Chris Withers wrote:
1) Squishdot is based on very old, pre 2.0 code. It does all it's own indexing instead of using ZCatalog (because it is based on old code). It should not do that, and if it does, it should track the same indexing code as the catalog. As is, Squishdot cannot take advantage of the new language independent features, or wildcard searching, or Japanese support etc...
I'm looking to pick up on Squishdot development until Butch Landigan has some free time, where can I find out how things 'should' be indexed? In general where should I look for documentation on the python-end of Zope development?
The Interfaces Wiki: http://www.zope.org/Members/michel/Projects/Interfaces
There are other techical reasons, such as the old indexing code being very ZTables centric. I made the same mistake the original index authors made in that I made the Catalog indexes very ZCatalog centric. Neither are very useful in the general case (which is the real root of the problem here).
I think I'm missing something 'cos I dont' understand this :S Are you saying Catalog indexes are bad or are you saying ZCatalogs are badly designed?
They are not bad nor are they badly designed. They were designed before we starting taking a more seriously disciplined process toward development, and it was also my first serious Zope product. If I could do it over, I would do some minor re-engineering that would have eliminated these indexing problems, like basing the new indexes on the old indexes (subclassing). The problem was, the old indexes were too ZTables centric, and we could not 'fix' them because this would break ZTables/Confera, so a compromise was taken to have two different interfaces. So because the old interfaces did not do what ZCatalog wanted to do I created new ones that did, and thus, they were more ZCatalog centric. -Michel