February 2000 - Zope-Dev

ZCatalog sort_on considered harmful
by Toby Dickenson 09 Feb '00

09 Feb '00

I have observed suprisingly poor ZCatalog performance when using the sort_on parameter in some specific (but fairly common) circumstances. The problem is due to the loop begining line 507 in Catalog.py, which traverses the whole index of the sort key. It is very efficient if the result set contains items with duplicate sort key values. For example sort_on='author', since many objects may have the same 'Author' property. However it is not efficient if the result set does not contain duplicate sort values, and it can be very inefficient if the result set only contains a small subset of the values in the whole sort key index (for example, sort_on='create_date'). The change below switches between two different schemes, based on the relative sizes of the index and result set. Any comments? # sort. If there are scores, then this block is not # reached, therefor 'sort-on' does not happen in the # context of text index query. This should probably # sort by relevance first, then the 'sort-on' attribute. if len(rs)>len(sort_index._index): # We have more hits than _unique_ values in the sort-key index. # This means that many of the hits have the same value sort-key. # We can use the index to efficiently partition all the hits into # those groups, then sort the groups. for k, intset in sort_index._index.items(): if type(rs) is IIBType: intset=rs.intersection(intset) else: intset=intset.intersection(rs) if intset: append((k,LazyMap(self.__getitem__, intset))) else: # We have more unique values in the sort-key index than # we do hits. I think the indexes are of no help here. for r in rs: append(sort_index._unindex[r],LazyMap(self.__getitem__,[r])) (also added to the collector) Toby Dickenson

1 0

Is object lookup in a container O(log n)?
by Kevin Dangoor 09 Feb '00

09 Feb '00

Hi, I just wanted to make sure of something before I make a change that will stick more than 1000 objects in one container... when you're looking up an object by ID, that's a O(log n) function, right? Kevin

3 3

Interbase and Zope
by Lee Catlett 09 Feb '00

09 Feb '00

Hi all, This is my first question and after looking in the archives decides this was the only place to find out for sure. I'm not saying it's not in the archive, only I didn't find my answer. Is anyone using Borland/Inprise's Interbase on a Linux box with Zope?? Am interested in putting a web site together with this combo if possible. Any comments welcome. Thanks, Lee

1 0

standard_html_header
by ethan mindlace fremen 09 Feb '00

09 Feb '00

Zopistas, I don't want to complicate zope, especially DTML, too much, but would it cause a huge problem if: <dtml> was aliased to <dtml-var standard_html_header> and </dtml> was aliased to <dtml-var standard_html_footer> ? I think it would be very handy and fairly logical. ~ethan

1 0

ZCatalog problems
by Dyon Balding 08 Feb '00

08 Feb '00

Hi, I setup an ZCatalog index "self._catalog.addIndex('description', 'TextIndex')". All the ZCatalog Aware objects have this description attrribute...which is just a string. Now when I go "<dtml-in "MyCatalog(description='blah')>" it should return all objects with 'blah' in the description attribute. But it is not! It is acting very strangly. It sometimes does after I do a reindex, but sometimes doesn't. My product inherits from ZCatatlog, and its default_catalog=self.id. Am I missing something? The fieldIndex's work fine. Is there some problem with the way lookups are done with TextIndex? I am using zope 2.1.3 from src. -- Dyon Balding, email: dyon(a)adroitnet.com.au work: +613 98888522, mobile: +61 0414992604 Adroit Internet Solutions

1 0

BLOBs with DCOracle and Oracle 8.1.5
by Dave Blasby 08 Feb '00

08 Feb '00

I'm trying to deal with Oracle 8.1 BLOBs inside my Zope application. Unfortunately, I've had no luck. I'm using the DCOracle 1.3.0 drivers. I've installed them, and they seem to work fine for non-BLOB data. First, I create a simple table, and add two row: SQL> create table blob_test ( the_blob BLOB, the_desc varchar(100) ); SQL> insert into blob_test values ( empty_blob(), 'test1'); SQL> insert into blob_test values ( empty_blob(), 'test2'); SQL> commit; Then, in python: % python Python 1.5.2 (#14, Feb 4 2000, 11:17:38) [GCC 2.8.1] on sunos5 Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam >>> import ocidb >>> import oci_8 >>> cnx = ocidb.Connect("scott/tiger") so far, so good. I can do simple SQL: >>> cnx.execute("select the_desc from blob_test") >>> cnx.fetchall() [('test1',), ('test2',)] Everything seems to be going well, until I try to manipulate a BLOB: >>> cnx.execute("select the_blob,the_desc from blob_test") >>> cnx.fetchone() Traceback (innermost last): File "<stdin>", line 1, in ? File "ociCurs.py", line 294, in fetchone self._error() File "ociCurs.py", line 81, in _error raise error, (rc, oci_.OracleErrorMessage(self._c.lda, rc)) oci.error: (24813, 'ORA-24813: cannot send or receive an unsupported LOB\012') OR, at a more simple level, you get the same result: ... >>> import oci_ >>> cnx.execute("select the_blob,the_desc from blob_test") >>> oci_.ofetch(cnx._c._c) -24813 According to the oracle documentation: ----- ORA-24813 cannot send or receive an unsupported LOB Cause: An attempt was made to send a LOB across the network, but either the server does not support the LOB sent by the client, or the client does not support the LOB sent by the server. This error usually occurs when the client and server are running different versions of Oracle. Action: Use a version of the Oracle that supports the LOB on both the client and the server. ---- So, I looked at the oci_8 module: ... >>> mylob = oci_8.lob(cnx._d) >>> print mylob <Lob object at 101d10> It appears that its created a LOB, but something is wrong: >>> mylob.length() -2130640896 >>> mylob.write('adslfjasf') -2 >>> mylob.length() -2130640896 Can anyone tell me what I'm doing wrong or point me somewhere? I've also tried using the perl DBI/DBD drivers and havent had any luck. thanks, dave

1 0

Helping Digital Creations fix bugs - suggestions
by Itamar Shtull-Trauring 08 Feb '00

08 Feb '00

I just answered a question on the Zope list about a bug in Gadfly which doesn't allow linebreaks. I was at least the second person to encounter this bug - I got the answer from someone else. The solution is to delete 1(!) character. I submitted it to the Collector, and now it says: Submitted: 2000/01/11 * * * This item is under review * * * okay, there was a conference, and it's not so big a bug, but I get the feeling nobody ever read the bug report, since fixing it takes 30 seconds at the most. I doubt it'll be fixed until someone from DC find the bug independently. I know DC asked about this at the conference, but I think people who weren't there might think of something too. Some uggested solutions (from what DC promised/suggested or): 1. Hand out portions of Zope to outside people (stuff like load_site.py or StructuredText comes to mind). I'd volunteer for load_site (proof of competence will be presented once I check that my merge with CVS actually works :). and/or Setup a (moderated?) mailing-list for each component devoted only to that specific component and force ZComponent's maintainer to read it. I'm thinking zope-zcatalog and zope-sql and zope-zpublisher, modeled after the mozilla and ptk lists. How many patches sent to zope-dev or zope lists got lost in the all the noise? 2. Change Collector so it has categories for stuff like Gadfly, StructuredText - for every Zope component. Remove obsolete stuff like ZTables. If I file a Gadfly bug, it should be emailed to the ZGadflyDA maintainer. Switch to Roundabout (http://www.lfw.org/python) when it's released? Anybody else have suggestions? -- Itamar S.T. itamars(a)ibm.net

1 0

Re: [Zope-PTK] Search broken: returns all entries?
by Tres Seaver 08 Feb '00

08 Feb '00

Mike Pelletier wrote: > > On Mon, 7 Feb 2000, Tres Seaver wrote: > > > The "quick search" field in the standard PTK header dumps the whole > > catalog -- I couldn't find a way to get any of the advanced search > > stuff to filter, either. > > > > Just FYI -- I do have the demo portal up and running. Thanks for a > > cool product, DC! > > This is fixed. The catalog creation bug should also finally be > fixed. I had appearently screwed up the check-in. I've been thinking about that one, just in general, trying to figure out which of the many WTDI (ways to do it) works best -- I think I will add a thumb rule to my list of Zopeisms as follows: * Any use of "manage_*" methods from within Python (not PythonMethods, though -- they don't have access to the Right Stuff) is highly suspect, and signals a need to refactor somewhere. This includes just about any method which takes REQUEST or RESPONSE as explicit parameters. -- ========================================================= Tres Seaver tseaver(a)palladion.com 713-523-6582 Palladion Software http://www.palladion.com

1 0

blah
by Terry Kerr 08 Feb '00

08 Feb '00

Hi, I setup an ZCatalog index "self._catalog.addIndex('description', 'TextIndex')". All the ZCatalog Aware objects have this description attrribute...which is just a string. Now when I go "<dtml-in "MyCatalog(description='blah')>" it should return all objects with 'blah' in the description attribute. But it is not! It is acting very strangly. It sometimes does after I do a reindex, but sometimes doesn't. My product inherits from ZCatatlog, and its default_catalog=self.id. Am I missing something? The fieldIndex's work fine. Is there some problem with the way lookups are done with TextIndex? I am using zope 2.1.3 from src. -- Terry Kerr (terry(a)adroitnet.com.au) Adroit Internet Solutions Pty Ltd (www.adroitnet.com.au) (03) 9888 8522 0414 938 124

1 0

Zope as a proxy - cache server
by Sin Hang Kin 08 Feb '00

08 Feb '00

Can I use zope as a proxy / cache server? Does it require to make changes on the zserver code or just zclient is enough? Rgs, Kent Sin

1 0