Casey Duncan <casey_duncan@yahoo.com> wrote:
--- Johan Carlsson <johanc@torped.se> wrote:
Hi, I need to check an assumption I have made on the ZCatalog Brains and Lazy Results from reading the source.
1. The Brains gets its schema from the ZCatalog and this schema is basicly fixed. E.g. a search result always has the attributes defined by the Meta Data in the ZCatalog?
Yes, you always get all the meta-data. However, does anyone know how the "used" argument of ZCatalog.searchResults works, or why you would use it?
'used' tells you the names of the indexes which actually participated in the search (as distinct from those you asked for).
I would rather be able to construct a Brain Schema at search time equivalent to the way theSELECT statement sets up the result attributes in SQL. Example: resultset =
Catalog.searchResults(query,schema=('id','title','keywords'))
Why is this important? I would suggest not putting big stuff in the meta-data that might warrant this. Just use the nifty new (v2.3) getObject method of the brain to access whatever properties/methods of the actual object you might need.
'getObject' forces you to "wake up" the catalogued object; in a large enough result set, the price is pretty punitive, compared to serving the results as cached metadata. Tres. -- =============================================================== Tres Seaver tseaver@digicool.com Digital Creations "Zope Dealers" http://www.zope.org
Casey Duncan <casey_duncan@yahoo.com> wrote:
Yes, you always get all the meta-data. However, does anyone know how the "used" argument of ZCatalog.searchResults works, or why you would use it?
'used' tells you the names of the indexes which actually participated in the search (as distinct from those you asked for).
What does this meen? Example, What's the different between: searchResults(used=(index1,index2), index1='kalle', index2='kaviar') and; searchResults(index1='kalle', index2='kaviar') (Are used supposed to be a tuple?) TIA, Johan Carlsson
Johan Carlsson wrote:
Casey Duncan <casey_duncan@yahoo.com> wrote:
Yes, you always get all the meta-data. However, does anyone know how the "used" argument of ZCatalog.searchResults works, or why you would use it?
'used' tells you the names of the indexes which actually participated in the search (as distinct from those you asked for).
What does this meen?
Example, What's the different between:
searchResults(used=(index1,index2), index1='kalle', index2='kaviar')
and;
Damn, where's this documented? I think this is _exactly_ what I've been looking for for ages, especially, I'm guessing, that if you don't specify 'used' then: 1. It's assumed all indexes take part in the search 2. All parameters that aren't supplied are assumed to be blank, and therefore will match anything? Am I on the right track? Furthermore, is there any way I can tell ZCatalog that nothing means nothing, not everything? For example, if I do searchResults(age=''), I want only records with a blank age record, _not_ all the records. Okay while I'm here, one final one. We have the sort_on and sort_order parameters, they'll return stuff ordered like: Aardvark Badger anteater bird Whereas most humans would think the following would be more readable: Aardvark anteater Badger bird I guess ZCatalog uses cmp somewhere along the line. IS there any way I can pass in a different cmp function for a given index? Hah, just thought of one last one ;-) Okay, we have sort_on, which lets me sort on one index. Is there any way I can provide a secondary sort key? something like searchResults(age='',sort_on=(('lastname','normal'),('firstname','normal'),'('salary','reverse'))) would eb a dream. Wow, that turned into a ZCatalog wishlist... any ideas if any of it is available, and if not, hwo hard it would be to achieve? cheers, Chris
On Thu, 22 Feb 2001, Chris Withers wrote:
I think this is _exactly_ what I've been looking for for ages, especially, I'm guessing, that if you don't specify 'used' then:
No, 'used' has *no* effect on the processing of the search; it exists only to allow keeping track of the indices used in searches across catalogs.
Okay while I'm here, one final one.
We have the sort_on and sort_order parameters, they'll return stuff ordered like:
Aardvark Badger anteater bird
Whereas most humans would think the following would be more readable:
Aardvark anteater Badger bird
Chris, this necrohippophilic sadism has to stop! "most humeans" does *not* map to "English speakers who prefer case-insensitive collation." Case-flattening is a lossy operation; there is *no way* you are going to get it by default. If you want the catalog to do a case-insensitive search / sort, then create a method on your object which does the approprate case-flattening, and index that instead of the "real" attribute. Tres. -- =============================================================== Tres Seaver tseaver@digicool.com Digital Creations "Zope Dealers" http://www.zope.org
Tres Seaver wrote:
Chris, this necrohippophilic sadism has to stop!
Erm? I think my education is lacking... what does that mean? :-S
"most humeans" does *not* map to "English speakers who prefer case-insensitive collation."
I disagree but it's not worth arguing about. However, the people who pay my salary feel that way which means I care ;-)
case-insensitive search / sort, then create a method on your object which does the approprate case-flattening, and index that instead of the "real" attribute.
Thanks, I always forget the flexibility of ZCatalog :-) That said, how hard would it be to alter ZCatalog so you could override the comparison function it uses? It feels like a ,cmp=cmp argument to the added search index consructor, then a self._cmp = cmp in the constructor, followed by a query and replace on cmp with self._cmp in the search index source. But I might be completely wrong :-( Apart from that, any ideas how I'd search for a record where the indexed attribute is blank or the secondary sort stuff? cheers, Chris
On Thu, 22 Feb 2001, Chris Withers wrote:
Tres Seaver wrote:
Chris, this necrohippophilic sadism has to stop!
Erm? I think my education is lacking... what does that mean? :-S
"Beating a dead horse". :)
"most humeans" does *not* map to "English speakers who prefer case-insensitive collation."
I disagree but it's not worth arguing about. However, the people who pay my salary feel that way which means I care ;-)
case-insensitive search / sort, then create a method on your object which does the approprate case-flattening, and index that instead of the "real" attribute.
Thanks, I always forget the flexibility of ZCatalog :-) That said, how hard would it be to alter ZCatalog so you could override the comparison function it uses? It feels like a ,cmp=cmp argument to the added search index consructor, then a self._cmp = cmp in the constructor, followed by a query and replace on cmp with self._cmp in the search index source. But I might be completely wrong :-(
The comparison function is built into the structure the of the index, and is not swappable "on-the-fly". When the catalog allows "drop-in" indexes, then you could define a CaseInsensitifeUnIndex class which implemented the semantics you prefer (but I still think it should be done at the application level, as I described above).
Apart from that, any ideas how I'd search for a record where the indexed attribute is blank or the secondary sort stuff?
"Trust the source, Luke!" is the best answer I've got. Tres. -- =============================================================== Tres Seaver tseaver@digicool.com Digital Creations "Zope Dealers" http://www.zope.org
Chris Withers wrote:
Tres Seaver wrote:
Chris, this necrohippophilic sadism has to stop!
Erm? I think my education is lacking... what does that mean? :-S
necrohippo philic sadism dead-horse-loving act of deliberate harm upon -- Steve Alexander Software Engineer Cat-Box limited http://www.cat-box.net
I disagree but it's not worth arguing about. However, the people who pay my salary feel that way which means I care ;-)
Ive been taking the Larry Wall approach to this when people tell me the correct way to do things: "A Perl script is "correct" if it gets the job done before your boss fires you." The simple fact is many non-technical users dont give a hoot about the underlying python reasons or computer science, its just common to see a case-insensitive sort. And yes it can be done in ZCatalog and dtml-in with a little fiddling. -- Andy McKay.
--- Andy McKay <andym@ActiveState.com> wrote:
I disagree but it's not worth arguing about. However, the people who pay my salary feel that way which means I care ;-)
Ive been taking the Larry Wall approach to this when people tell me the correct way to do things:
"A Perl script is "correct" if it gets the job done before your boss fires you."
[snip] Well hell, what's good enough for Perl is good enough for... *slaps self silly* C'mon, are you really seriously gonna back up reasoning for Zope features by quoting Larry Wall??? 8^) ===== | Casey Duncan | Kaivo, Inc. | cduncan@kaivo.com `-----------------> __________________________________________________ Do You Yahoo!? Yahoo! Auctions - Buy the things you want at great prices! http://auctions.yahoo.com/
On Thu, 22 Feb 2001, Chris Withers wrote:
Apart from that, any ideas how I'd search for a record where the indexed attribute is blank or the secondary sort stuff?
Somebody contributed a patch for the secondary sort stuff a while back. Whether it still works with 2.3.0 is a good question. I also remember a *long* time ago (back when I was first starting with zope about a year ago) someone posting a method for picking up all the records in a catalog with a blank value for a field. How you'd find it in the mail archives, though, I have no clue, nor do I remember what the method involved. --RDM
Hi Chris, RDM.. There's also Zieve for this sort of thing... it seems to work well enough for me... -steve
"RDM" == R David Murray <" <bitz@bitdance.com>> writes:
RDM> On Thu, 22 Feb 2001, Chris Withers wrote: >> Apart from that, any ideas how I'd search for a record where >> the indexed attribute is blank or the secondary sort stuff? RDM> Somebody contributed a patch for the secondary sort stuff a RDM> while back. Whether it still works with 2.3.0 is a good RDM> question. RDM> I also remember a *long* time ago (back when I was first RDM> starting with zope about a year ago) someone posting a method RDM> for picking up all the records in a catalog with a blank RDM> value for a field. How you'd find it in the mail archives, RDM> though, I have no clue, nor do I remember what the method RDM> involved. RDM> --RDM RDM> _______________________________________________ Zope-Dev RDM> maillist - Zope-Dev@zope.org RDM> http://lists.zope.org/mailman/listinfo/zope-dev ** No cross RDM> posts or HTML encoding! ** (Related lists - RDM> http://lists.zope.org/mailman/listinfo/zope-announce RDM> http://lists.zope.org/mailman/listinfo/zope )
"R. David Murray" wrote:
On Thu, 22 Feb 2001, Chris Withers wrote:
Apart from that, any ideas how I'd search for a record where the indexed attribute is blank or the secondary sort stuff?
Somebody contributed a patch for the secondary sort stuff a while back. Whether it still works with 2.3.0 is a good question.
Any idea where I'd find that? I may need to give it a go...
I also remember a *long* time ago (back when I was first starting with zope about a year ago) someone posting a method for picking up all the records in a catalog with a blank value for a field. How you'd find it in the mail archives, though, I have no clue, nor do I remember what the method involved.
Likewise with this. The only solution I can think of at the moment is to have a seperate 'x_is_blank' index that set to 'not x', but that seems terribly inefficient to me :-S Any ideas gratefully recieved :-) Chris
On Thu, 22 Feb 2001, Johan Carlsson wrote:
Casey Duncan <casey_duncan@yahoo.com> wrote:
Yes, you always get all the meta-data. However, does anyone know how the "used" argument of ZCatalog.searchResults works, or why you would use it?
'used' tells you the names of the indexes which actually participated in the search (as distinct from those you asked for).
What does this meen?
Example, What's the different between:
searchResults(used=(index1,index2), index1='kalle', index2='kaviar')
and;
searchResults(index1='kalle', index2='kaviar')
There is no practical difference, unless you are "federating" queries across multiple catalogs; the 'used' parameter is there solely to accumulate a record of the indices used in the search.
(Are used supposed to be a tuple?)
It should be something supporting the mapping interface. Here is how it is used (from 'lib/python/Products/ZCatalog/Catalog.py'): if used is None: used={} for i in self.indexes.keys(): try: index = self.indexes[i].__of__(self) if hasattr(index,'_apply_index'): r=index._apply_index(args) if r is not None: r, u = r for name in u: used[name]=1 In other words, "Nothing to see here; move along." :) Tres. -- =============================================================== Tres Seaver tseaver@digicool.com Digital Creations "Zope Dealers" http://www.zope.org
participants (8)
-
Andy McKay -
Casey Duncan -
Chris Withers -
Johan Carlsson -
R. David Murray -
Steve Alexander -
Steve Spicklemire -
Tres Seaver