pcplantdb

[Chad Knepp <pyg@galatea.org>]

Richard Morris writes:
 > Chad Knepp wrote:
 > > If we don't use Zope are you thinking that we would be using PHP?  My
 > > commitment to using python for this project is close to a blocking
 > > level issue.  If we use a python/RDBMS solution I am not at all sure
 > > php would integrate with this easily.  Were you thinking you would
 > > learn python then?
 > > 
 > No problem with learning python.

Good, no matter what we do at least it will be in python.

 > >  > Yes we can have different products running, but we could end up with
 > >  > having different ports for different parts, say
 > >  > http://permaculture.info/ - for non zope bits
 > >  > http://permaculture.info:8080/ - for zope bits
 > >  > (unless of course we go for an apache plug in and the sysadmins allow 
 > > it).
 > > 
 > > Why?  The URL structure would allow for this.  Example:
 > > permaculture.info/eden to access the old Eden engine,
 > > permaculture.info/pfaf to access PFAF, permaculture.info/hg, etc. and
 > > permaculture.info to be whatever the default is.  'mod_rewrite' is
 > > pretty much a standard part of Apache anyway.
 > > 
 > Because as Zope comes out of the box, it runs on a different port to
 > apache. Port 8000 - apache port 8080 - Zope.
 > mod_rewrite is a way round problem (might make things a bit slower?)

In $ZOPEINSTANCE/etc/zope.conf you just set 'base-port 80'

Using Apache with mod_rewrite:
<http://zope.org/Members/shaw/HowTo/ApacheFrontEnd2/>
<http://zope.org/Members/jec/virtual_howto_html/>

Using Apache with mod_proxy:
<http://zope.org/Members/anser/apache_zserver/>

 > >  > Also theres integration under the hood, how easy is it to pull things
 > >  > from the Z-ODB from outside. I believe people have posted stuff
 > >  > on how to do this from python. But will it limit developers to using 
 > >  > python in the future.
 > > 
 > > I don't really understand where you're going here, but I can't see the
 > > limitation.  As I said above Zope provides adapters for php and perl
 > > development.  Moving the dataset to an RDBMS would require and object
 > > relational mapper.  Such tools do exist although they are designed
 > > more for using an RDBMS while retaining an object model more so than
 > > transferring an existing dataset to an RDBMS.
 > 
 > Do you have details of these adapters

Jeez!  <http://zope.org/search?SearchableText=database+adapter>

 > >  > That is one advantage of sticking to MySQL for the plant database, 
 > >  > interms that basically every language can communicate relativly 
 > >  > painlessly with it.
 > > 
 > > As I've said before Zope can talk to any RDBMS as well.
 > > 
 > > Richard have you really thought about the kind of database schema it
 > > would require to implement relationships the way we have been talking
 > > about?  If you really think that MySQL is the way to go we can always
 > > use it as the database instead of ZODB.  I personally think this is a
 > > completely lossy idea.
 > 
 > Yes one table does it (object version rather similar)
 > 
 > Table relationships
 > Fields
 >      source - id for source object
 >      target - id for target object
 >      linktype - describes the exact nature of the link
 >                      "Guild member" "pest"
 >      metadata - who submitted it etc.
 > 
 > I've done something like this with the "White Oak Plant Guild page"
 > http://www.ibiblio.org/pfaf/pcplantdb/showthing.php?BOTNAME=White+Oak%2FHazelnut+community
 > The parentchild table is used to encode these relationships.

I don't think an adequate solution is as simple as that.  Since every
single thing has several to many important relationships with
practically every other thing a single table would approach
data_items^2 rows for only one relationship.  The only sane way I can
see to do this is with sets because most of these relationships are
not unique.  The relationship a rabbit in your garden has with most of
the stuff in your garden is pretty much the same (as I well know).  To
say Rabbit --Eats--> Beans, Rabbit --Eats--> Kale, etc. is going to
take a bit of space.  {rodents} --Eats--> {vegetables} is a bit more
compact.  Also what about locale specific differences?

 > There may actually been some benifit in following the structure of the 
 > XML XLink spec which is used to describe some very general and flexiable 
 > linking structure.
 > 
 > What we might need is some universal nameing structure so with a single
 > id we can point to any item in the system.
 > 
 > In an OO system
 > if all resources are sub classes of a single base class Say HGBaseClass 
 > then id's are simple a reference to an object of this type.
 > 
 > In a RDBM system then there is probably a need for a master table
 > 
 > Table: MasterReferences
 >      MasterID     - global id
 >      ResourceType - is this a wiki page, plant db page, photo etc.
 >      ResourceSpecificID - more specific id for that particular resource
 > 
 > One benifit of URL's is that the do provide this global namespace idea.

Yeah, that's almost as good as just using objects where all you would
need is a unique global id.

-- 
Chad Knepp
python -c 'import base64;print base64.decodestring("cHlnQGdhbGF0ZWEub3Jn")'