[sqlalchemy] Re: Problem with using global metadata/connection in a threaded context
Currently, I create a single instance of my persistence layer API, which makes a call to global_connect(), and optionally to default_metadata.create_all() (for tests). I then pass this instance to all the threads that need to talk to my database. global_connect() uses a DynamicMetadata, which is thread-specific: you need to define it each time a new thread is created. Depending on the architecture, you might want to add a func called on each thread (eg. on_start_thread_list within cherrypy,...) which will do the job. It would look like (not tested): import sqlalchemy as sa # blabla def init_thread(): if sa.default_metadata.is_bound(): pass # ok, thread is already init' else: sa.global_connect() # assert sa.default_metadata.is_bound() == True Hope that helps. Cheers Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.net http://www.sirloon.net http://sirbot.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] Re: @property
I know that I could use a mapper extension for this (with before_insert, before_update, etc) but could it not be possible to do it with my @property ? I think when SA apply the mapper (that is, assign), it overrides your property definition. You can ensure that by tracking if your property code is called. When I need to do that, I rename the original property in the mapper: assign_mapper(Bla,blabla,properties=dict(_score=blabla.c.score)) and then deal with _score within the property code. One problem is now, your mapper doesn't have any score attribute, so your selects need to be aware of this and work with _score. Hope that helps cheers Seb -- Sébastien LELONG http://www.sirloon.net sebastien.lelong[at]sirloon.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: how to do many-to-many on the same table?
Hi, Hi, This is causing very messy code though, especially if youa re deleteing an article. Here is now i have it now. The flush is needed, or else i get an AssertionError saying SA is trying to flush 0 into a primary_key. Could you give more details ? I tried using the previous attached code, seems to work nice: c = Article.get(###) c.delete() c.flush() # delete article and its article_related You can find this code in attachment. You may also want to play with backrefs if you just want to have a reference to the relater in the related article (A relates to B, B would have reference to A without saying B relates to A) Hope that helps Cheers, Seb -- Sébastien LELONG http://www.sirloon.net sebastien.lelong[at]sirloon.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 -~--~~~~--~~--~--~--- articles.py Description: application/python
[sqlalchemy] Re: DynamicMetaData question
Hey list, are you confused by the current system ? Please let me know. the only change I would favor here would be to merge connect into MetaData, BoundMetaData and DynamicMetaData stay around for backwards compat for probably forever. Not actually confused by the system. Since I need to define the connection parameter in a seperate file (config file), I *need* the DND.connect() method. So, DND is the only I use... The global_connect() is not useful for me since I need multiple datasources. Again, beeing able to define connection parameter later is crucial, so having a MetaData.connect() would be great. but I need to get at least 20 or 30 users on this list telling me how they are using metadata. A typical initialization of my models looks like: _metadata = DynamicMetaDate(I'll be bound later...) one_table = Table(blabla,_metadata, ...) _engine = None def init_model(dburi,**kwargs): global _engine if _engine: # Engine already defined # check if metadata bound if _metadata.is_bound(): # nothing to do pass else: # need to connect metadata. # occurs when starting new thread _metadata.connect(_engine) else: # first we access the model, init everything _engine = sa.create_engine(dburi,**kwargs) settlement_metadata.connect(_engine) BTW, I'd posted on the TG list to know how they handle multiple datasources with SA. Answer was for now, only one datasource [probably due to global_connect], we'll see next release). So maybe global_connect will not even be used anymore by TG... Cheers Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.net http://www.sirloon.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: DynamicMetaData question
[...] I couldn't use BoundMetaData because I don't know the connection parameters until much after import time, so I am using the only other option I know of, which is DynamicMetaData [...]. I have exactly the same problem. One option is to use global_connect and default_metadata (doc: Using the global Metadata object), but as my apps are essentially multi-sources, I need to define datasources separatly. And I definitly cannot use the BoundMetaData and hardcode the connection parameters... So I'm using the DynamicMetaData and its connect method. Seems to be two independant features in this DynamicMetaData... Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.net http://www.sirloon.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: Ping
I need to ping the Mysql database to keep it alive. Is there a simple way to do this with SA? Maybe you'd want to have a look at the pool_recycle argument of create_engine. It'll make the connection pool beeing checked if active, and potentially recycle the connections if they are too old. Search for MySQL has gone away, you'll have different threads talking about that. Hope it helps. Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.net http://www.sirloon.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: Get error: 'list' object has no attribute 'accept_visitor'
select_columns = [table.c.col1, table.c.col2] statement = table.select(select_columns) You probably want to: select_columns = [table.c.col1, table.c.col2] statement = sqlalchemy.select(select_columns) res = statement.execute().fetchall() # or the like... The table.select form you try to get working is used to specify eg. where clauses, not to specify which columns to get. -- Sébastien LELONG sebastien.lelong[at]sirloon.net On Tuesday 20 February 2007 16:10, vinjvinj wrote: I'm getting the following error when I build the select clause. select_columns = [table.c.col1, table.c.col2] statement = table.select(select_columns) File build\bdist.win32\egg\sqlalchemy\sql.py, line 1351, in select File build\bdist.win32\egg\sqlalchemy\sql.py, line 65, in select File build\bdist.win32\egg\sqlalchemy\sql.py, line 1503, in __init__ File build\bdist.win32\egg\sqlalchemy\sql.py, line 1575, in append_whereclause File build\bdist.win32\egg\sqlalchemy\sql.py, line 1581, in _append_condition AttributeError: 'list' object has no attribute 'accept_visitor' Any idea what this means? VJ --~--~-~--~~~---~--~~ 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: improving connection resiliency
I am wondering if there isn't a better way to handle connection issues than how I am doing it (see the sample below)? You should give a try with SA 0.3.3 (at least), since several connection problems have been solved. See thread MySQL has gone way. Although it talks about MySQL, the bugfix is about the connection pool/object: http://groups.google.com/group/sqlalchemy/browse_thread/thread/9412808e695168ea/93ac328e84206c4b More precisely, also have a look at this thread How to regenerate connections in the QueuePool. You'll find a test case which reconnects/regenerates connections when the database goes down: http://groups.google.com/group/sqlalchemy/browse_thread/thread/f8f5a86ed26d19a7/857f4c70371ea96d FWIW, you'll find a example of retry mechanism which handle database recovering (well, actually, this code is not that clean...). Finally, IIRC, be aware databases exceptions encapulsation into SA's exception has been improved since 0.2.8: some weren't caught resulting in a raise of the original, underlying database exceptions. Hope it helps, Cheers Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.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: new setuptools vs local SA copy
[...] I challenged him to name *any* scenario where an administrator would want a local-environment-based PYTHONPATH to be overridden by an application-wide configuration and he didnt reply to that one. he sees it as a if youre using .eggs, then you must accept that PYTHONPATH only points to installation directories, not runtime directories...so basically breaking PYTHONPATH's documented behavior into something repurposed is by design. I completely agree. I too have spent a lot of time finding why my PYTHONPATH was not considered anymore. Having it overridden is definitly not the expected behavior. -- Sébastien LELONG sebastien.lelong[at]sirloon.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: making crud with assign_mapper
Thanks. I've been using the inspect module but there's such a lot in there it's a bit dizzying! FWIW, you can also use ipython and its dynamic object introspection, completion, etc... This helps a *lot* finding what you want. It's also usefull to dive into the sources, precisely where you want, since you can invoke the edit command on several python object. Give a try with: In [9]: edit mytable.__class__.columns Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.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: list of strings/unicode strings
Is there anyway I can specify the list of tags like this, instead? book.tags = [upython, uprogramming] Well you could use a PickleType column in your book table... Of course, you'll loose functionnalities you could have using a real Tag object. The problem is how you consider a tag in your model... See attachment for the pickletype col. Hope it helps. Cheers, Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.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 -~--~~~~--~~--~--~--- with_pickletype.py Description: application/python
[sqlalchemy] Re: got problems where map to selectable
If you put a refresh just before your assert, that is working: ... user = session.query(User).selectone_by(name='user1') session.refresh(user)# here ! assert user.post_count==2, 'user.post_count is %s'%user.post_count Not sure to understand why though... I bet its because of SA's cache (identity_map). Since the user was already loaded before, SA doesn't re-perform sql access to the db. You can see it by activating engine.echo = True, and selecting users by id (eg. session.query(User).get(1)) Hope it helps. Cheers, Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.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: access mapper properties
In [1]: oo=Tariffa().select(limit=1)[0] In [2]: for jjj in oo.c: print jjj.name codice aliquota_iva aliquota_enpav centro_costo cod_funzione_calcolo unita_misura oo.c will iterate over the attributes of the mapped table. You can iterate over all attributes, including relations defined in mappers, with: for one_attr in oo.mapper.props.keys(): print getattr(oo,one_attr) Not sure however if that's the correct way... For mapper only properties, you canget them with: mapper_props = oo.mapper.properties.keys() Actually, oo.mapper.props.keys() is the (almost) concatenation of table + mapper properties since: set(sorted(cl.c.keys() + cl.mapper.properties.keys())) == set(sorted(cl.mapper.props.keys())) = True Hope it helps. Cheers, Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.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: Inspect and __init__ of a mapped class
btw if u find a way to obtain some externaly-bound non-global reference from func's code (see what dis.dis() shows), DO mail... Bot sure to understand precisely what you want to do, but it seems like monkey patching: class Klass(object): def incredible_func(self): a = 2 print a: %s % a o = Klass() o.incredible_func() # a: 2 def awesome_func(self): b = 17 print b: %s % b Klass.incredible_func = awesome_func another_o = Klass() o.incredible_func() # b: 17 It implies rewriting the whole body func, not just find/replace as in your example. Maybe that's what you want to avoid... Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.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: ORM ramblings 2 - and announcing some o2r wrapper
Having a sqlalchemy.__version__ and revision (or whatever named vars) is definitly usefull, even more when there are frequent project releases such as with SA. +1 for this feature... Maybe just add one simple .revision as of svn's $Revision$, that should be okay for flying on devel. But the file where you put this $Revision$ property has to be changed on every commit so the property is updated, no ? If so, this would give wrong revision number... but u'll need to setup a svn autoprop for that, add . ~/.subversion [miscellany] enable-auto-props = yes [auto-props] *.py = svn:keywords=Id Revision You can also: svn propset svn:keywords Revision the_file_you_want_to_tag avoiding the autoprop for all files. Cheers, Seb -- Sébastien LELONG [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] Row attributes from a table with renamed columns
Dear SQLAlchemists,
I'm trying to fetch some specific row attributes from a select statement on a
table. So far so good, this shouldn't be a problem... except this table
declares some renamed columns, using key, as:
sa.Column('the_label', sa.String(30), key='label')
When I'm inserting rows, the insert statement uses the renamed column
(label):
one_table.insert().execute(id=0,label='oula')
But when I'm selecting results and fetch some attributes on a given row, only
the original column names remain:
one_row = one_table.select().execute().fetchall()[0]
print attr = %s % one_row .label # error !
print orig attr = %s % one_row the_label # ok it works...
Is there any way to get the renamed columns on a RowProxy object ?
As attachment, a test case:
$ python table_with_key.py
Keys in row: ['id', 'the_label']
r.the_label: blabla
Traceback (most recent call last):
File table_with_key.py, line 25, in ?
print r.label: %s % r.label # won' work
File build/bdist.linux-i686/egg/sqlalchemy/engine/base.py, line 730, in
__getattr__
AttributeError: label
Thanks Cheers
Seb
--
Sébastien LELONG
sebastien.lelong[at]sirloon.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
-~--~~~~--~~--~--~---
table_with_key.py
Description: application/python
[sqlalchemy] Re: count(*) function
I have tried select([func.count(*)], from_obj=[table_name]).execute() but it didn't work I think you should try to specify a column in your count or leave it empty (didn't try). If you're using mapped objects, you can use the SelectResults extension: from sqlalchemy.ext.selectresults import SelectResults SelectResults(your_session.query(YourMappedClass)).count() Cheers, Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.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: Performance of data mapping and plain access
one thing that could make ORM loads much faster would be if you knew the objects would not need to be flushed() at a later point, and you disabled history tracking on those instances. this would prevent the need to create a copy of the object's attributes at load time. This reminds me a functionality I was looking for a few weeks ago. I'd liked to be able to load read-only objects from a database. Read-only objects were (are still) useful for me when retrieving data from a precious database (production db), ensuring no update/insert/delete operations would occur. Another scenario is an app which can actually do these operations, and another app which musn't. While this can be done using grants, it could be useful (and easier) if things could be done at the ORM level. Is there any other way to get those read-only objects, without implementing this functionality ? Cheers, -- Sébastien LELONG sebastien.lelong[at]sirloon.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: Performance of data mapping and plain access
[...] simpliest thing would be to use a Session that has flush() overidden. or an engine that overrides execute() to check for INSERT/UPDATE/DELETE statements and throws an error [...] I tried the ReadOnlySession class which overrides the flush() func. Works like a charm, this adds a security net in my apps as I cannot actually even add any grants, so this is very useful for me. Thanks. Moreover, in the 2nd scenario, if app #1 can insert rows and app #2 can't, this means there must be different db users (per app) with specific grants for each, which adds complexity and may results in an evil maintainance... Cheers, -- Sébastien LELONG sebastien.lelong[at]sirloon.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: SA 0.3.x: performance issues = proposal
As attachement, here's the patch (against rev 2132). It's local to sqlalchemy/orm/, so: cd /sqlalchemy/orm patch -p0 attribute_cache.patch About the patch itself: 1. the cache should be a WeakKeyDictionary OK, done. Performances are still OK, differences between built-in dict are negligible. 2. the cache needs to be cleared for a particular cache (class ?) If I understand right, while calling register_attribute(), the cache will be cleared for the given class, if this one has been cached. Is there anywhere else the cache needs to cleared ? 3. the cache should *probably* be at the module level, and not within the AttributeManager itself; SA uses only a single AttributeManager per application so it probably doesnt matter I let the cache within the AttributeManager... Things can be moved at the module level, but if it's not necessary right now, it might not be a good idea to do it right now (when it will be necessary, the design may have been changed and modifications not valid anymore). 4. the raise should use an exception class of some kind I've made it raise a TypeError... Yes, string based exceptions are very a bad thing ! Attributes are cached while using the managed_attributes() *and* noninherited_managed_attributes() funcs. Those are very similar. I hesitated refactor them, but centralize cache management is probably a good thing. This is the purpose of the second patch, which include this refactoring. As expected, performances decrease. Here's the traditionnal benchmark results: SA 0.3.1, rev 2127: total time 4.29376721382 real0m5.420s user0m4.088s sys 0m0.108s SA with attr. cache, no refactoring (~2X faster, back closed to 0.2.8 speed): total time 2.34819602966 real0m3.013s user0m2.344s sys 0m0.088s SA with attr. cache, with refactoring (30% slower than without refactoring) total time 3.05679416656 real0m3.747s user0m2.988s sys 0m0.068s It's up to you choosing the patch ! IMO, I'm *not* in favor to use refactoring in this case :). Finally, I've put a clear_attribute_cache func which, ... clear the attribute cache. While client code may not have to worry about caching, it may need to clear it... thanks much, this is the kind of user involvement i like. Well, you're very welcome ! Glad to help ! Cheers, Seb -- Sébastien LELONG sebastien.lelong[at]sirloon.net P.S: im beginning to suspect that yield introduces overhead vs. a straight list ? (python interpreter storing stack frames ? dunno). dunnotoo :) --~--~-~--~~~---~--~~ 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: attributes.py === --- attributes.py (révision 2132) +++ attributes.py (copie de travail) @@ -609,8 +609,18 @@ class AttributeManager(object): allows the instrumentation of object attributes. AttributeManager is stateless, but can be -overridden by subclasses to redefine some of its factory operations. +overridden by subclasses to redefine some of its factory operations. Also be aware AttributeManager +will cache attributes for a given class, allowing not to determine those for each objects (used +in managed_attributes() and noninherited_managed_attributes()). This cache is cleared for a given class +while calling register_attribute(), and can be cleared using clear_attribute_cache() +def __init__(self): +# will cache attributes, indexed by class objects +self._attribute_cache = weakref.WeakKeyDictionary() + +def clear_attribute_cache(self): +self._attribute_cache.clear() + def rollback(self, *obj): retrieves the committed history for each object in the given list, and rolls back the attributes each instance to their original value. @@ -639,21 +649,31 @@ def managed_attributes(self, class_): returns an iterator of all InstrumentedAttribute objects associated with the given class. +if self._attribute_cache.has_key(class_): +return self._attribute_cache[class_] + +self._attribute_cache[class_] = [] if not isinstance(class_, type): -raise repr(class_) + is not a type +raise TypeError(repr(class_) + is not a type) for key in dir(class_): value = getattr(class_, key, None) if isinstance(value
[sqlalchemy] Re: MySQL Has Gone Away
I'm still having this problem with MySQL (both 4.1 and 5.0), SA (tested with both 0.2.8 and 0.3.1) and TurboGears 1.0b1. Almost the same for me (MySQL 5.0.22, SA 0.2.8 and 0.3.1, using CherryPy). Using pool_recycle works nice on a non-high load environments. But while I benchmark my app (using siege or ab), this always produces Lost connection during MySQL query..., and sometimes MySQL server has gone away Note the version of MySQLdb is important: using 1.2.2b1 (didn't tested the last 1.2.2b2) can reduce those errors, but not completely though. Also, MySQLdb is not thread-safe, so using it in a multithreaded environment can be painful... (but I'm sure that's transparent using TG). Since SA 0.2.4 (pool_recycle did not exist at this time), I use a specific pool, derived from the QueuePool class, which *always* check the current checked-out connection is valid, using connection.ping() (MySQL specific, I think not all DBs support this). I'm still using my pool as of 0.3.1, to prevent those kind of errors. I didn't find out why this occurs: having a test-case is very difficult. I'd also investigated the way the app access MySQL. IIRC, while using a socket, if a connection is timed-out, reaccessing the server automatically regenerate the connection (well, actually, I've not observed this behavior for all the mysql servers I use), without producing those errors. I have this kind of lines in MySQL logs when this occurs: Connect [EMAIL PROTECTED] on dev blablabla Well, that's a lot of obscure observations... without a clean solution ! But maybe someone will have some ideas... Cheers, -- Sébastien LELONG sebastien.lelong[at]sirloon.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 -~--~~~~--~~--~--~---
