[sqlalchemy] Re: Replicating a transaction
On Tuesday 20 March 2007 07:47:53 Benno Rice wrote: Hi, I'm wondering if it would be possible to have a situation where transactions coming in from the ORM system could be sent to multiple database servers at once. If I were to look into adding this, where would be the best place to start? although i'm not the authority to answer, i would probably subclass Engine and makeing it multi-sub-engine. Don't know how diff.dialects will be tackled though.. e.g one db is sqlite for quick and dirty caching, another is oracle for secure back-storage, etc. Also no idea how would u synchronise the db-responses, i.e. when to consider a transaction ready - when 1st db is ready, or when last db is ready etc. Same for errors - if one db throws and others do not.. i guess this can be ery usable also for full-replication mechanisms. svil --~--~-~--~~~---~--~~ 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: How to model an odd many-to-many relationship?
On 3/20/07, Mike Kent [EMAIL PROTECTED] wrote: I have a many-to-many relationship between two entities, call them A and B. This is easy enough to model in sqlalchemy using Elixir. However, to complicate things, I need an integer column somewhere called 'priority'. In the relationship between an A and multiple Bs, I want the Bs to be ordered by the value of the 'priority' column. In the relationship between a B and multiple As, the value of 'priority' is irrelevant. The problem is, I don't know what entity needs to have the 'priority' column in it, or how to model this in the relationship between A and B. I thought that maybe the 'priority' column should be in the secondary table that handles the many-to-many relationship between A and B, but I'm not sure how to set that up, and I'm not sure that Elixir can handle that at all. Basically Elixir doesn't handle that case. For now, your only option is to make a separate entity for the intermediary table, which would hold two belongs_to relationships, and use has_many relationships in your existing entities (A and B) to that new entity, instead of using has_and_belongs_to_many relationships. Now on the SQLAlchemy's side, you can either use that solution too, or use an AssociationProxy. See: http://www.sqlalchemy.org/docs/plugins.html#plugins_associationproxy I personally feel the associationproxy should be better integrated into the normal relation function (preferably as an option), but Michael Bayer doesn't agree with me on this. My point is that many-to-many relationship are already doing some kind of associationproxy behind the scenes, so I don't see why you should use an immensely different way to set things up simply because you have extra fields in your intermediary table. As for Elixir, I've been thinking to integrate the associationproxy into the has_and_belongs_to_many relationship for some time now. I'll do it if enough people are interested. -- Gaëtan de Menten http://openhex.org --~--~-~--~~~---~--~~ 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] Fix on ADODBAPI
I found that the conenctionString is not complete, the port is
missing. Not like anywhere else, host and port must be separated by a
comma !
def make_connect_string(keys):
connectors = [Provider=SQLOLEDB]
connectors.append (Data Source=%s,%s % (keys.get(host),
keys.get(port, 1433)))
connectors.append (Initial Catalog=%s % keys.get(database))
user = keys.get(user)
if user:
connectors.append(User Id=%s % user)
connectors.append(Password=%s % keys.get(password, ))
else:
connectors.append(Integrated Security=SSPI)
return [[;.join (connectors)], {}]
--~--~-~--~~~---~--~~
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: SELECT 'a fixed string', ...
Bertrand Croq wrote:
hi,
I am currently using sqlalchemy to build SQL queries and it's a fantastic
tool! By now, I am looking for a way to build:
SELECT 'a_fixed_string', atable.col1, atable.col2
FROM atable
using the syntax:
select([XXX, atable.c.col1, atable.c.col2])
but I don't know what to put at XXX.
Does anyone have a clue
try this:
select([literal('a_fixed_string'), atable.c.col1, atable.c.col2])
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: Fix on ADODBAPI
I also found that when trying to connect within a thread the connection hangs. I had to use pythoncom.CoInitialize(). Example: # import pythoncom import sys sys.coinit_flags = 0 pythoncom.CoInitialize() [ COM code here... ] pythoncom.CoUninitialize() More informations: http://twistedmatrix.com/pipermail/twisted-python/2004-June/007997.html --~--~-~--~~~---~--~~ 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: SELECT 'a fixed string', ...
Le Mardi 20 Mars 2007 10:38, jose a écrit :
try this:
select([literal('a_fixed_string'), atable.c.col1, atable.c.col2])
Perfect ! Thanks a lot.
--
Bertrand Croq,
Ingénieur Développement
___
Net-ng Tel : +33 (0)223 21 21 52
Immeuble Germanium Fax : +33 (0)223 21 21 60
80 av. des Buttes de CoesmesWeb : http://www.net-ng.com
35000 RENNESe-mail: [EMAIL PROTECTED]
FRANCE
Ce message et tout document joint sont confidentiels. Toute diffusion ou
publication en est interdite. Si vous recevez ce message par erreur,
merci d'en avertir immédiatement l'expéditeur par e-mail et de
supprimer ce message et tout document joint.
This message and any attachment are confidential. Any use is prohibited
except formal approval. If you receive this message in error, please
notify the sender by return e-mail and delete this message and any
attachment from your system.
--~--~-~--~~~---~--~~
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: migrate tool
Thanks Evan, Hope that soon I'll use your tool. Stefan --~--~-~--~~~---~--~~ 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: Issue using rev 2425
Michael, thanks for working through this and for taking the time to explain what's going on and to provide alternative ways to getting this done. Your efforts to support the users of sqlalchemy are really extraordinary. --~--~-~--~~~---~--~~ 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: How to model an odd many-to-many relationship?
Gaetan, Thanks for your response. I'd like to follow up with two questions: 1. I understand what you mean about the current way I'd have to model this relationship in Elixir. This would allow me full control over the intermediate table, which means I could put my 'priority' field in it. However, I believe this will also prevent me from getting 'all Bs for a given A, in priority order', by simply accessing the correct property in an A. Instead, when accessing that property, I could get all occurances of the intermediate table for an A. Can you give me some idea of how I would set up a select that would give me all Bs for an A, in 'priority' order? 2. When you talk about 'Now on the SQLAlchemy side', it makes me thing that in addition to setting up the relationships in Elixir as you specified, I would also need to do something directly in SQLAlchemy. Is that what you meant, or am I misinterpreting you? Finally, I do hope you will add the capability for finer-grained control over the intermediate table of a many-to-many relationship to Elixir. I would think that this kind of ordered many-to-many relationship is a prevalent enough use case to warrant 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: [ticket:336] Informix support and some enhancement for oracle and pgsql
label length of 18 is very small, we might have to resolve ticket #512 first (makes the truncation size non-global). please check my patch. we add a attribute in dialet for this case. --~--~-~--~~~---~--~~ 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: small bug/feature in PassiveDefault for timestamp in MySQL (or am I doing something wrong?)
mysql.py gets MSTimestamp added, which probably subclasses
MSDateTime, which sends along TIMESTAMP, and gets a fix in
ischema_names to support reflection. colspecs dict gets
sqltypes.TIMESTAMP: MSTimestamp added.
*perhaps*, types.py should have Timestamp added to it, TIMESTAMP will
subclass Timestamp. but i dont think this is needed quite yet.
can you send me a patch ?
On Mar 20, 2007, at 11:11 AM, eppye wrote:
I'm using SQLalchemy 0.3.4 with MySQL 5.0 and MySQLdb 1.2.2 and try
to insert this table:
tas = Table(
'ta', metadata,
Column('idta',Integer,primary_key=True,autoincrement=False),
Column('ts',TIMESTAMP,PassiveDefault(text(CURRENT_TIMESTAMP))),
I get the next error (did some formatting on it):
sqlalchemy.exceptions.SQLError: (OperationalError) (1067, Invalid
default value for 'ts')
'\nCREATE TABLE ta (
\n\tidta INTEGER NOT NULL,
\n\tts DATETIME DEFAULT CURRENT_TIMESTAMP,
\n\tPRIMARY KEY (idta)\n
) ENGINE=InnoDB\n\n' ()
MySQL does not accept this.
When using SQL-directly this SQL is accepted:
'''CREATE TABLE ta (
idtaINTEGER NOT NULL,
ts TIMESTAMP DEFAULT
CURRENT_TIMESTAMP,
PRIMARY KEY (idta)
)
Then I realised that SQLAlchemy passes a TIMESTAMP as DATETIME.
I fixed this in databases\mysql.py so SQLAlchemy now passes
TIMESTAMP;
and all goes OK.
Is this a 'bug', a feature or is there a better way?
Best regards,
eppye
--~--~-~--~~~---~--~~
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: Replicating a transaction
SessionTransaction will start transactions across any number of databases, as different engines come into its querying scope. when SessionTranasction commits, it commits all underlying transacitons. so if you want to start all transactions at once, you can create a new SessionTransaction, and just add() each engine you want. at the point of add(), it retrieves a connection from the pool and calls begin() on it. however, this is not using two phase commit, which means that if one commit fails, all previous commit's stay committed. if you want true two phase commit I'd look into Zalchemy which has this feature. On Mar 20, 2007, at 1:47 AM, Benno Rice wrote: Hi, I'm wondering if it would be possible to have a situation where transactions coming in from the ORM system could be sent to multiple database servers at once. If I were to look into adding this, where would be the best place to start? Many thanks. -- Benno Rice [EMAIL PROTECTED] http://jeamland.net/ --~--~-~--~~~---~--~~ 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: Replicating a transaction
On Mar 20, 2007, at 11:52 AM, Michael Bayer wrote: however, this is not using two phase commit, which means that if one commit fails, all previous commit's stay committed. if you want true two phase commit I'd look into Zalchemy which has this feature. actually i should correct myself - technically the commit strategy inside of SessionTransaction is more or less equivalent to what Zalchemy is doing right now - when you say trans.commit(), flush() will be called on the underlying session which will issue SQL to all involved databases. *then* it will go through and commit each database individually. since all SQL has been issued, its usually* the case that any exceptions that would have been raised have been raised already, and the commits are guaranteed** to succeed. * if you have your database set up to defer constraint checking until commit, then this wont work; exceptions will all get thrown at the COMMIT phase. zalchemy doesnt solve this either. ** guaranteed as close as we can get without accessing a true two- phase API on the database itself, which AFAICT is not possible with DBAPI. also if constraint checking is deferred until commit, like above the whole strategy fails. --~--~-~--~~~---~--~~ 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] Dealing with uncommited data
In a traditional application you can insert a new row and read the row within the same transaction. What is the typical usage pattern to deal with this in SA? In our particular setup (Zope) a new session is created for each new HTTP request and flushed automatically at the end of request when Zope commits the transaction. A solution might be to flush the session explicitly in such a case however I don't think that this is a smart idea. Any better idea? Andreas pgp3WFPsVJhgb.pgp Description: PGP signature
[sqlalchemy] Re: Dealing with uncommited data
Andreas Jung wrote: In a traditional application you can insert a new row and read the row within the same transaction. What is the typical usage pattern to deal with this in SA? In our particular setup (Zope) a new session is created for each new HTTP request and flushed automatically at the end of request when Zope commits the transaction. A solution might be to flush the session explicitly in such a case however I don't think that this is a smart idea. Any better idea? You can call session.flush() in between. This writes to the transaction but does not commit anything yet If you need to directly query the tables afterwards (triggers, text query, whatever), but inside the same transaction: session.flush() conn = session.context.get_current().connection(SomeMappedClass) conn.execute(sa.select(...)...) (at least, works for me with turbogears that wraps requests with transactions the same way as zope) --~--~-~--~~~---~--~~ 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] table name/Foreign key
Hi,
I have a dump question about naming conventions for foreign keys.
Using ORACLE as back-end all table names are in capital letters. So
Table object looks like:
Table('BRANCH',DynamicMetaData(),Column('id',OracleInteger(),primary_key=True,nullable=False),
Column('name',OracleString(length=500),nullable=False),
Column('lastmodifiedby',OracleInteger(),
ForeignKey('person.id')),
Column('lastmodificationdate',OracleTimestamp(timezone=False),
default=PassiveDefault(sqlalchemy.sql._TextClause object at
0xb6b142ac)),
Column('creationdate',OracleTimestamp(timezone=False),
default=PassiveDefault(sqlalchemy.sql._TextClause object at
0xb6b1430c)),
Column('createdby',OracleInteger(),ForeignKey('person.id')),schema=None)
The question is why Foreign key is in lower case, since the table
person name is actually in capital
Table('PERSON',DynamicMetaData())
Thanks,
Valentin.
--~--~-~--~~~---~--~~
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: table name/Foreign key
it would appear the query to read back constraint information is:
SELECT
ac.constraint_name,
ac.constraint_type,
LOWER(loc.column_name) AS local_column,
LOWER(rem.table_name) AS remote_table,
LOWER(rem.column_name) AS remote_column,
LOWER(rem.owner) AS remote_owner
FROM all_constraints%(dblink)s ac,
all_cons_columns%(dblink)s loc,
all_cons_columns%(dblink)s rem
WHERE ac.table_name = :table_name
AND ac.constraint_type IN ('R','P')
AND ac.owner = :owner
AND ac.owner = loc.owner
AND ac.constraint_name = loc.constraint_name
AND ac.r_owner = rem.owner(+)
AND ac.r_constraint_name = rem.constraint_name(+)
-- order multiple primary keys correctly
ORDER BY ac.constraint_name, loc.position, rem.position
so...theres your lower case.
On Mar 20, 2007, at 4:38 PM, vkuznet wrote:
Hi,
I have a dump question about naming conventions for foreign keys.
Using ORACLE as back-end all table names are in capital letters. So
Table object looks like:
Table('BRANCH',DynamicMetaData(),Column('id',OracleInteger
(),primary_key=True,nullable=False),
Column('name',OracleString(length=500),nullable=False),
Column('lastmodifiedby',OracleInteger(),
ForeignKey('person.id')),
Column('lastmodificationdate',OracleTimestamp(timezone=False),
default=PassiveDefault(sqlalchemy.sql._TextClause object at
0xb6b142ac)),
Column('creationdate',OracleTimestamp(timezone=False),
default=PassiveDefault(sqlalchemy.sql._TextClause object at
0xb6b1430c)),
Column('createdby',OracleInteger(),ForeignKey
('person.id')),schema=None)
The question is why Foreign key is in lower case, since the table
person name is actually in capital
Table('PERSON',DynamicMetaData())
Thanks,
Valentin.
--~--~-~--~~~---~--~~
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: Feature request: Session.get_local()
A patch containing tests and cleaned up identity key is attached.
~ Daniel
Michael Bayer wrote:
committed, r2409.
the row needs to have a class and entity_name present to determine
what mapper you want to use to extract from the row, so i put those as
keyword arguments for now.
also, I notice the usage of plain old assert for argument checking.
should we make this change across the board and get rid of
exceptions.ArgumentError ? i feel like we have to go one way or the
other with that.
also i didnt do any testing of this, we might want to add some tests to
test/orm/session.py .
On Mar 12, 2007, at 9:21 PM, Daniel Miller wrote:
def identity_key(self, *args, **kwargs):
Get an identity key
Valid call signatures:
identity_key(class_, ident, entity_name=None)
class_ - mapped class
ident - primary key, if the key is composite this is a tuple
entity_name - optional entity name. May be given as a
positional arg or as a keyword arg.
identity_key(instance=instance)
instance - object instance (must be given as a keyword arg)
identity_key(row=row)
row - result proxy row (must be given as a keyword arg)
if args:
kw = {}
if len(args) == 2:
class_, ident = args
entity_name = kwargs.pop(entity_name, None)
assert not kwargs, (unknown keyword arguments: %s
% (kwargs.keys(),))
else:
assert len(args) == 3, (two or three positional args are
accepted, got %s % len(args))
class_, ident, entity_name = args
mapper = _class_mapper(class_, entity_name=entity_name)
return mapper.instance_key_from_primary_key(ident,
entity_name=entity_name)
else:
try:
instance = kwargs.pop(instance)
except KeyError:
row = kwargs.pop(row)
assert not kwargs, (unknown keyword arguments: %s
% (kwargs.keys(),))
mapper = # not sure how to get the mapper form a row
return mapper.identity_key_from_row(row)
else:
assert not kwargs, (unknown keyword arguments: %s
% (kwargs.keys(),))
mapper = _object_mapper(instance)
return mapper.identity_key_from_instance(instance)
--~--~-~--~~~---~--~~
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
-~--~~~~--~~--~--~---
Index: lib/sqlalchemy/orm/session.py
===
--- lib/sqlalchemy/orm/session.py (revision 2432)
+++ lib/sqlalchemy/orm/session.py (working copy)
@@ -452,54 +452,57 @@
identity_key(class\_, ident, entity_name=None)
class\_
-mapped class
-
+mapped class (must be a positional argument)
+
ident
primary key, if the key is composite this is a tuple
entity_name
-optional entity name. May be given as a
-positional arg or as a keyword arg.
+optional entity name
identity_key(instance=instance)
instance
object instance (must be given as a keyword arg)
-identity_key(row=row, class=class\_, entity_name=None)
+identity_key(class\_, row=row, entity_name=None)
+class\_
+mapped class (must be a positional argument)
+
row
result proxy row (must be given as a keyword arg)
-
+
+entity_name
+optional entity name (must be given as a keyword arg)
if args:
-kw = {}
-if len(args) == 2:
+if len(args) == 1:
+class_ = args[0]
+try:
+row = kwargs.pop(row)
+except KeyError:
+ident = kwargs.pop(ident)
+entity_name = kwargs.pop(entity_name, None)
+elif len(args) == 2:
class_, ident = args
entity_name = kwargs.pop(entity_name, None)
-assert not kwargs, (unknown keyword arguments: %s
-% (kwargs.keys(),))
+elif len(args) == 3:
+class_, ident, entity_name = args
else:
-assert len(args) == 3, (two or three positional args are
-
[sqlalchemy] Re: [ticket:336] Informix support and some enhancement for oracle and pgsql
unit tests: - cant remove any unit tests - many of them are commented out here I will check the whole unit tests to fix the comment blocks. - changing identifier names in unit tests to handle informix ident - e maybe, would rather informix skip those the old test may shrink the length of table name or field name but this need more work. maybe we can skip them in the beginning. sql.py: - i want to implement ticket #512. the way youre doing it still wont work for selectables that have no engine attached (which is a major use case). ive been thinking a lot about 512 and i will bump it up in priority...i want to release 0.3.6 so that people can start breaking all the latest code, then ill do 512 which is going to change some internals significantly. en, i can fix the informix backend for ticket #512. schema.py: - enable argument - i dont see you using this argument. it seems to be about visitor traversal. the visitor system has been improved in the trunk so you can visit schema items any way you want without any flags on the schema items. definitely dont want to add anything to schema.py our requirment is only using FK in logical level( SA ) not in database level. when setting the enable argument's value to False , the real database will dont create the FK constraint. oracle.py: Added DATE and TIME type support postgres.py: Added SAVEPOINT support. without SAVEPOINT support, the next code cannot run: ::python rs = con.execute( select * from sometable1 ) for row in rs: a = row.a try: con.execute( insert into sometable2( a ) values( %s ) , ( a ) ) except: pass with SAVEPOINT support: ::python rs = con.execute( select * from sometable1 ) for row in rs: a = row.a try: con.execute( insert into sometable2( a ) values( %s ) , ( a ) ) except: if con.engine.name == 'postgres': con.execute( 'rollback to sp' ) -- james.zhang [EMAIL PROTECTED] --~--~-~--~~~---~--~~ 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] [patch] Data partitioning via bind functions
So one thing we're attempting to do in a project I'm working on is partitioning our data set across several databases based on object keys. I've come up with a way to implement this in SQLAlchemy which I've provided here as a mostly-complete patch. Session objects now have a bind_func method which associates a function taking a mapper and an ORM object and returning a Connectable to a mapper. When a transaction is begun on an object associated with that mapper, this function is called in order to work out which Connectable should be used to handle this object. The main area where I haven't been able to work this through completely is in the dependency handling of many-to-many relationships. I'm very interested in what people think of this and whether they feel it's a candidate for inclusion in SQLAlchemy. The patch is against SQLAlchemy 0.3.5. --~--~-~--~~~---~--~~ 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-partitioning.patch Description: Binary data -- Benno Rice [EMAIL PROTECTED] http://jeamland.net/
