*** SachaS_ has quit IRC
*** btami has joined #gnuenterprise
<btami> good morning
*** reinhard has joined #gnuenterprise
<reinhard> good morning all
*** johannesV has joined #gnuenterprise
*** kilo has joined #gnuenterprise
<kilo> good morning
*** sjc has joined #gnuenterprise
*** lekma has joined #gnuenterprise
<lekma> morning everyone
<lekma> reinhard: u here?
<lekma> brb
*** wt has joined #gnuenterprise
<lekma> reinhard: i have a question about dirty reads
<lekma> from what i understand the postchange occurs
<lekma> during onvalidate if i call an external python method
<lekma> to make the object available to a session.find()
<lekma> but the object used during onvalidate is already available, and if it's not returned by a find
<lekma> it's not really a matter
<lekma> so unless i missed smthing
<lekma> you can delay the postChanges
<ajmitch_> hi all
<ajmitch_> reinhard: common & appserver packages done
<reinhard> lekma: the problem is not the object calling the onValidate
<reinhard> it's all the other objects
<ajmitch_> putting off sponsorship for now
<reinhard> so if I have object a and b
<reinhard> I must postchanges for object a so object b can see it in a find
<reinhard> and vice versa
<lekma> ok
<reinhard> ajmitch_: cool, thanks
<lekma> so should i try solution a) (delete all not null constraints from db except for gnue_id)
<lekma> to have a temp working solution
<lekma> ?
<reinhard> if it would really hold you back otherwise, then yes
<reinhard> it will take at least a week until we have the "clean" solution working
<reinhard> probably more
<reinhard> :(
<lekma> well well... we have to decide here what to do...
<lekma> reinhard: strange thing happening here
<lekma> when i session.request() with properties like
<lekma> sales_saleOrder
<lekma> everything is fine
<lekma> but if property is : sales_saleOrder.sales_orderId
<lekma> and sales_saleOrder is None
<lekma> xmlrpc does not send back anything
<lekma> s/xmlrpc/xmlrpc server
<reinhard> lekma: what would you think is the correct behaviour?
<lekma> sending back an empty value
<reinhard> in SQL DB terms, sales_saleOrder.sales_orderId results in a join
<reinhard> so you would think indirect properties should all yield "None" if the reference points to None?
<reinhard> so if sales_salesOrder is None
<lekma> well in fact the pb is in session.count, request returns listid correctly, but count returns 0, even thougth there are records that should be returned
<lekma> reinhard: yes
<reinhard> sales_salesOrder.sales_orderId, sales_salesOrder.sales_customer etc should all be none?
<lekma> it should at least return smthing
<reinhard> do you have any chance to debug the SQL statement that is issued by the request?
<reinhard> and would you know how the SQL statement would have to be constructed to result in that what would be more useful?
* reinhard thinks it should probably be an outer join or somehting
<reinhard> but not sure if all db's support that
<lekma> ok i'll try to get the debug and test sql statements in pg
*** nickr__ has joined #gnuenterprise
<reinhard> thanks
<lekma> what would be the smallest debug-level to get sql statement without everything else
<lekma> ?
<reinhard> I usually searchfor the sql statements in postgres' own log files
<reinhard> i think you need debug level 5 for sql statements
<reinhard> :q
<reinhard> err
<reinhard> wrong window
*** nickr has quit IRC
*** sjc has quit IRC
*** johannesV has quit IRC
*** johannesV has joined #gnuenterprise
*** sjc has joined #gnuenterprise
<reinhard> lekma: if you have a really easy reproducible test case, can you please try what changes if you change in geasSession.py after line 156 "elements = string.split (propertyname, '.')"  insert the line "if len (elements) > 1: return None"
<lekma> ok
<reinhard> you might not be able to use indirect fields in sort ordr and conditions then
<reinhard> but just to see what it would change
<lekma> well i stoped syncing with svn to avoid dirty read pbs, so in line 156 of geasSession.py i don't have the same, would you give me the method
<lekma> name
<lekma> pls
<reinhard> __getFieldname
<reinhard> at the very start
<reinhard> after elements = ...
<lekma> ok found it
<reinhard> however I still believe that an outer join would be much much better
<reinhard> does anybody know how the different db's implement outer joins?
<reinhard> and is gnue-common's dbdriver system capable of doing them?
<reinhard> bbl lunch
<lekma> reinhard: doesn't work
<lekma> error: column "none" does not exist
<lekma> i think the correct sql would be LEFT OUTER JOIN\
<lekma> i think the correct sql would be LEFT OUTER JOIN
<lekma> but i don't know if other dbs handle this
<reinhard> back for a minute (toast is in the machine)
<reinhard> thanks for testing
<reinhard> re outer join, how would the full SQL statement look like? (for example)
<reinhard> or do you have any pointer how OUTER JOIN syntax is?
<reinhard> hmmm... I guess I could search in the postgres docs myself...
<reinhard> seems like OUTER JOIN is a fairly standard feature of databases
<reinhard> pg, mysql, and firebird support it
<reinhard> even sqlite supports left outer join (but not right outer join)
<reinhard> so it should not be a too big problem
<reinhard> hmmm
<reinhard> except maybe
<reinhard> is outer join possible over more tables?
<reinhard> i want A outer join B, A outer join C, and C outer join D in the same SQL statement
<lekma> sorry i was away
<lekma> i don't know if outer join would solve my problem... i'm testing different sql statement right now
<lekma> cause it seems the error is coming from and AND clause
<lekma> cause it seems the error is coming from an AND clause
<reinhard> well
<reinhard> one part of the AND clause should actually be the LEFT OUTER JOIN ... ON... clause
<reinhard> (AFAICT now after having read a bit about JOIN syntax
<reinhard> )
<lekma> that's what i just found out
<lekma> that's it the AND clause : AND (t1.gnue_id = t0.sales_saleOrderRef)
<lekma> should not be there
<lekma> if i delete it and
<lekma> replace FROM clause by :FROM sales_SaleOrders t0 LEFT JOIN sales_SaleOrders t1 ON (t1.gnue_id = t0.sales_saleOrderRef)
<lekma> it works
<lekma> (please note that fk is on same class)
<lekma> but the pb is the same if fk is on an other class
<lekma> (i tested it)
<reinhard> lekma: yes
<reinhard> lekma: thanks for testing, I will look what we can do on this
<lekma> thx
<reinhard> lekma: what will you do about dirty reads?
<reinhard> should we fix the outer join stuff for current svn or for last release?
<reinhard> shall we revert the dirty read changes so you can use svn again?
<lekma> i'm a bit afraid about deleting NOT NULL constraints in db, and the modif in reading gcd stuff...
<lekma> if nobody uses dirty reads atm
<lekma> i would be glad if you revert it so that i can svn up again
<reinhard> ok
<reinhard> so you will wait for the "clean" solution
<lekma> i will try to hold others as long as i can
<lekma> :)
<reinhard> good
<reinhard> wish you luck in this :)
<lekma> bbl
*** neilt has joined #gnuenterprise
<neilt> good morning
*** mnemoc has joined #gnuenterprise
<kilo> bbl
*** kilo has quit IRC
<reinhard> hi neilt
<neilt> reinhard: hello
<neilt> reinhard:  if you have time I have a couple of questions?
<reinhard> neilt: just a sec, i am on the phone
<neilt> cool
<reinhard> neilt: sorry, still on the phone
<neilt> :)
<reinhard> ok
<reinhard> now i am here
<reinhard> please ask whatever you want to know
<neilt> do you have a tiome schedule for version 4 appserver?
<reinhard> yes
<reinhard> ASAP
<reinhard> :)
<neilt> is that realistic?
<reinhard> we are already working on it
<reinhard> for all the functions we want to have in 0.4 with acceptable stability it might be a month or so
<neilt> do you thing 1 week, 1 month or ?
<reinhard> there will be prereleases with (hopefully) good stability but not yet all functions in ROADMAP
<neilt> excellent
<johannesV> reinhard, i've just reverted the dirty-read stuff successfully ...
<neilt> are there any major issues with using appserver in a production CRM/Sales Order tracking process?
<johannesV> will do some cleanup (remove debug-code) and then commit it
<reinhard> what is it what you would regard most urgend of the 0.4 points?
<reinhard> urgent
<reinhard> for major issues you might really want to ask lekma
<neilt> reinhard: re 4.0 stability and security
<reinhard> as he uses appserver in a production evironment with 10 users á 5 connections 10 hour a day
<reinhard> actually neither stability nor security are issues on the roadmap for 0.4 AFAICT
<neilt> does lekma use it at his company or a client company?
<reinhard> there are no issues with stability currently AFAICT
<lekma> neilt: client company
<reinhard> lekma: thanks for jumpint in :)
<reinhard> jumping
<reinhard> sigh, can't type any more...
<neilt> reinhard:  from TODO
<neilt>   The 0.4 release will add some features to make Appserver usable in production
<neilt>   environments with limited security requirements.
<reinhard> oh
<reinhard> is that bad English?
<lekma> reinhard, johannesV : i may be wrong but it seems the memory leak we've seen, is back. But only when running server in debug mode.
<reinhard> actually I wanted to say something like "production environments where security is not an issue"
<lekma> maybe the gDebug thingy...
<reinhard> lekma: might well be, we didn't investigate too much in debug mode
<reinhard> neilt: I don't want to say that security is low on our priority list
<lekma> i forgot the server was running in debug mode... and boom
<reinhard> it's just that a lot of things are higher...
<neilt> reinhard: security is never not an issue.
<neilt> what is the current state of security
<reinhard> you can have a user database with user name, password, and a list of classes the user may access
<reinhard> this is very very primitive and only a temporary solution
<neilt> reinhard: what is the security in regards to processes that are not clients? For Example,
<reinhard> wrt server security against attacs it's problable that things like connection hijacking etc are possible
<neilt> is the DB open to all connections or do they have to have a password also/
<reinhard> db can have a password
<reinhard> you can set up the db with or without password however you like
<neilt> is anyone running the server with passwords?
<reinhard> and you can configure appserver to log onto the db with a password
<reinhard> i don't know
<reinhard> personally, i don't
<reinhard> (it's easier as my db isn't open to the internet)
<neilt> lekma: can you join #gnue?
<neilt> reinhard: you too if you have time
<lekma> neilt: i am alone in #gnue
<lekma> oops sorry
*** lekma has quit IRC
*** lekma has joined #gnuenterprise
*** jamest has joined #gnuenterprise
*** wt has quit IRC
*** jamest has left #gnuenterprise
*** R45 has quit IRC
*** R45 has joined #gnuenterprise
*** neilt has quit IRC
*** dcmwai has joined #gnuenterprise
<reinhard> johannesV: is the change of sort order upwards compatible?
<reinhard> i.e. can i still do find ('address_person', sortorder = ['property1', 'property2'])
<reinhard> ?
<johannesV> reinhard, yes you can
<johannesV> have you read the mail ?=
<johannesV> if so, i probably should explain it a bit better then ...
<reinhard> oops sorry
<johannesV> using your example would be treated as [('property1', False), (property2, False)]
<reinhard> i see you answer that question in your mail
<reinhard> must have scrolled over it when reading
<johannesV> ok, great, so my english isn't that bad :)
<reinhard> can i also do
<johannesV> mix it up ?
<reinhard> sortorder = [('foo', True), 'bar']
<reinhard> yes
<reinhard> you read my mind, don't you?
<reinhard> :)
<johannesV> haven't tried yet
<johannesV> but looks crazy ...
<johannesV> just a sec
<reinhard> ...johannesV: haven't tried yet
<reinhard> didn't ou write the program??
<reinhard> ;-)
*** titopbs has joined #gnuenterprise
<johannesV> hmm, well i did, but who said i should know what it's gonna do ?
<johannesV> reinhard, you *can* mix it up !
<reinhard> cool
<johannesV> hmm, i did a really defensive job then :)
<johannesV> but if you look at 'geasSession.py' line 370 you can see it
<johannesV> every item of the order-sequence will be extended with a fallback if it's missing
<johannesV> so, i'll try to create a test-case for the LEFT JOIN thingy
<reinhard> johannesV: the failed dirty read approach is reverted now, correct?
<johannesV> yes, it's reverted
<reinhard> so lekma could svn up again in his test environment
<johannesV> yeah, but he better updates his test-environment and do some tests before taking this svn version into production
<lekma> johannesV: yes i know
<reinhard> I take that for granted ;-)
<johannesV> :)
<lekma> :)
*** jamest has joined #gnuenterprise
*** btami has quit IRC
*** jcater has joined #gnuenterprise
*** kilo has joined #gnuenterprise
<reinhard> malek == lekma, correct?
<lekma> yes
<reinhard> :)
<lekma> old nickname that sticked
<lekma> brb
<johannesV> ok, i've a test-case for joins available now ...
<johannesV> this is quite scary ... :)
<johannesV> we really *need* that join-patch
*** wendall911 has joined #gnuenterprise
*** fx has joined #gnuenterprise
*** fx has left #gnuenterprise
*** cilkay has quit IRC
*** cilkay has joined #gnuenterprise
*** holycow has joined #gnuenterprise
*** cilkay has quit IRC
*** cilkay has joined #gnuenterprise
*** btami has joined #gnuenterprise
<lekma> bye
*** lekma has quit IRC
<johannesV> bbl
*** johannesV has quit IRC
*** johannesV has joined #gnuenterprise
<johannesV> hmm, is there any chance for a python class to implement some code which get's executed whenever it's involved in a "x is y" statement ?
<johannesV> like: if foo is None: .....
<jcater> "is" only applies to None
<jcater> and that's pretty specific statement
<jcater> what are you trying to accomplish?
<johannesV> i try to create a class which provides attribute access
<johannesV> like "foo.bar.baz"
<johannesV> but if bar is None .baz can't be evaluated
<johannesV> so it just should return None
<johannesV> i've created a class called NullObject
<johannesV> which has __getattr__ () returning itself, for a given amount of iterations
<johannesV> and if the max is reached it returns None
<johannesV> so it can handle foo.bar.baz, return None at the end
<johannesV> but if the number of steps (like foo.bar) is lower than the max i've got a little problem with statements like if foo.bar.baz is None:
<johannesV> so that's why i'm looking for something like __is__
<johannesV> but since "is" is the identity-operater it looks like i do have a problem
*** jamest has left #gnuenterprise
<johannesV> or do you have any idea ?
<johannesV> i could change all code of the type "if x is None" or "if x is not None" ... but that could be quite a lot
<reinhard> johannesV: why doesn't NullObject simply return None on __getattr__?
<johannesV> because it cannot handle "foo.bar.baz"
<johannesV> cause foo.bar would return None
<johannesV> so foo.bar   .baz ---> None.baz
<johannesV> which is not allowed
<reinhard> ah
<reinhard> ok
<reinhard> what about
<johannesV> if i could get the number of "planned" __getattr__-calls
<reinhard> NullObject knows what type it is of
<johannesV> of course it does know ... ?
<reinhard> and can look in the class dicitionary if it should return a "None" (for normal properties), another NullObject (for reference  props) and an exception (for procedures)
<reinhard> i mean what appserver class it stands for
<johannesV> hmm
<johannesV> that sounds like a good idea
<johannesV> will try that
<reinhard> bbl
<johannesV> but have to do a little sort on joined tables first
*** jamest has joined #gnuenterprise
<reinhard> hahaha
<reinhard> Linux and Windows - I want the best from both worlds
<reinhard> so I take Linux
<johannesV> reinhard: left outer joins and the new language interface are working !
<reinhard> seriously?
<johannesV> yes
<reinhard> did you test outer joins with different db's?
<johannesV> but i need to tweak geasFilter and geasSession a bit
<johannesV> no, not yet, but i will within the next 15 minutes
<reinhard> cool
<johannesV> (since i've a test-case with gcd/gld and .py)
<johannesV> (will commit that into samples/testing/joins)
*** kilo_ has joined #gnuenterprise
*** kilo has quit IRC
<johannesV> reinhard, it's working with sqlite !
<johannesV> reinhard, it's working with mysql
<johannesV> reinhard, it's working with firebird
<reinhard> great
<johannesV> if i would remember how to set the maxdb online i could test this db too
<reinhard> johannesV: i'm not especially worried about maxdb
<johannesV> well, but i am curious about *lol
<reinhard> i would be quite confused if maxdb missed a feature that even sqlite has
<reinhard> well
<reinhard> off to tv
<johannesV> me too
<johannesV> will commit tomorrow
<reinhard> johannesV: 21:45 - your time is running out ;-)
<johannesV> right
<reinhard> good night
<johannesV> good night
*** johannesV has quit IRC
*** btami has quit IRC
*** reinhard has quit IRC
*** nickr has quit IRC
*** nickr has joined #gnuenterprise
*** kilo_ has quit IRC
*** jamest has left #gnuenterprise
*** dcmwai has quit IRC
*** jcater has quit IRC
*** gnardo has joined #gnuenterprise
*** nickr has quit IRC
*** XEN has joined #gnuenterprise
*** holycow has quit IRC
*** titopbs has quit IRC
*** XEN has quit IRC
*** sjc has quit IRC
*** nickr__ has joined #gnuenterprise
*** nickr has quit IRC
*** jamest has joined #gnuenterprise
*** jamest has quit IRC
*** jamest has joined #gnuenterprise
*** XEN has joined #gnuenterprise
*** XEN has quit IRC
*** wendall911 has quit IRC
*** holycow has joined #gnuenterprise
*** jcater has joined #gnuenterprise
*** jamest has quit IRC
*** jcater has quit IRC