[sqlalchemy] Conflicting columns with secondaryjoin
In my project, I have users, who receive messages, which can be
assigned any number of 'tags'. (In the same manner as Gmail's labels)
I've got (amongst others) the following tables:
- UserMessages (userId, messageId, ...)
- UserMessageTags (userId, messageId, tagId)
- UserTags (userId, tagId, ...)
What I'm trying to do is make all UserTags rows available in the
UserMessages objects. At the moment I have:
orm.mapper (UserMessages, userMessagesTable
,properties={
'tags': orm.relation(UserTags
,secondary=userMessageTagsTable
,primaryjoin=sql.and_(
userMessagesTable.c.userId==userMessageTagsTable.c.userId
,userMessagesTable.c.messageId==userMessageTagsTable.c.messageId
)
,secondaryjoin=sql.and_(
userMessageTagsTable.c.userId==userTagsTable.c.userId
,userMessageTagsTable.c.tagId==userTagsTable.c.tagId
)
,lazy=False
,uselist=True
)
}
)
This works find for reading in the data, but when I attempt to remove
an entry from the list:
userMessage.tags.pop()
session.flush()
I get an:
sqlalchemy.exceptions.InvalidRequestError: Column
'userMessageTags.userId' is not available, due to conflicting property
'userId':sqlalchemy.orm.properties.ColumnProperty object at
0xb75ba2ac
When I changed the primaryjoin and secondaryjoin conditions to only
use messageId and tagId, respectively, it worked as expected. Changing
them to both only use the userId columns caused the same
IvalidRequestError.
Is this a bug in SA, or in my implementation (which, admittedly, is
usually the case :) )?
Takk,
- Mel C
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups
sqlalchemy group.
To post to this group, send email to sqlalchemy@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en
-~--~~~~--~~--~--~---
[sqlalchemy] Re: pyodbc and tables with triggers
Rick Morrison wrote: Sorry, Stephen, I replied too early; your second email arrived before the first. A whole day before the first. So until we get a real cleanup, you're looking to try modules in this order: ['pyodbc', 'adodbapi', 'pymssql'] Sounds OK to me -- any objections out there? Looks good to me. I got slightly confused somewhere through this thread. When I was putting a test together for the passthrough patch, I ended up using an Import hook to force a particular dbapi module to be used programatically (given that I have all three installed). Obviously there are variations on that (manually renaming one etc) but have I missed anything more sophisticated using SA itself? Didn't look like it to me from the code. TJG --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: MS-SQL 2005: unicode inserts fail?
Hi, This could be a problem with reflection (autoload'ing the table). Try defining the table explicitly. Paul metaperl wrote: The following simple program dies (stack trace follows). It is in fact because of unicode character strings, because the same program works if you remove the u from in front of the strings. --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Inconsistent results in session.flush()
Hi, I'm having a problem where the results of session.flush() vary from one run to another of my test suite. The unit of work transaction dump is significantly different from one run to the next, similar to the issue in ticket 461. I haven't managed to make a test case small enough to post to the list yet, and I think I need to delve a little further into the code to find out why it's failing. (This is with both 0.3.5 and rev2416) The logs from the UOWTransaction on a failing run and a passing run are below. As well as the ordering being different, there is at least one class (ReleaseLine) that doesn't appear in the bad run. Unfortunately I don't know how to go about debugging this. I think I need to see exactly what is going on in the dependency sort. Do you have any suggestions for suitable places to add some extra logging? This is a failing run: INFO:sqlalchemy.orm.unitofwork.UOWTransaction.0x..30:Task dump: UOWTask(0x184b2b0, Component/component/None) (save/update phase) | |- UOWTask(0x184bb50, User/user/None) (save/update phase) | |- Save User(0x1851870) | | |- Process User(0x1851870).branches | | |- Process User(0x1851870).reviews | | |- Process User(0x1851870).labels | | |- Process Branch(0x17ee310).user | | |- Process Branch(0x184b190).user | | | |- UOWTask(0x184bb70, ChangeOrigin/change_origin/None) (save/update phase) | | | |- Process Branch(0x17ee310).change_origin | | | |- Process Branch(0x184b190).change_origin | | | | | |- UOWTask(0x184b590, Label/label/None) (save/update phase) | | | | | | | |- UOWTask(0x184b1b0, Branch/branch/None) (save/update phase) | | | | |- Save Branch(0x17ee310) | | | | |- Save Branch(0x184b190) | | | | | |- Process Branch(0x17ee310).review | | | | | |- Process Branch(0x184b190).review | | | | | | | | | |- UOWTask(0x183f470, Review/review/None) (save/update phase) | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb30, sqlalchemy.orm.dependency.MapperStub object at 0x0183FDF0) (save/update phase) | | | | | | |- Process Branch(0x17ee310).label | | | | | | |- Process Branch(0x184b190).label | | | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bfb0, Counter/counter/None) (save/update phase) | | | | |- Save Counter(0x184b0f0) | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb50, User/user/None) (delete phase) | | | |- UOWTask(0x184bb70, ChangeOrigin/change_origin/None) (delete phase) | | | | | |- UOWTask(0x184b590, Label/label/None) (delete phase) | | | | | | | |- UOWTask(0x184b1b0, Branch/branch/None) (delete phase) | | | | | | | | | |- UOWTask(0x183f470, Review/review/None) (delete phase) | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb30, sqlalchemy.orm.dependency.MapperStub object at 0x0183FDF0) (delete phase) | | | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bfb0, Counter/counter/None) (delete phase) | | | | | | | | | | | | | | | | | | | | | | | | | And on a good run looks like this: INFO:sqlalchemy.orm.unitofwork.UOWTransaction.0x..f0:Task dump: UOWTask(0x17f2610, User/user/None) (save/update phase) |- Save User(0x17fe9f0) | |- Process User(0x17fe9f0).labels | |- Process Branch(0x17caa10).user | |- Process Branch(0x17f2470).user | |- Process User(0x17fe9f0).reviews | |- Process User(0x17fe9f0).branches | |- UOWTask(0x17f2a50, ChangeOrigin/change_origin/None) (save/update phase) | | |- Process Branch(0x17caa10).change_origin | | |- Process Branch(0x17f2470).change_origin | | | |- UOWTask(0x17f29f0, Component/component/None) (save/update phase) | | | | | |- UOWTask(0x17f2ad0, ReleaseLine/release_line/None) (save/update phase) | | | | |- Process Branch(0x17caa10).release_line | | | | |- Process Branch(0x17f2470).release_line | | | | |- Process Counter(0x17fed10).release_line | | | | | | | |- UOWTask(0x17f2430, Counter/counter/None) (save/update phase) | | | | |- Save Counter(0x17fed10) | | | | | | | | | | | | | | | | |- UOWTask(0x17f2450, Branch/branch/None) (save/update phase) | | | | |- Save Branch(0x17caa10) | | | | |- Save Branch(0x17f2470) | | | | | |- Process Branch(0x17caa10).review | | | | |
[sqlalchemy] use_labels =30 vs MAX_LABEL_LENGTH
Hi Michael, I see that sql.py uses a limit of 30 characters to create the column label when use_labels is set to True. If name is greater than 30 char long, the label is trunked at position 24 and is appended a random integer to it. Since the name created in this way is less useful, I would like to sugest you to customize the maxlength of column names. (PostgreSQL accepts until 63 characters for names, with SQLite you may use very, very long names) def _get_label(self): if self.__label is None: if self.table is not None and self.table.named_with_column(): self.__label = self.table.name + _ + self.name #if self.table.c.has_key(self.__label) or len(self.__label) = 30: if self.table.c.has_key(self.__label) or len(self.__label) = MAX_LABEL_LENGTH: self.__label = self.__label[0:24] + _ + hex(random.randint(0, 65535))[2:] else: self.__label = self.name self.__label = .join([x for x in self.__label if x in legal_characters]) --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: error compile
Sébastien LELONG ha scritto: Seems that SA compiles in a wrong way my query... Can't what's wrong is happening... subvet appers to be a sub-select, so probably SA made some optimizations. You should print the whole query (print sql) and not the sub-query (as in your code: print subvet) to check if your query is actually compiled the wrong way. Seb sql=select([unita_aziendale.c.id], and_(unita_aziendale.c.id.in_( select([azienda_veterinario.c.id_unita_aziendale], and_(azienda_veterinario.c.id_veterinario==3, azienda_veterinario.c.id_unita_aziendale==unita_aziendale.c.id, azienda_veterinario.c.data_inizioNone, azienda_veterinario.c.data_fine==None ), from_obj=[ azienda_veterinario, unita_aziendale], ), )), ) print sql SELECT unita_aziendale.id FROM unita_aziendale WHERE unita_aziendale.id IN (SELECT azienda_veterinario.id_unita_aziendale AS id_unita_aziendale FROM azienda_veterinario WHERE azienda_veterinario.id_veterinario = %(azienda_veterinario_id_veterinario)s AND azienda_veterinario.id_unita_aziendale = unita_aziendale.id AND azienda_veterinario.data_inizio IS NOT NULL AND azienda_veterinario.data_fine IS NULL) -- As you can see the from_obj of subselect is wrong, the FROM should be: FROM azienda_veterinario, unita_aziendale j --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: Inconsistent results in session.flush()
On Mar 16, 2007, at 6:59 AM, King Simon-NFHD78 wrote:
Unfortunately I don't know how to go about debugging this. I think I
need to see exactly what is going on in the dependency sort. Do you
have
any suggestions for suitable places to add some extra logging?
these issues are often due to bugs in unitofwork.py and are
unbelievably tough to debug. which is why youre probably going to
have to get a test case for me since I really need to nail all of
these. if you want to turn on all logging available, youd say
logger.getLogger('sqlalchemy.orm').setLevel(logging.DEBUG), and youll
get a whole lot of stuff...after seeing all the logger names you can
tune out some of the noise with additional setLevel statements.
theres also a monkeypatch thats in the test suite which i use for
these, it intentionally reverses the order of topological sorts in
order to flush these out, it looks like:
from sqlalchemy.orm import unitofwork
from sqlalchemy import topological
class RevQueueDepSort(topological.QueueDependencySorter):
def __init__(self, tuples, allitems):
self.tuples = list(tuples)
self.allitems = list(allitems)
self.tuples.reverse()
self.allitems.reverse()
topological.QueueDependencySorter = RevQueueDepSort
unitofwork.DependencySorter = RevQueueDepSort
but when i get these, it takes me a few hours to wrap my head around
the issue, and im the one who wrote the code...so dont knock yourself
out on it...
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups
sqlalchemy group.
To post to this group, send email to sqlalchemy@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en
-~--~~~~--~~--~--~---
[sqlalchemy] Re: error compile
put correlate=False in your subquery. On Mar 16, 2007, at 12:43 PM, Jose Soares wrote: Hi, Seems that SA compiles in a wrong way my query... In [9]: sql=select([UnitaAziendale.c.id]) In [10]: subvet = select([azienda_veterinario.c.id_unita_aziendale], : and_(azienda_veterinario.c.id_veterinario==3, : azienda_veterinario.c.id_unita_aziendale==UnitaAziendale.c.id, : azienda_veterinario.c.data_inizioNone, : azienda_veterinario.c.data_fine==None : ) : ) In [11]: In [11]: print subvet SELECT azienda_veterinario.id_unita_aziendale FROM azienda_veterinario, unita_aziendale WHERE azienda_veterinario.id_veterinario = %(azienda_veterinario_id_v_5338)s AND azienda_veterinario.id_unita_aziendale = unita_aziendale.id AND azienda_veterinario.data_inizio IS NOT NULL AND azienda_veterinario.data_fine IS NULL In [12]: sql.append_whereclause(not_(UnitaAziendale.c.id.in_(subvet))) In [13]: print subvet (SELECT azienda_veterinario.id_unita_aziendale AS id_unita_aziendale FROM azienda_veterinario WHERE azienda_veterinario.id_veterinario = %(azienda_veterinario_id_v_5338)s AND azienda_veterinario.id_unita_aziendale = unita_aziendale.id AND azienda_veterinario.data_inizio IS NOT NULL AND azienda_veterinario.data_fine IS NULL) jo --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: error compile
Sébastien LELONG ha scritto: As you can see the from_obj of subselect is wrong, the FROM should be: FROM azienda_veterinario, unita_aziendale OK, I see... You probably mean that since your sub-select occurs on two tables, those have to be present in the FROM clause. I've tested this kind on query (select a from A where a in (select b from B where a = b)) and this works ok (MySQL). Your sub-query becomes invalid if taken alone, but the whole keeps correct. What's the error when performing manually the query (using mysql client or the like) ? Seb I know it works but the problem is the performance. this query takes 0m31.073s: SELECT unita_aziendale.id FROM unita_aziendale WHERE unita_aziendale.id IN (SELECT azienda_veterinario.id_unita_aziendale AS id_unita_aziendale FROM azienda_veterinario WHERE azienda_veterinario.id_veterinario = 3 AND azienda_veterinario.id_unita_aziendale = unita_aziendale.id AND azienda_veterinario.data_inizio IS NOT NULL AND azienda_veterinario.data_fine IS NULL) this one takes 0m0.686s SELECT unita_aziendale.id FROM unita_aziendale WHERE unita_aziendale.id IN (SELECT azienda_veterinario.id_unita_aziendale AS id_unita_aziendale FROM azienda_veterinario, unita_aziendale WHERE azienda_veterinario.id_veterinario = 3 AND azienda_veterinario.id_unita_aziendale = unita_aziendale.id AND azienda_veterinario.data_inizio IS NOT NULL AND azienda_veterinario.data_fine IS NULL) jo --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: pyodbc and tables with triggers
I don't think you missed anything; that more sophisticated approach is exactly what we're discussing. Seeing as how you have all three modules installed and switch between them, you probably have your own ideas about how it should work. Please pick up the discussion on ticket #480 and give us your input. Rick Rick Morrison wrote: Sorry, Stephen, I replied too early; your second email arrived before the first. A whole day before the first. So until we get a real cleanup, you're looking to try modules in this order: ['pyodbc', 'adodbapi', 'pymssql'] Sounds OK to me -- any objections out there? Looks good to me. I got slightly confused somewhere through this thread. When I was putting a test together for the passthrough patch, I ended up using an Import hook to force a particular dbapi module to be used programatically (given that I have all three installed). Obviously there are variations on that (manually renaming one etc) but have I missed anything more sophisticated using SA itself? Didn't look like it to me from the code. TJG --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: Conflicting columns with secondaryjoin
the error youre getting typically occurs when theres more than one
instance of a userMessageTags.c.userId Column instance. how there
might be more than one of the same Column instance can sometimes
happen when using table reflection and things get confused...although
I was pretty sure I had fixed all the issues with that.
so i would ask that you are on the latest version of SA (0.3.5), and
also let me know if youre using autoload=True to get your tables,
as well as any other tricks you might be doing with Table/MetaData,
and otherwise try to send me a reproducing test case. The joins you
are using should be valid, and i would suspect they are not the
problem, but are just leading to what the problem actually is; I
made a test program on this end using a many-to-many table with four
columns and it works.
On Mar 16, 2007, at 4:12 AM, Mel Collins wrote:
In my project, I have users, who receive messages, which can be
assigned any number of 'tags'. (In the same manner as Gmail's labels)
I've got (amongst others) the following tables:
- UserMessages (userId, messageId, ...)
- UserMessageTags (userId, messageId, tagId)
- UserTags (userId, tagId, ...)
What I'm trying to do is make all UserTags rows available in the
UserMessages objects. At the moment I have:
orm.mapper (UserMessages, userMessagesTable
,properties={
'tags': orm.relation(UserTags
,secondary=userMessageTagsTable
,primaryjoin=sql.and_(
userMessagesTable.c.userId==userMessageTagsTable.c.userId
,userMessagesTable.c.messageId==userMessageTagsTable.c.message
Id
)
,secondaryjoin=sql.and_(
userMessageTagsTable.c.userId==userTagsTable.c.userId
,userMessageTagsTable.c.tagId==userTagsTable.c.tagId
)
,lazy=False
,uselist=True
)
}
)
This works find for reading in the data, but when I attempt to remove
an entry from the list:
userMessage.tags.pop()
session.flush()
I get an:
sqlalchemy.exceptions.InvalidRequestError: Column
'userMessageTags.userId' is not available, due to conflicting property
'userId':sqlalchemy.orm.properties.ColumnProperty object at
0xb75ba2ac
When I changed the primaryjoin and secondaryjoin conditions to only
use messageId and tagId, respectively, it worked as expected. Changing
them to both only use the userId columns caused the same
IvalidRequestError.
Is this a bug in SA, or in my implementation (which, admittedly, is
usually the case :) )?
Takk,
- Mel C
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups
sqlalchemy group.
To post to this group, send email to sqlalchemy@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en
-~--~~~~--~~--~--~---
[sqlalchemy] Re: Inconsistent results in session.flush()
I can actually read a fair degree from these dumps, i need mostly to know what the actual dependencies are (i.e. which classes are dependent on what, whats the error). also when you do the full debug echoing the UOW should illustrate a series of dependency tuples which will show what pairs of classes the UOW perceives as dependent on each other. On Mar 16, 2007, at 6:59 AM, King Simon-NFHD78 wrote: Hi, I'm having a problem where the results of session.flush() vary from one run to another of my test suite. The unit of work transaction dump is significantly different from one run to the next, similar to the issue in ticket 461. I haven't managed to make a test case small enough to post to the list yet, and I think I need to delve a little further into the code to find out why it's failing. (This is with both 0.3.5 and rev2416) The logs from the UOWTransaction on a failing run and a passing run are below. As well as the ordering being different, there is at least one class (ReleaseLine) that doesn't appear in the bad run. Unfortunately I don't know how to go about debugging this. I think I need to see exactly what is going on in the dependency sort. Do you have any suggestions for suitable places to add some extra logging? This is a failing run: INFO:sqlalchemy.orm.unitofwork.UOWTransaction.0x..30:Task dump: UOWTask(0x184b2b0, Component/component/None) (save/update phase) | |- UOWTask(0x184bb50, User/user/None) (save/update phase) | |- Save User(0x1851870) | | |- Process User(0x1851870).branches | | |- Process User(0x1851870).reviews | | |- Process User(0x1851870).labels | | |- Process Branch(0x17ee310).user | | |- Process Branch(0x184b190).user | | | |- UOWTask(0x184bb70, ChangeOrigin/change_origin/None) (save/update phase) | | | |- Process Branch(0x17ee310).change_origin | | | |- Process Branch(0x184b190).change_origin | | | | | |- UOWTask(0x184b590, Label/label/None) (save/update phase) | | | | | | | |- UOWTask(0x184b1b0, Branch/branch/None) (save/update phase) | | | | |- Save Branch(0x17ee310) | | | | |- Save Branch(0x184b190) | | | | | |- Process Branch(0x17ee310).review | | | | | |- Process Branch(0x184b190).review | | | | | | | | | |- UOWTask(0x183f470, Review/review/None) (save/update phase) | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb30, sqlalchemy.orm.dependency.MapperStub object at 0x0183FDF0) (save/update phase) | | | | | | |- Process Branch(0x17ee310).label | | | | | | |- Process Branch(0x184b190).label | | | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bfb0, Counter/counter/None) (save/ update phase) | | | | |- Save Counter(0x184b0f0) | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb50, User/user/None) (delete phase) | | | |- UOWTask(0x184bb70, ChangeOrigin/change_origin/None) (delete phase) | | | | | |- UOWTask(0x184b590, Label/label/None) (delete phase) | | | | | | | |- UOWTask(0x184b1b0, Branch/branch/None) (delete phase) | | | | | | | | | |- UOWTask(0x183f470, Review/review/None) (delete phase) | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb30, sqlalchemy.orm.dependency.MapperStub object at 0x0183FDF0) (delete phase) | | | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bfb0, Counter/counter/None) (delete phase) | | | | | | | | | | | | | | | | | | | | | | | | | And on a good run looks like this: INFO:sqlalchemy.orm.unitofwork.UOWTransaction.0x..f0:Task dump: UOWTask(0x17f2610, User/user/None) (save/update phase) |- Save User(0x17fe9f0) | |- Process User(0x17fe9f0).labels | |- Process Branch(0x17caa10).user | |- Process Branch(0x17f2470).user | |- Process User(0x17fe9f0).reviews | |- Process User(0x17fe9f0).branches | |- UOWTask(0x17f2a50, ChangeOrigin/change_origin/None) (save/update phase) | | |- Process Branch(0x17caa10).change_origin | | |- Process Branch(0x17f2470).change_origin | | | |- UOWTask(0x17f29f0, Component/component/None) (save/update phase) | | | | | |- UOWTask(0x17f2ad0, ReleaseLine/release_line/None) (save/update phase) | | | | |- Process Branch(0x17caa10).release_line | | | |
[sqlalchemy] Re: cascading question
Humm, this doesn't help as if a site is deleted, it deletes the options
that where related to that option.
I want to be able to:
- delete an option without deleting a site
- delete a site without deleting an option
just delete (clean up) the related rows inside the weak
options_has_sites table.
I just can't figure it out...
Michael Bayer wrote:
youd need to add some delete cascades to your relationship, maybe
on just the backref (using the backref() function), e.g.
option_mapper = assign_mapper(ctx, Option, option_table,
properties={
'sites':relation(Site, backref=backref(options,
cascade=save-update, delete),
secondary=options_has_sites, cascade=save-update),
},
order_by=option_table.c.name,
)
On Mar 16, 2007, at 10:38 AM, Alexandre CONRAD wrote:
Hello,
I have a many-to-many relation between an option table and a site
table.
Deleting an option correctly deletes the related rows in
options_has_sites table.
But when I delete a site, I have some orphan rows in the
options_has_sites table. How can I avoid this ?
# SITE TABLE ---
site_table = Table('sites', meta,
Column('id', Integer, primary_key=True),
Column('name', Unicode(20), nullable=False, unique=True),
)
class Site(object):
pass
site_mapper = assign_mapper(ctx, Site, site_table,
order_by=site_table.c.name,
)
# OPTION TABLE
option_table = Table('options', meta,
Column('id', Integer, primary_key=True),
Column('name', Unicode(20), unique=True, nullable=False),
)
options_has_sites = Table('sites_has_options', meta,
Column('id_site', None, ForeignKey('sites.id'),
primary_key=True),
Column('id_option', None, ForeignKey('options.id'),
primary_key=True),
)
class Option(object):
pass
option_mapper = assign_mapper(ctx, Option, option_table,
properties={
'sites':relation(Site, backref=options,
secondary=options_has_sites, cascade=save-update),
},
order_by=option_table.c.name,
)
Should I play with backref() ?
Regards,
--
Alexandre CONRAD
---
Texte inséré par Platinum 2007:
S'il s'agit d'un mail indésirable (SPAM), cliquez sur le lien suivant pour
le reclasser : http://127.0.0.1:6083/Panda?ID=pav_31913SPAM=true
---
--
Alexandre CONRAD - TLV FRANCE
Research Development
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups
sqlalchemy group.
To post to this group, send email to sqlalchemy@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en
-~--~~~~--~~--~--~---
[sqlalchemy] Re: cascading question
the rows in the M2M table should be deleted automatically.it
*might* require that your instances are present in the session but
its not supposed to. can you make me a short test case for this
one ? its not the first time ive heard about it.
technically you can just put ON DELETE CASCADE on the tables too but
id like to fix this issue.
On Mar 16, 2007, at 12:19 PM, Alexandre CONRAD wrote:
Humm, this doesn't help as if a site is deleted, it deletes the
options
that where related to that option.
I want to be able to:
- delete an option without deleting a site
- delete a site without deleting an option
just delete (clean up) the related rows inside the weak
options_has_sites table.
I just can't figure it out...
Michael Bayer wrote:
youd need to add some delete cascades to your relationship, maybe
on just the backref (using the backref() function), e.g.
option_mapper = assign_mapper(ctx, Option, option_table,
properties={
'sites':relation(Site, backref=backref(options,
cascade=save-update, delete),
secondary=options_has_sites, cascade=save-update),
},
order_by=option_table.c.name,
)
On Mar 16, 2007, at 10:38 AM, Alexandre CONRAD wrote:
Hello,
I have a many-to-many relation between an option table and a site
table.
Deleting an option correctly deletes the related rows in
options_has_sites table.
But when I delete a site, I have some orphan rows in the
options_has_sites table. How can I avoid this ?
# SITE TABLE ---
site_table = Table('sites', meta,
Column('id', Integer, primary_key=True),
Column('name', Unicode(20), nullable=False, unique=True),
)
class Site(object):
pass
site_mapper = assign_mapper(ctx, Site, site_table,
order_by=site_table.c.name,
)
# OPTION TABLE
option_table = Table('options', meta,
Column('id', Integer, primary_key=True),
Column('name', Unicode(20), unique=True, nullable=False),
)
options_has_sites = Table('sites_has_options', meta,
Column('id_site', None, ForeignKey('sites.id'),
primary_key=True),
Column('id_option', None, ForeignKey('options.id'),
primary_key=True),
)
class Option(object):
pass
option_mapper = assign_mapper(ctx, Option, option_table,
properties={
'sites':relation(Site, backref=options,
secondary=options_has_sites, cascade=save-update),
},
order_by=option_table.c.name,
)
Should I play with backref() ?
Regards,
--
Alexandre CONRAD
-
--
Texte inséré par Platinum 2007:
S'il s'agit d'un mail indésirable (SPAM), cliquez sur le lien
suivant pour le reclasser : http://127.0.0.1:6083/Panda?
ID=pav_31913SPAM=true
-
--
--
Alexandre CONRAD - TLV FRANCE
Research Development
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups
sqlalchemy group.
To post to this group, send email to sqlalchemy@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en
-~--~~~~--~~--~--~---
[sqlalchemy] Re: SQLAlchemy test suite - no success/failure exit code
all tests pass on Mac OSX with the latest sqlite (as well as postgres, mysql). solaris, not so sure, ill have a look later. On Mar 16, 2007, at 1:19 PM, [EMAIL PROTECTED] wrote: Michael OK this was something small, the alltests.py scripts needed Michael to call testbase's main() function and not its runTests() Michael method, so that the exit code is propigated. rev 2414. skip Excellent. Trying a test run now... The good news is the test failures are now properly reported. The bad news is there are test failures. I'm going to try and get things set up to test on the Solaris systems at work, but it would help if people could take a peek at the test logs: http://www.python.org/dev/buildbot/community/all/g5%20OSX% 20trunk/builds/37/step-SQLAlchemy/0 http://www.python.org/dev/buildbot/community/all/g5%20OSX%202.5/ builds/16/step-SQLAlchemy/0 and let me know if any of those are expected. It would also help if someone using SA on Mac OSX could run the tests to see what (if anything) fails. Thx, -- Skip Montanaro - [EMAIL PROTECTED] - http://www.webfast.com/~skip/ The hippies and the hipsters did some great stuff in the sixties, but the geeks pulled their weight too. -- Billy Bragg --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: SQLAlchemy test suite - no success/failure exit code
I just ran my python 2.5 install (its a framework install) on OSX against the test suite on r2418, which uses the built-in sqlite3, and i got all tests passed. On Mar 16, 2007, at 1:19 PM, [EMAIL PROTECTED] wrote: Michael OK this was something small, the alltests.py scripts needed Michael to call testbase's main() function and not its runTests() Michael method, so that the exit code is propigated. rev 2414. skip Excellent. Trying a test run now... The good news is the test failures are now properly reported. The bad news is there are test failures. I'm going to try and get things set up to test on the Solaris systems at work, but it would help if people could take a peek at the test logs: http://www.python.org/dev/buildbot/community/all/g5%20OSX% 20trunk/builds/37/step-SQLAlchemy/0 http://www.python.org/dev/buildbot/community/all/g5%20OSX%202.5/ builds/16/step-SQLAlchemy/0 and let me know if any of those are expected. It would also help if someone using SA on Mac OSX could run the tests to see what (if anything) fails. Thx, -- Skip Montanaro - [EMAIL PROTECTED] - http://www.webfast.com/~skip/ The hippies and the hipsters did some great stuff in the sixties, but the geeks pulled their weight too. -- Billy Bragg --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: SQLAlchemy test suite - no success/failure exit code
geez, what version of sqlite / pysqlite / python is running there ? tons of those errors ive never seen before. On Mar 16, 2007, at 1:19 PM, [EMAIL PROTECTED] wrote: Michael OK this was something small, the alltests.py scripts needed Michael to call testbase's main() function and not its runTests() Michael method, so that the exit code is propigated. rev 2414. skip Excellent. Trying a test run now... The good news is the test failures are now properly reported. The bad news is there are test failures. I'm going to try and get things set up to test on the Solaris systems at work, but it would help if people could take a peek at the test logs: http://www.python.org/dev/buildbot/community/all/g5%20OSX% 20trunk/builds/37/step-SQLAlchemy/0 http://www.python.org/dev/buildbot/community/all/g5%20OSX%202.5/ builds/16/step-SQLAlchemy/0 and let me know if any of those are expected. It would also help if someone using SA on Mac OSX could run the tests to see what (if anything) fails. Thx, -- Skip Montanaro - [EMAIL PROTECTED] - http://www.webfast.com/~skip/ The hippies and the hipsters did some great stuff in the sixties, but the geeks pulled their weight too. -- Billy Bragg --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: SQLAlchemy test suite - no success/failure exit code
ah, in fact one failure youre having, the testcast failure, is indicative of running against a very old sqlite. FAIL: testcast (sql.select.SelectTest) -- Traceback (most recent call last): File /private/tmp/sqlalchemy/test/sql/select.py, line 737, in testcast check_results(sqlite.dialect(), ['NUMERIC(10, 2)', 'NUMERIC(12, 9)', 'DATE', 'TEXT', 'VARCHAR(20)'], '?') File /private/tmp/sqlalchemy/test/sql/select.py, line 723, in check_results self.assertEqual(str(cast(tbl.c.v1, Numeric).compile (dialect=dialect)), 'CAST(casttest.v1 AS %s)' %expected_results[0]) AssertionError: 'casttest.v1' != 'CAST(casttest.v1 AS NUMERIC(10, 2))' the CAST isnt happening for the sqlite dialect. this only occurs if this logic takes place: self.supports_cast = (sqlite is not None and vers (sqlite.sqlite_version) = vers(3.2.3)) if youre running py2.5 it should be using the sqlite3 thats included, which on my install is 3.3.5. otherwise you should run 3.3.13 since there was an order by bug in 3.3.12 and a few previous versions. On Mar 16, 2007, at 1:19 PM, [EMAIL PROTECTED] wrote: Michael OK this was something small, the alltests.py scripts needed Michael to call testbase's main() function and not its runTests() Michael method, so that the exit code is propigated. rev 2414. skip Excellent. Trying a test run now... The good news is the test failures are now properly reported. The bad news is there are test failures. I'm going to try and get things set up to test on the Solaris systems at work, but it would help if people could take a peek at the test logs: http://www.python.org/dev/buildbot/community/all/g5%20OSX% 20trunk/builds/37/step-SQLAlchemy/0 http://www.python.org/dev/buildbot/community/all/g5%20OSX%202.5/ builds/16/step-SQLAlchemy/0 and let me know if any of those are expected. It would also help if someone using SA on Mac OSX could run the tests to see what (if anything) fails. Thx, -- Skip Montanaro - [EMAIL PROTECTED] - http://www.webfast.com/~skip/ The hippies and the hipsters did some great stuff in the sixties, but the geeks pulled their weight too. -- Billy Bragg --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: SQLAlchemy test suite - no success/failure exit code
Michael all tests pass on Mac OSX with the latest sqlite (as well as Michael postgres, mysql). Hmmm... Then I wonder what's different about the buildbot setup and a normal interactive run. Michael solaris, not so sure, ill have a look later. There's no particular need. I was just hoping to get a different data point to compare my buildbot results with. You provided that with your Mac OSX results. Skip --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: SQLAlchemy test suite - no success/failure exit code
Michael geez, what version of sqlite / pysqlite / python is running Michael there ? tons of those errors ive never seen before. That will have to wait for me to get home. Probably not until tomorrow at the earliest. I'll let you know what I figure out. Perhaps I can tweak the buildslave script to emit the relevant version information. Skip --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: Inconsistent results in session.flush()
I've just run the attached script about thirty times, and it succeeded 5 times and failed the rest. I've cut out a lot of unnecessary stuff, but it's still a bit long I'm afraid. I'll cut it down some more, but since you seemed so eager to see it ;-) I thought I'd send it along as is. On a bad run, the dependency tuples look like this: DEBUG:sqlalchemy.orm.unitofwork.UOWTransaction.0x..50:Dependency sort: Mapper|User|user Mapper|Component|component Mapper|ChangeOrigin|change_origin Mapper|Label|label (cycles: [Mapper|Label|label, Mapper|ReleaseLine|release_line]) Mapper|Counter|counter Mapper|Branch|branch sqlalchemy.orm.dependency.MapperStub object at 0x00D59590 And on a good run they look like this: DEBUG:sqlalchemy.orm.unitofwork.UOWTransaction.0x..10:Dependency sort: Mapper|User|user Mapper|Component|component Mapper|ChangeOrigin|change_origin Mapper|ReleaseLine|release_line (cycles: [Mapper|ReleaseLine|release_line, Mapper|Label|label]) Mapper|Counter|counter Mapper|Branch|branch sqlalchemy.orm.dependency.MapperStub object at 0x00D59670 Thanks a lot for looking at this, Simon Michael Bayer wrote: I can actually read a fair degree from these dumps, i need mostly to know what the actual dependencies are (i.e. which classes are dependent on what, whats the error). also when you do the full debug echoing the UOW should illustrate a series of dependency tuples which will show what pairs of classes the UOW perceives as dependent on each other. On Mar 16, 2007, at 6:59 AM, King Simon-NFHD78 wrote: Hi, I'm having a problem where the results of session.flush() vary from one run to another of my test suite. The unit of work transaction dump is significantly different from one run to the next, similar to the issue in ticket 461. I haven't managed to make a test case small enough to post to the list yet, and I think I need to delve a little further into the code to find out why it's failing. (This is with both 0.3.5 and rev2416) The logs from the UOWTransaction on a failing run and a passing run are below. As well as the ordering being different, there is at least one class (ReleaseLine) that doesn't appear in the bad run. Unfortunately I don't know how to go about debugging this. I think I need to see exactly what is going on in the dependency sort. Do you have any suggestions for suitable places to add some extra logging? This is a failing run: INFO:sqlalchemy.orm.unitofwork.UOWTransaction.0x..30:Task dump: UOWTask(0x184b2b0, Component/component/None) (save/update phase) | |- UOWTask(0x184bb50, User/user/None) (save/update phase) | |- Save User(0x1851870) | | |- Process User(0x1851870).branches | | |- Process User(0x1851870).reviews | | |- Process User(0x1851870).labels | | |- Process Branch(0x17ee310).user | | |- Process Branch(0x184b190).user | | | |- UOWTask(0x184bb70, ChangeOrigin/change_origin/None) (save/update phase) | | | |- Process Branch(0x17ee310).change_origin | | | |- Process Branch(0x184b190).change_origin | | | | | |- UOWTask(0x184b590, Label/label/None) (save/update phase) | | | | | | | |- UOWTask(0x184b1b0, Branch/branch/None) (save/update phase) | | | | |- Save Branch(0x17ee310) | | | | |- Save Branch(0x184b190) | | | | | |- Process Branch(0x17ee310).review | | | | | |- Process Branch(0x184b190).review | | | | | | | | | |- UOWTask(0x183f470, Review/review/None) (save/update phase) | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb30, sqlalchemy.orm.dependency.MapperStub object at 0x0183FDF0) (save/update phase) | | | | | | |- Process Branch(0x17ee310).label | | | | | | |- Process Branch(0x184b190).label | | | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bfb0, Counter/counter/None) (save/ update phase) | | | | |- Save Counter(0x184b0f0) | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb50, User/user/None) (delete phase) | | | |- UOWTask(0x184bb70, ChangeOrigin/change_origin/None) (delete phase) | | | | | |- UOWTask(0x184b590, Label/label/None) (delete phase) | | | | | | | |- UOWTask(0x184b1b0, Branch/branch/None) (delete phase) | | | | | | | | | |- UOWTask(0x183f470, Review/review/None) (delete phase) | | | | | | | | | | | | | | |
[sqlalchemy] Use pyPgSQL module with SA?
Hi-
My apologies if this has been asked and answered, but I was unable to
find any reference. The SA documentation says it's possible to
pyPgSQL instead of psycopg2 but I can't make it work. Basically,
here's what I'm trying:
from sqlalchemy import *
from pyPgSQL import libpq
def myConn():
return libpq.PQconnectdb(host=localhost port=5432 dbname=...
user=... password=...)
db=create_engine('postgres://', creator=myConn)
This results in sqlalchemy.exceptions.InvalidRequestError: Cant get
DBAPI module for dialect 'sqlalchemy.databases.postgres.PGDialect
object
Any idea what dumb mistake I'm making? Thanks.
Larry
--~--~-~--~~~---~--~~
You received this message because you are subscribed to the Google Groups
sqlalchemy group.
To post to this group, send email to sqlalchemy@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/sqlalchemy?hl=en
-~--~~~~--~~--~--~---
[sqlalchemy] Re: SQLAlchemy test suite - no success/failure exit code
Michael ah, in fact one failure youre having, the testcast failure, Michael is indicative of running against a very old sqlite. Can you tell me what version I'm using other than very old? Here's what I see: $ otool -L lib-dynload/_sqlite3.so lib-dynload/_sqlite3.so: /usr/lib/libsqlite3.0.dylib (compatibility version 9.0.0, current version 9.6.0) /usr/lib/libmx.A.dylib (compatibility version 1.0.0, current version 93.0.0) /usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 88.1.8) and the stuff in /usr/lib looks fairly new: $ ls -Ll libsqlite3.* -rwxr-xr-x 1 root wheel 284508 Mar 14 20:04 libsqlite3.0.8.6.dylib -rwxr-xr-x 1 root wheel 284508 Mar 14 20:04 libsqlite3.0.dylib -rwxr-xr-x 1 root wheel 284508 Mar 14 20:04 libsqlite3.dylib I suppose it's possible that I'm picking up another version. I have an old version of Darwin Ports on the machine and Python comes with sqlite of some version. Skip --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: SQLAlchemy test suite - no success/failure exit code
go into python and type: from pysqlite2 import dbapi2 dbapi2.sqlite_version On Mar 16, 2007, at 3:23 PM, [EMAIL PROTECTED] wrote: Michael ah, in fact one failure youre having, the testcast failure, Michael is indicative of running against a very old sqlite. Can you tell me what version I'm using other than very old? Here's what I see: $ otool -L lib-dynload/_sqlite3.so lib-dynload/_sqlite3.so: /usr/lib/libsqlite3.0.dylib (compatibility version 9.0.0, current version 9.6.0) /usr/lib/libmx.A.dylib (compatibility version 1.0.0, current version 93.0.0) /usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 88.1.8) and the stuff in /usr/lib looks fairly new: $ ls -Ll libsqlite3.* -rwxr-xr-x 1 root wheel 284508 Mar 14 20:04 libsqlite3.0.8.6.dylib -rwxr-xr-x 1 root wheel 284508 Mar 14 20:04 libsqlite3.0.dylib -rwxr-xr-x 1 root wheel 284508 Mar 14 20:04 libsqlite3.dylib I suppose it's possible that I'm picking up another version. I have an old version of Darwin Ports on the machine and Python comes with sqlite of some version. Skip --~--~-~--~~~---~--~~ You received this message because you are subscribed to the Google Groups sqlalchemy group. To post to this group, send email to sqlalchemy@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/sqlalchemy?hl=en -~--~~~~--~~--~--~---
[sqlalchemy] Re: Inconsistent results in session.flush()
oh, this is easy. you have a circular insert relationship between label and release_line, which you can see in the sort there (the cycles:) as well as evidenced by the fact that you have a use_alter needed in order to create the foreign keys on those two tables. add post_update=True to one of the relationships which will break that particular cycle by using a second UPDATE statement, and youre in. On Mar 16, 2007, at 1:58 PM, King Simon-NFHD78 wrote: I've just run the attached script about thirty times, and it succeeded 5 times and failed the rest. I've cut out a lot of unnecessary stuff, but it's still a bit long I'm afraid. I'll cut it down some more, but since you seemed so eager to see it ;-) I thought I'd send it along as is. On a bad run, the dependency tuples look like this: DEBUG:sqlalchemy.orm.unitofwork.UOWTransaction.0x..50:Dependency sort: Mapper|User|user Mapper|Component|component Mapper|ChangeOrigin|change_origin Mapper|Label|label (cycles: [Mapper|Label|label, Mapper|ReleaseLine|release_line]) Mapper|Counter|counter Mapper|Branch|branch sqlalchemy.orm.dependency.MapperStub object at 0x00D59590 And on a good run they look like this: DEBUG:sqlalchemy.orm.unitofwork.UOWTransaction.0x..10:Dependency sort: Mapper|User|user Mapper|Component|component Mapper|ChangeOrigin|change_origin Mapper|ReleaseLine|release_line (cycles: [Mapper|ReleaseLine|release_line, Mapper|Label|label]) Mapper|Counter|counter Mapper|Branch|branch sqlalchemy.orm.dependency.MapperStub object at 0x00D59670 Thanks a lot for looking at this, Simon Michael Bayer wrote: I can actually read a fair degree from these dumps, i need mostly to know what the actual dependencies are (i.e. which classes are dependent on what, whats the error). also when you do the full debug echoing the UOW should illustrate a series of dependency tuples which will show what pairs of classes the UOW perceives as dependent on each other. On Mar 16, 2007, at 6:59 AM, King Simon-NFHD78 wrote: Hi, I'm having a problem where the results of session.flush() vary from one run to another of my test suite. The unit of work transaction dump is significantly different from one run to the next, similar to the issue in ticket 461. I haven't managed to make a test case small enough to post to the list yet, and I think I need to delve a little further into the code to find out why it's failing. (This is with both 0.3.5 and rev2416) The logs from the UOWTransaction on a failing run and a passing run are below. As well as the ordering being different, there is at least one class (ReleaseLine) that doesn't appear in the bad run. Unfortunately I don't know how to go about debugging this. I think I need to see exactly what is going on in the dependency sort. Do you have any suggestions for suitable places to add some extra logging? This is a failing run: INFO:sqlalchemy.orm.unitofwork.UOWTransaction.0x..30:Task dump: UOWTask(0x184b2b0, Component/component/None) (save/update phase) | |- UOWTask(0x184bb50, User/user/None) (save/update phase) | |- Save User(0x1851870) | | |- Process User(0x1851870).branches | | |- Process User(0x1851870).reviews | | |- Process User(0x1851870).labels | | |- Process Branch(0x17ee310).user | | |- Process Branch(0x184b190).user | | | |- UOWTask(0x184bb70, ChangeOrigin/change_origin/None) (save/update phase) | | | |- Process Branch(0x17ee310).change_origin | | | |- Process Branch(0x184b190).change_origin | | | | | |- UOWTask(0x184b590, Label/label/None) (save/update phase) | | | | | | | |- UOWTask(0x184b1b0, Branch/branch/None) (save/update phase) | | | | |- Save Branch(0x17ee310) | | | | |- Save Branch(0x184b190) | | | | | |- Process Branch(0x17ee310).review | | | | | |- Process Branch(0x184b190).review | | | | | | | | | |- UOWTask(0x183f470, Review/review/None) (save/update phase) | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb30, sqlalchemy.orm.dependency.MapperStub object at 0x0183FDF0) (save/update phase) | | | | | | |- Process Branch(0x17ee310).label | | | | | | |- Process Branch(0x184b190).label | | | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bfb0, Counter/counter/None) (save/ update phase) | | | | |- Save Counter(0x184b0f0) | | | | | | | | | | | | | | | | | | | | | | | | | |- UOWTask(0x184bb50, User/user/None) (delete phase) | | | |- UOWTask(0x184bb70,
