Product Management,

This is a list of feature requests from a response by cvh@LE (Chris) to a blog I posted recently. I’ve summarized them before the detailed message.

Must Have:
- EXPLAIN plan for Infobright executor
- ANALYZE plan to test if a query will executed by Infobright or by MYSQL without running the query
- GROUP BY function-result

Nice to Have:
- enable partitioning for MyISAM
- enable more MySQL engines, specifically BLACKHOLE
- improve (singe record) error handling [SREH] in LOAD
- enable SET Functionality in LOAD
- enable CREATE [TEMPORARY] TABLE ... SELECT ...

Hi David,

Thanks for giving us the opportunity to post our wish list.

Here’s mine:

Absolute Requirement:
- Something like EXPLAIN for Infobright executor, or a way to test if a query is
executed by Infobright or by MYSQL before running a query ()

- Group by Function-Result (e.g date-functions)

Nice to have:
- and this one should be relatively easy: enable partitioning for MyISAM in the ICE release

- also enable by default some more engines, specifically BLACKHOLE, then the ICE could serve as a
true working machine only producing reports and posting into BLACKHOLE tables which are replicated to
slaves from where users can make heavily use of them w/o producing load on the ICE ..

- Some Form of Error Handling in LOAD, e.g. Some form of hickup in the data (particularly when migrating large datasets)
are to my experience not uncommon . If LOAD could simply skip and report those rows in a way that
rejected rows wont break the loading but are for instance written to a log table (whole row as text, no colums)
for further processing at a later time
(I am well aware that this is not SQL standard , but you asked for it: ).)

- SET Functionality in LOAD, which allows some minimal form of Transformation during LOAD in SQL

- CREATE [TEMPORARY] TABLE ... SELECT ... for Brighthouse , afaik it only works for MyISAM Engines.
I know this almost touches DML, which is reserved for IEE, but, again, you asked for it..

I am pretty sure I will come up with more soon ..
Author: cvh@LE (Chris)
Date: 02/05/09

David Lutz - 12 February 2009 11:26 AM

Product Management,

This is a list of feature requests [....]

cvh@LE (chris) - 12 February 2009 11:26 AM

- Some Form of Error Handling in LOAD, e.g. Some form of hickup in the data (particularly when migrating large datasets)
are to my experience not uncommon . If LOAD could simply skip and report those rows in a way that
rejected rows wont break the loading but are for instance written to a log table (whole row as text, no colums)
for further processing at a later time

FWIW, Raju (Volomedia) is also asking for this.

Geoffrey

Hi, David & Geoffrey,

I just came up with some more features to my wishlist, I know that the infobright developers are already working on some of the above items and maybe the following are on their also roadmap already.

1.) conditional aggregegations or functional calculations in the form of
  SELECT year,  sum( IF (state=“DE”,sales,null) ) as Sales_Germany,    sum( IF (state=“F”,sales,null) ) as Sales_France ...
    or
  SELECT year, avg( IF (sales>1000,sales,null) ) , .005* sum( IF (sales>1000,sales,null) )  ...

2.) calculating / working with variables inside select in the form of :
  > set @top=0;
  > select @top:=@top+1 TOP, sdate, count(*) sales from abfrage_flug group by sdate order by count(*) desc limit 10;

3.) random subsets / subset of randomly selected items
  While Jabusz and Jakub have already pointed out some ways around this with Infobright, and there is always the way via MyISAM it would be nice to   have a true Infobright implementation to get a random subset of a table/set (randomly selected items) Something in the form of :

    SELECT * from table ORDER BY rand() LIMIT 1000;

Considering the priorities of my requests I would admit that none of these are true must-haves, but nice-to-haves. However, the conditional aggregations of (1) are certainly much-nicer-have if not very-close-to-must-haves since they allow for easy production of pivot tables inside MYSQL without joining derived tables

Chris

Hi Chris!
It looks that all the features you have mentioned above, except variables, are included in the current work plan.

As for variables I think it is a new area, so Product Management should first investigate it.

Cheers,

Although this was not in response to my blog posting, it is nonetheless an ICE feature request:

Victoria, Jakub, Carl & Miriam,

You guys are truly amazing working nights and weekends to answer questions and helping the community using Infobright. I must say I was rather skeptical when I found out you stripped DML capabilities out of ICE (but can we please, please, PLEASE have at least ‘truncate table’?) but it’s a good way of protecting your intellectual property and try people to move to a subscription. I know of a case where the database is simply dropped and rebuilt each night and they are very happy with the performance increase they’re getting over PostgreSQL so you can get around the limitations if you’re willing to invest some time. I still think you need to buy support when you run a product like yours in a production environment but that’s a different discussion. Now regarding TPC-H: it’s too bad I cannot run all queries yet and have to wait a bit for the upcoming releases but when I look at the vigor with which you’re pushing IB forward I’m confident that the wait won’t be long. Thanks again for your support and do try to enjoy what’s left of the weekend!

Best regards, Jos van Dongen
2009-02-07