R help - Oct 2010 - merging and working with BIG data sets. Is sqldf the best way??

Hi everyone,



I’m working with some very big datasets (each dataset has 11 million rows
and 2 columns). My first step is to merge all my individual data sets
together (I have about 20)



I’m using the following command from sqldf

               data1 <- sqldf("select A.*, B.* from A inner join B
using(ID)")



But it’s taking A VERY VERY LONG TIME to merge just 2 of the datasets (well
over 2 hours, possibly longer since it’s still going).





I was wondering if anyone could suggest a better way, or maybe some
suggestions on how I could tweak my computer set up to speed it up?





I’ve looked at the following packages and this is the only way I’ve found to
actually merge large data sets in R. These packages seem great for accessing
large data sets by avoiding storing them in RAM….but I can’t see how they
can be used to merge data sets together:

·        ff

·        filehash

·        bigmemory



Does anyone have any ideas?



At the moment my best idea is to hand it over to someone with a dedicated
database server and get them to do the merges (and then hope package biglm
can do the modelling)



Thanks for any ideas at all!!







Chris Howden

Founding Partner

Tricky Solutions

Tricky Solutions 4 Tricky Problems

Evidence Based Strategic Development, IP development, Data Analysis,
Modelling, and Training

(mobile) 0410 689 945

(fax / office) (+618) 8952 7878

chris@trickysolutions.com.au

	[[alternative HTML version deleted]]

Chris Howden wrote:
> Hi everyone,
>
> I?m working with some very big datasets (each dataset has 11 million rows
> and 2 columns). My first step is to merge all my individual data sets
> together (I have about 20)
>
> I?m using the following command from sqldf
>                data1 <- sqldf("select A.*, B.* from A inner join B
> using(ID)")
>
> But it?s taking A VERY VERY LONG TIME to merge just 2 of the datasets (well
> over 2 hours, possibly longer since it?s still going).
>
> I was wondering if anyone could suggest a better way, or maybe some
> suggestions on how I could tweak my computer set up to speed it up?
>
> I?ve looked at the following packages and this is the only way I?ve found
to
> actually merge large data sets in R. These packages seem great for
accessing
> large data sets by avoiding storing them in RAM?.but I can?t see how they
> can be used to merge data sets together:
>
> ?        ff
> ?        filehash
> ?        bigmemory
>
> Does anyone have any ideas?
>
> At the moment my best idea is to hand it over to someone with a dedicated
> database server and get them to do the merges (and then hope package biglm
> can do the modelling)
>   
SQLite is not optimized for large datasets. If you don't have
familiarity with index generation in SQL databases, then you
may well be best served by finding someone who has such
experience.

If the data are already sorted by ID, it may be feasible to do the
merge with a Perl or Python script, processing records
from two files at a time incrementally.

Good luck!
>
> Thanks for any ideas at all!!
>
>
> Chris Howden
> Founding Partner
> Tricky Solutions
> Tricky Solutions 4 Tricky Problems
> Evidence Based Strategic Development, IP development, Data Analysis,
> Modelling, and Training
> (mobile) 0410 689 945
> (fax / office) (+618) 8952 7878
> chris at trickysolutions.com.au
>
>

On Tue, Oct 12, 2010 at 2:39 AM, Chris Howden
<chris at trickysolutions.com.au> wrote:
> I?m working with some very big datasets (each dataset has 11 million rows
> and 2 columns). My first step is to merge all my individual data sets
> together (I have about 20)
>
> I?m using the following command from sqldf
>
> ? ? ? ? ? ? ? data1 <- sqldf("select A.*, B.* from A inner join B
> using(ID)")
>
> But it?s taking A VERY VERY LONG TIME to merge just 2 of the datasets (well
> over 2 hours, possibly longer since it?s still going).
You need to add indexes to your tables.   See example 4i on the sqldf home page
http://sqldf.googlecode.com
This can result in huge speedups for large tables.

-- 
Statistics & Software Consulting
GKX Group, GKX Associates Inc.
tel: 1-877-GKX-GROUP
email: ggrothendieck at gmail.com