Hi David:

Would you be interested in describing for us all - here on the list - where some of that code is and how an implementation was arrived at to "properly handle millions or even billions of rows" in a target table. I, for one would be very interested in knowing just where that code might be and how it was tested.

Regards,
Ruth

David E Jones wrote:
I wouldn't say that OFBiz OOTB is a middle of the road compromise, that would 
imply consistency in the code which is not the case. The fact is some code is 
designed and tested to properly handle millions or even billions of rows in its 
target table, and other code can only have a few thousand or perhaps tens of 
thousands. There is no general rule or easy way to simplify that down. Things 
that are used more by larger organization have tended to be written or improved 
over time to handle more data. Things that aren't haven't, it's that simple.

-David


On Aug 17, 2010, at 10:19 AM, Ruth Hoffman wrote:

Hi David:
Thanks for chiming in! I understand and my experience has been exactly as you say: 
"Each bit of code and each query needs to be designed with a certain expected data 
set size"

What you get with OFBiz out-of-the-box is a middle of the road "compromise". 
The really good news is that OFBiz is flexible enough to handle any necessary tweeks.

FYI - I wanted to do more exploration into testing with large data sets and try and 
arrive at some guidelines to help determine just how those "tweeks" might work. 
I had intended to document the same, but then my database went away and I turned my 
attention to other things.

Regards,
Ruth

David E Jones wrote:
One thing to keep in mind is that with database queries and OFBiz there is no 
inherent nature of how it handles large data sets. Each bit of code and each 
query needs to be designed with a certain expected data set size, and typically 
the more data you want to allow for in the data store the more effort and 
testing it requires.

In other words, a query and bit of code that doesn't handle a certain large 
table means absolutely nothing about how other queries and other code will 
handle similarly large tables.

-David


On Aug 17, 2010, at 9:33 AM, Ruth Hoffman wrote:

Hi Micha:

Could I ask how big are your data sets? The reason I ask is that several years 
ago, prior to 9.04 I had an out-of-memory problem that was never resolved. The 
database tables that I was querying had between 4 - 12 million records. My 
database (and OFBiz usage of the same) went away before I couldn't do any 
further investigation into the problem.

So, I'm still curious about OFBiz support for really large data sets.

Regards,
Ruth

Michał Cukierman wrote:
Thanks for the quick answer.

I have allocated 2GB of memory, so it's not a case. The problem is that
I work on huge data sets. Version is 10.4 but 9.04 is has the same code.
I have ommited those informations 'couse as you see - it's not relevant
to the topic.

I don't understand your last assumption. Maybe it will someting in my
understanding?

For now I think the best would be to improve the code I have proposed as
it fixes:
1) My issues
2) Coding way of selecting everything to get a couple of categories.

So, thats why I asked about:

1) How to implement Right/Left joins on DynamicViewEntity
(I was able to find out how to creat inner join only)
2) How to implement my whereString
(Haven't found in a code such a use case)


Regards,
Michał



Dnia 2010-08-17, wto o godzinie 07:54 -0700, BJ Freeman pisze:
as far as performance how much memory have you allocated to ofbiz?
what version of ofbiz are you using?
There is a top Category that should be checked for first.
have not been in that part of the code for a while.

Michał Cukierman sent the following on 8/17/2010 7:41 AM:
Hi,

I have a problem with performence (and Out of Memmory Issue) with
catalog management module. There is a section invoked after
Choose Top Category

Collection<GenericValue>  allCategories =
delegator.findList("ProductCategory", null, null, null, null, false);
for (GenericValue curCat: allCategories) {
        Collection<GenericValue>  parentCats =
curCat.getRelatedCache("CurrentProductCategoryRollup");

if (parentCats.isEmpty())
        results.add(curCat);
}

As we see. To get categories without parents we need to fetch for all
categories and for each of them, find a parent.

I do not know Ofbiz api very well for now, so I have implemented:
EntityWhereString where =
EntityCondition.makeConditionWhere("PRODUCT_CATEGORY.PRODUCT_CATEGORY_ID
not in" +
" (select PRODUCT_CATEGORY_ROLLUP.PRODUCT_CATEGORY_ID from
PRODUCT_CATEGORY_ROLLUP)");
Set<String>  fieldsToSelect = FastSet.newInstance();
                        fieldsToSelect.add("productCategoryId");
                        results.addAll(delegator.findList("ProductCategory", 
where,
fieldsToSelect, null, null, false));


I simply do invoke a SQL query in the database:

select PRODUCT_CATEGORY.PRODUCT_CATEGORY_ID from PRODUCT_CATEGORY where
PRODUCT_CATEGORY.PRODUCT_CATEGORY_ID not in
(select PRODUCT_CATEGORY_ROLLUP.PRODUCT_CATEGORY_ID from
PRODUCT_CATEGORY_ROLLUP);

Is there a way ( for sure is ) to implement it in a better manner? For
now it works on my system, but after improving the query it could be
passed to a production.

So I got two questions in fact:

1) How to implement Right/Left joins on DynamicViewEntity
2) How to implement my whereString


Thank you in advance,




Reply via email to