Large BTreeFolder2 batching/pagination

View: New views

7 Messages — Rating Filter: Alert me

Large BTreeFolder2 batching/pagination

by Ken Ara :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

We have some BTreeFolder2's containing hundreds of thousands of objects, even one containing over 2 million instances of Folder. That is really no problem in itself.

Under our setup, beyond 300-400,000 contained objects, the batching provided by getBatchObjectListing becomes unusable. I've seen this problem mentioned somewhere but never any hint of a solution.

We have a new application that will require pagination, or ideally, the ability to go forward and back through the items (any order). Is there a way to refer to the contained items by something like offset? It might be easy if the objects had sequential, consecutive numeric ids, but sadly this cannot be the case.

We are using the BTreeFolder2-1.0.2 version that came with Zope 2.10.

Thanks.

Ken

_______________________________________________
Zope maillist - Zope@...
https://mail.zope.org/mailman/listinfo/zope
** No cross posts or HTML encoding! **
(Related lists -
https://mail.zope.org/mailman/listinfo/zope-announce
https://mail.zope.org/mailman/listinfo/zope-dev )

Re: Large BTreeFolder2 batching/pagination

by Andreas Jung-5 :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

What kind of practical sense does it make to batch 400k objects? I can no imagine single usecase where one
would be interested in walking through such an amount of objects manually. Better organize your data in a more
handy way or implement some search logic for bringing the batch size to a number of relevant object.s

-aj

On Mon, Nov 2, 2009 at 00:04, Ken Ara <feedreader@...> wrote:

Under our setup, beyond 300-400,000 contained objects, the batching provided by getBatchObjectListing becomes unusable. I've seen this problem mentioned somewhere but never any hint of a solution.

Re: Large BTreeFolder2 batching/pagination

by Ken Ara :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

I agree it is hard to imagine, but I am just the web guy...

If I dare to guess, I would say that what they want to do is improve access to the information for the users. The two million objects are strongly interlinked, so this is the navigation system used up to now. But why not open it up to browsing?

For the content team, it might also be nice to use the ZMI or similar when working with this content.

We have not been successful to create a Catalog of this many objects. The process seemed to time out after many hours, probably hardware-bound. Also, the size of the resulting ZODB is of concern, but we may try again with the Catalog on a mounted database.

My general question remains: Is there a way to address the objects contained in a BTreeFolder2 using something like an 'offset' or other identifier? Has anyone found a strategy that scales better than getBatchObjectListing?

Thanks

Ken

--- On Mon, 11/2/09, Andreas Jung <lists@...> wrote:

From: Andreas Jung <lists@...>
Subject: Re: [Zope] Large BTreeFolder2 batching/pagination
To: "Ken Ara" <feedreader@...>
Cc: zope@...
Date: Monday, November 2, 2009, 6:23 AM

What kind of practical sense does it make to batch 400k objects? I can no imagine single usecase where one
would be interested in walking through such an amount of objects manually. Better organize your data in a more
handy way or implement some search logic for bringing the batch size to a number of relevant object.s

-aj

On Mon, Nov 2, 2009 at 00:04, Ken Ara <feedreader@...> wrote:

Under our setup, beyond 300-400,000 contained objects, the batching provided by getBatchObjectListing becomes unusable. I've seen this problem mentioned somewhere but never any hint of a solution.

Re: Large BTreeFolder2 batching/pagination

by Andreas Jung-5 :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

Am 02.11.09 11:28, schrieb Ken Ara:
> I agree it is hard to imagine, but I am just the web guy...
>
> If I dare to guess, I would say that what they want to do is improve
> access to the information for the users. The two million objects are
> strongly interlinked, so this is the navigation system used up to now.
> But why not open it up to browsing?
>

Because no human can deal in a reasonable way with 500k object..build
your *own*
custom and working navigation throughout the data records...presenting
the ZMI view for BTreefolders to a human is just sick.
>
> For the content team, it might also be nice to use the ZMI or similar
> when working with this content.
>
> We have not been successful to create a Catalog of this many objects.
> The process seemed to time out after many hours, probably
> hardware-bound. Also, the size of the resulting ZODB is of concern,
> but we may try again with the Catalog on a mounted database.
>
we have ZCatalogs with millions of objects..likely you are indexing
everything within one big transaction
without using savepoints or subtransactions..

>
-aj

[lists.vcf]

begin:vcard fn:Andreas Jung n:Jung;Andreas org:ZOPYX Ltd. & Co. KG adr;quoted-printable:;;Charlottenstr. 37/1;T=C3=BCbingen;;72070;Germany email;internet:info@... title:CEO tel;work:+49-7071-793376 tel;fax:+49-7071-7936840 tel;home:+49-7071-793257 x-mozilla-html:FALSE url:www.zopyx.com version:2.1 end:vcard

Re: Large BTreeFolder2 batching/pagination

by Ken Ara :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

I should also mention that one of these large collections contains objects with sequential, but not consecutive, numeric ids (timestrings) like some blogs. Users of this database can benefit from the ability to access items added around the same time.

With or without a catalog, how would one get the next or previous id in such a case?

--- On Mon, 11/2/09, Andreas Jung <lists@...> wrote:

From: Andreas Jung <lists@...>
Subject: Re: [Zope] Large BTreeFolder2 batching/pagination
To: "Ken Ara" <feedreader@...>
Cc: zope@...
Date: Monday, November 2, 2009, 11:36 AM

Am 02.11.09 11:28, schrieb Ken Ara:
> I agree it is hard to imagine, but I am just the web guy...
>
> If I dare to guess, I would say that what they want to do is improve
> access to the information for the users. The two million objects are
> strongly interlinked, so this is the navigation system used up to now.
> But why not open it up to browsing?
>

Because no human can deal in a reasonable way with 500k object..build
your *own*
custom and working navigation throughout the data records...presenting
the ZMI view for BTreefolders to a human is just sick.
>
> For the content team, it might also be nice to use the ZMI or similar
> when working with this content.
>
> We have not been successful to create a Catalog of this many objects.
> The process seemed to time out after many hours, probably
> hardware-bound. Also, the size of the resulting ZODB is of concern,
> but we may try again with the Catalog on a mounted database.
>
we have ZCatalogs with millions of objects..likely you are indexing
everything within one big transaction
without using savepoints or subtransactions..

>
-aj

Re: Large BTreeFolder2 batching/pagination

by Ken Ara :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

I take your suggestion to find a Catalog-based solution. I have set the subtransactions very low (100).

But I have got the following error:

Traceback (innermost last):
Module ZPublisher.Publish, line 119, in publish
Module ZPublisher.mapply, line 88, in mapply
Module ZPublisher.Publish, line 42, in call_object
Module Products.ZCatalog.ZCatalog, line 348, in manage_catalogFoundItems
Module Products.ZCatalog.ZCatalog, line 800, in ZopeFindAndApply
Module Products.ZCatalog.ZCatalog, line 559, in catalog_object
Module transaction._manager, line 110, in savepoint
Module transaction._transaction, line 312, in savepoint
Module transaction._transaction, line 309, in savepoint
Module transaction._transaction, line 737, in __init__
Module ZODB.Connection, line 1034, in savepoint
Module ZODB.Connection, line 543, in _commit
Module ZODB.Connection, line 571, in _store_objects
Module ZODB.Connection, line 1168, in store
IOError: [Errno 28] No space left on device

I am guessing that this process is using the /tmp directory which unfortunately was incorrectly sized at only 1.9Gb. Can someone confirm this?

Then either I must find a way to change the /tmp size, or is there a solution through Zope?

Thank you.

Ken

--- On Mon, 11/2/09, Andreas Jung <lists@...> wrote:
>
> We have not been successful to create a Catalog of this many objects.
> The process seemed to time out after many hours, probably
> hardware-bound. Also, the size of the resulting ZODB is of concern,
> but we may try again with the Catalog on a mounted database.
>
we have ZCatalogs with millions of objects..likely you are indexing
everything within one big transaction
without using savepoints or subtransactions..

>
-aj

_______________________________________________
Zope maillist - Zope@...
https://mail.zope.org/mailman/listinfo/zope
** No cross posts or HTML encoding! **
(Related lists -
https://mail.zope.org/mailman/listinfo/zope-announce
https://mail.zope.org/mailman/listinfo/zope-dev )

Re: Large BTreeFolder2 batching/pagination

by Andreas Jung-5 :: Rate this Message:

Reply to Author | View Threaded | Show Only this Message

Am 08.11.09 13:17, schrieb Ken Ara:

> I take your suggestion to find a Catalog-based solution. I have set the subtransactions very low (100).
>
> But I have got the following error:
>
> Traceback (innermost last):
> Module ZPublisher.Publish, line 119, in publish
> Module ZPublisher.mapply, line 88, in mapply
> Module ZPublisher.Publish, line 42, in call_object
> Module Products.ZCatalog.ZCatalog, line 348, in manage_catalogFoundItems
> Module Products.ZCatalog.ZCatalog, line 800, in ZopeFindAndApply
> Module Products.ZCatalog.ZCatalog, line 559, in catalog_object
> Module transaction._manager, line 110, in savepoint
> Module transaction._transaction, line 312, in savepoint
> Module transaction._transaction, line 309, in savepoint
> Module transaction._transaction, line 737, in __init__
> Module ZODB.Connection, line 1034, in savepoint
> Module ZODB.Connection, line 543, in _commit
> Module ZODB.Connection, line 571, in _store_objects
> Module ZODB.Connection, line 1168, in store
> IOError: [Errno 28] No space left on device
>
> I am guessing that this process is using the /tmp directory which unfortunately was incorrectly sized at only 1.9Gb. Can someone confirm this?
>
> Then either I must find a way to change the /tmp size, or is there a solution through Zope?

export TMPDIR= ...

-aj

[lists.vcf]