Any idea what’s wrong?

active is set to true on all rows in that table

Can you show a traceback?

Should it not be FacebookPage.select()

without the ()

LOL yes

I feel silly 😄

Happens 😅

existing_facebook_post = FacebookPost.get(lambda p: p.facebook_id == post_id and p.facebook_page == self.facebook_page)

This query works most of the time

Sometimes, it gives AssertionError:

I have updated Pony to 0.7.13

Any ideas?

stacktrace: https://gist.github.com/matthewrobertbell/595368323a17496d010c7c0894c7c4af

The query works fine with that exact data when I run it on its own

The same error if I do a select query rather than get

Hi Matthew! This is indeed a bug. I pushed the fix on GitHub, you can check if it works for you https://github.com/ponyorm/pony/commit/09ed56fa2ed36e5deae1cf1aeb17bdecc5191816

Sorry but I want to do the same thing for their total. But I get an error (AssertionError) The only place I changed is this: @property def point(self): return coalesce(select(c.point for c in self.comments).sum(), 0.0)

Hi Muhammed! I fixed this bug and pushed the fix on GitHub, will be included in the next release

https://github.com/ponyorm/pony/commit/bcfec2cc50d5725a54643610453577e14e4c3d8e

https://github.com/ponyorm/pony/commit/2f635b667031ab374adbedc4834dccdc95dc65d7

https://github.com/ponyorm/pony/commit/ed379264afb23e6882ea6b030ea5ef389e6e5a36

I've the following class: class PatientCollection: def __init__(self, user): self.user = user def query(self): return Patient.select(user=self.user)[:] When testing it I get the error > return Patient.select(user=self.user)[:] E TypeError: select() got an unexpected keyword argument 'user' When using def query(self): return Patient.select(lambda p: p.user == self.user)[:] Everything is fine. Is this a bug or something wrong in my code?

Now Entity.select(...) method accepts kwargs, as well as obj.collection.select(...) https://github.com/ponyorm/pony/commit/0910f768629da3b37fdd0fa773c168d98037eb05

Hi Muhammed! I fixed this bug and pushed the fix on GitHub, will be included in the next release

Thanks Alexander :)

Hi Matthew! This is indeed a bug. I pushed the fix on GitHub, you can check if it works for you https://github.com/ponyorm/pony/commit/09ed56fa2ed36e5deae1cf1aeb17bdecc5191816

Thank you so much! I will check this morning

@metaprogrammer with pony installed from github master I get the same AssertionError, it looks like the exact same stack trace

I also tried pip install git+https://github.com/ponyorm/pony.git@orm to install from orm branch but the same error occurs

Hmm, maybe I was too fast. I fixed very similar problem just before your report, and thought it is the same. Let me check...

Hi @matthewrobertbell, can you check again? It should work now

With git orm branch? I will when I get back to my computer

yes, orm branch

thanks

This worked, thank you very much! Do you plan a new release soon which would include recent bug fixes?

Yes, I hope we will make release on this week

Thank you, please let the channel know when you do 🙂

Sure!

What was the problem which caused the bug?

During query execution Pony creates so-called PefetchContext which stores details about additional data that should be fetched along with this query. Until now Pony assumed that it is not necessary to create new PrefetchContext while there is another active PrefetchContext, and checked this using assert Also when Pony prepares execution of a query it performs flush() to save all modified objects before query execution to be sure that the query sees all modifications that was done in current db_session Performing flush() causes hooks on modified objects to be called, such as before_insert. If inside before_insert you make another query, it can lead to situation when we need to perform select during preparation of another select: top-level select creating PrefetchContext performing flush saving previously created object calling before_insert nested select creating nested PrefetchContext In previous tests for before_insert we didn't check the situation when before_insert was called inside a flush caused by select, and this before_insert executes another select as well So turns out we actually need to manage a stack of PrefetchContexts. I added this stack and it fixes the problem

cool, good to know

This is greatest thing! In my application I’ve implemented my own powerful “events” layer based on before/after hooks. Sometimes on some complex scenarios I ran into assertion errors related to prefetch context. To resolve these error I made some refactoring and workarounds. Thank you!!!

Now Entity.select(...) method accepts kwargs, as well as obj.collection.select(...) https://github.com/ponyorm/pony/commit/0910f768629da3b37fdd0fa773c168d98037eb05

Great Alexander!

Hi, i'm checking reconnections forcing a database shutdown. but On a pny.exists() query, and i'm getting a : " psycopg2.OperationalError: could not connect to server: Connection refused Is the server running on host "127.0.0.1" and accepting TCP/IP connections on port 19999? " This qquery is the first step in a long runing process(reading kafka topic) raising this exception, break de flow and read a new msg so... i lost the control... i'm expecting to keep the flow in a neverending loop tring to reconect... is any way to achieve that?

Hi, if this is the first step, and you expect that connection will work after some time, you can write: while True: try: pny.exists(some_test_sql) except OperationalError: rollback() else: break

but... is something that could be articulate on a general way

as... i have a lot of db access in per msg/flow

and i guess that i'm getting this exception because is the first query on the flow

on my case i need to freeze the flow when i detect a db disruption

If I understand the use-case correctly, you can wrap this code into some function and use it as a first step of the flow def wait_for_connection(): while True: try: pny.exists(some_test_sql) except OperationalError: rollback() else: break def process_message(): wait_for_connection() ... # start doing other stuff msg = read_message() process_message(msg)

If the goal is to wait some time until the server is up again

The use case could be summarize as a long running job doing: Read Msg from Queue -> Per msg make different operations(including 6/7 db calls) -> ack msg

my best guest could be freeze the operation flow when i detect a db problem to resume later(when db up)

I am getting this fairly frequently in my background workers, is this a bug or a problem with my code?

I am using db_session(strict=True)

This is also happening without strict=True. Could it be another bug?

Can I update a table's structure without deleting the table from the database and create it again? I tried adding the field on the class but it doesn't work :(

At this moment you need to manually perform ALTERE TABLE command

This is also happening without strict=True. Could it be another bug?

It is possible. Can you give more details so I can reproduce it?

At this moment you need to manually perform ALTERE TABLE command

Ok thank you

What other details would be helpful?

Can you comment out this assertion and look what happens after that?

What other details would be helpful?

Also, did you have some exceptions in your hooks?

if you mean things like before_insert, I don’t think so, but it is hard to tell from sentry

the issue is this happens very rarely

so it is hard to investigate

maybe you can add try/except to before_insert with some error reporting

If before_insert failed, wouldn’t it show up in the exception?

Currently it should

If you dont' catch it somewhere

In general, what should Pony do if during flush() it need to save 10 objects, and in before_insert the second object throws an exception? Should it suppress the exception and continue saving the rest 8 objects? Or should it immediately abort the current transaction?

abort I think

So, when before_insert hook of any object throws an exception, Pony should switch db_session's state to "invalid". Even if the user code catch the exception and try to continue current db_session, Pony will refuse to do this

Is that what happens now or what you want to happen?

What I want to happen. Right now I think it stop saving unsaved objects and raises an exception. And if a user code catch this exception and continue working in current db_session it should get the assert that you have get

Ok, so I’ll try and look to see if there is an exception being caught in my code somewhere

Hi, how can I combine more than sets? For example category1.products + category2.products and .products = Set(Product) Can I use + sign?

try set(cat1.products) | set(cat2.products)

if you want products in either

(that’s not in a query)

I use sort_by, filter_by etc. after merging