________________________ ===\===\===\===\======== <--- ZMI tabs % :::::::::::::::::::=== TP.NAME..... .SZ. ..LM.. []---------- ---- ------ []---------- ---- ------ []---------- ---- ------ []---------- ---- ------ []---------- ---- ------ [RN][CT][][][][][Hide] <--- New HideBtn ________________________ Hidden Items: [Select Item... [v] [Open][Unhide][ViewAll] <--- Buttons for doing things ^ with hidden | items | +-----------> ViewAll => pop-up new window ____________________________________________ This would be more visually convenient. I'm thinking of trying to monkey-patch OFS.Folder to just do this. Of course, with these kind of modifications, for any given folder, there's no mandatate that anything need to be hidden at all. Thoughts? Sean -----Original Message----- From: Casey Duncan [mailto:c.duncan@nlada.org] Sent: Wednesday, March 06, 2002 12:02 PM To: sean.upton(a)uniontrib.com; Eric.Roby(a)noaa.gov; zope-dev(a)zope.org Subject: Re: [Zope-dev] Zope 2.6 and OrderedFolders On Wednesday 06 March 2002 02:49 pm, sean.upton(a)uniontrib.com allegedly wrote: > Is this a suggestion that TF go into the core? One other (major) problem > with TF is that it breaks session tracking site-wide, or at least the > ability to have more than one session id manager per Zope instance in CST; > I'm not sure about sessions in Zope 2.5... > > TF seems to have too much normalcy-hijacking magic to go into the core, if > you ask me... > > Sean -1 on TF in the core. Nice hack, but too much magic and there are significant performance implications if I remember correctly. Also Zope 3 will render TF superfluous. /---------------------------------------------------\ Casey Duncan, Sr. Web Developer National Legal Aid and Defender Association c.duncan(a)nlada.org \---------------------------------------------------/

4 3

RE: [Zope-dev] More Zope 2.6 requests
by sean.upton＠uniontrib.com 08 Mar '02

08 Mar '02

+1 on #2 - definitely useful. Sean -----Original Message----- From: Jim Washington [mailto:jwashin@vt.edu] Sent: Friday, March 08, 2002 5:26 AM To: zope-dev(a)zope.org Subject: [Zope-dev] More Zope 2.6 requests While we are asking, I have two things that I would consider valuable additions to Zope 2.6: 1. UserSniffer. Currently an External Method, but has functionality that should be available OOTB to assist making those (horrors!) browser-dependent hacks. It has other uses, too, like explaining to clients that their browser is five years old and needs the free upgrade that is available. 2. Support for HTTP_X_FORWARDED_FOR. Many (most?) of us run zope behind Apache ProxyPass or Squid, and the Zope logs therefore save the ip address of the proxying machine rather than any semblance of where the client browser is. I realize that other proxies do not always follow the rules, etc., etc., but I think that using HTTP_X_FORWARDED_FOR if not null would be better than a log full of 192.168.1.1. Apologies if these have already been discussed under another heading. -- Jim Washington _______________________________________________ Zope-Dev maillist - Zope-Dev(a)zope.org http://lists.zope.org/mailman/listinfo/zope-dev ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope )

2 1

the segfault saga continues
by Leonardo Rochael Almeida 08 Mar '02

08 Mar '02

So now I have installed a brand new MySQL-python-2.9.2a2 explicitly compiled against reentrant mysql libraries (libmysqlclient_r, in Debian's libmysqlclient10-dev package). I also installed ZMySQLDA-2.0.9b1 which compounds MySQL-python reentrancy by serializing all mysql requests with a lock. We've also completely reformed the site so as to get rid of that huge number of pythonscripts (they were mostly duplicates and we manage to remove the redundancy using ZClasses). Segfaults are still with me, however, unless I disable the cycle-gc, of course. Tomorrow I'll try and get a core to see what gdb can tell me. Anything you guys suggest I take a look at? (besides the C stack trace which will probably be pointing somewhere inside the GC) I'll also try go get the remaining non recompiled python scripts fixed, so as to please Anthony Baxter :-) But for that I'll need one of these: * a way to make the 'needs recompiling' log message give me the path to the object. Since that message happens at __setstate__, getPhysicalPath() doesn't return anything useful (no acquisition wrappers). * a way to know whether a Python script needs recompiling or not. So that I can ZCatalog only the non recompiled python scripts. I tried asking the ZCatalog for all the python scripts and later recompiling them, but I managed to miss a lot of them somehow. It's as if the ZCatalog decided that it wasn't going to really catalog ALL python scripts. -- Ideas don't stay in some minds very long because they don't like solitary confinement.

2 2

RE: [Zope-dev] Manual object-to-relational persistency framework
by Bjorn Stabell 08 Mar '02

08 Mar '02

I think the Storage interface wroks at a lower level, dealing with pickled Python objects. I'd like to do it at a higher level, so I can define my own object-relational mapping, unique for each object type I want to use the Zope framework (acquisition, access control, automatic persistence) and still have the objects stored in a nice domain specific relational schema, not storing anything (even "ghost" objects) in the ZODB, but having "live" objects available as usual. Ideally, it should be possible to get the the data for the objects from anywhere. I guess it would be possible to make a special Storage interface that calls hooks in objects that have them to actually implement the storage/persistency functionality, but that seems kind of overkill and it'd mean I have to mount another Storage etc. There must be any easier way? -----Original Message----- From: t.foerster(a)biologie.hu-berlin.de [mailto:t.foerster@biologie.hu-berlin.de] Posted At: Friday, March 08, 2002 17:27 Posted To: Zope Developer Conversation: [Zope-dev] Manual object-to-relational persistency framework Subject: Re: [Zope-dev] Manual object-to-relational persistency framework On Friday, 8. March 2002 09:50, Bjorn Stabell wrote: > I'm trying to use Zope without storing everything in ZODB while still > having the perception of an object-oriented database. I want to > control the object-to-relational mapping layer, but would like the > loading and saving of objects to be automatic. Is there some way to > do that using Zope? > Where would I, e.g., hook into to get automatic > loading/cacheing/saving of objects that weren't stored in ZODB, and > where I could actually create the methods to load/save objects myself > (I want to control the data structure). Sounds like changing the (default) backend of the ZODB. That way you use Zope with the ZODB but your objects are stored in a RDBMS too. ZODB comes with several backends (a.k.a Storages), the default FileStorage, DemoStorage, OracleStorage. The latter doing already object relational mapping to Oracle RDBMS. If you want to control the mapping, just define your own Storage (I do not know how hard it is, maybe the 'gurus' do?) and then tweak your Zope instance to use your custom Storage instead of the default FileStorage. That way you do not have to care about automatic loading/saving of your objects. Regards, Thomas -- Webmaster Innovationskolleg Theoretische Biologie http://itb.biologie.hu-berlin.de _______________________________________________ Zope-Dev maillist - Zope-Dev(a)zope.org http://lists.zope.org/mailman/listinfo/zope-dev ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope )

2 1

WebDAV locks disappearing...
by Jerome Alet 07 Mar '02

07 Mar '02

Hi, I've just tested webDAV access for the very first time using cadaver, so maybe this is a known problem. I've locked some objects using cadaver's lock command, and then opened a browser keeping cadaver's connection opened. then I've searched for this locks using the ZMI and also a method of my own. All worked fine, the locks were found. Then some time later I've retried to find the locks, and both the ZMI and my method returned no lock. However in cadaver the locks still seemed to be there... Then I've unlocked the objects and relocked them in cadaver and retried, this time the locks were found again... I hadn't the time to do some more testing but I find this surprising. Does anyone have seen the same problem ? FYI Zope 2.5.0 + Python 2.1.2 both up-to-date Debian Woody packages. bye, Jerome Alet

3 4

multiple database connections
by Jerome Alet 07 Mar '02

07 Mar '02

Hi, First I have to say I've not tried what is discussed below, I'm just thinking about how this could be done... Say I have several relationnal databases, both have an identical structure. Since these databases are identical, all my ZSQL methods can be used with any of them. So I plan to architecture my Folder tree like that : Level0 | +-- ZSQLMethod1 +-- ZSQLMethod2 +-- ... +-- Level1-1 | | | +-- DAConnection | +-- ... some stuf specific to this database ... | +-- Level1-2 | +-- DAConnection +-- ... some stuff specific to this database ... Each DAConnection connects to a different database, but the id is identical in order for the ZSQL methods to find it in whichever folder they are called. Now the problem is : one can't create a ZSQL method if there's no DAConnection at the same level or above. I suppose that creating all my ZSQL methods in for example Level1-1, then cutting&pasting them in Level0 could work, but I find this very counter intuitive. And if this would work, then the ZSQL methods wouldn't be editable anymore since at Level0 there's no DA... So my question is : why the hell can't developpers create ZSQL methods as they see fit, and be allowed to type-in the name of a possibly not-yet-existing Database Connection ? After all, creating ZSQL methods isn't for end users, so maybe Zope could trust developpers in knowing exactly what they want to do... Adding a free text field and allowing a "Free" entry in the drop-down list in the ZSQL method creation form would help. The Test tab form could also be modified to allow the developper to choose any DA at test time. Any comment or a solution to do this the right way ? thanks in advance. Jerome Alet - alet(a)unice.fr - http://cortex.unice.fr/~jerome Fac de Medecine de Nice http://wwwmed.unice.fr Tel: (+33) 4 93 37 76 30 Fax: (+33) 4 93 53 15 15 28 Avenue de Valombrose - 06107 NICE Cedex 2 - FRANCE

3 3