On 01/05/14 19:50, Seb wrote:
> Hello,
>
> I've been looking for a way to write a table into multiple files, and am
> wondering if there are some clever suggestions. Say we have a table
> that is too large (several Gb) to write to a file that can be used for
> further analyses in other languages. The table consists of a timestamp
> field and several numeric fields, with records every 10th of a second.
> It could be meaningfully broken down into subsets of say 20 minutes
> worth of records. One option is to write a shell script that loops
> through the timestamp, selects the corresponding subset of the table,
> and writes it as a unique file. However, this would be extremely slow
> because each select takes several hours, and there can be hundreds of
> subsets. Is there a better way?
# copy (select * from generate_series(1,1000)) to program 'split -l 100
- /tmp/xxx';
COPY 1000
# \q
$ ls -l /tmp/xxxa*
-rw------- 1 postgres postgres 292 May 1 19:08 /tmp/xxxaa
-rw------- 1 postgres postgres 400 May 1 19:08 /tmp/xxxab
-rw------- 1 postgres postgres 400 May 1 19:08 /tmp/xxxac
-rw------- 1 postgres postgres 400 May 1 19:08 /tmp/xxxad
-rw------- 1 postgres postgres 400 May 1 19:08 /tmp/xxxae
-rw------- 1 postgres postgres 400 May 1 19:08 /tmp/xxxaf
-rw------- 1 postgres postgres 400 May 1 19:08 /tmp/xxxag
-rw------- 1 postgres postgres 400 May 1 19:08 /tmp/xxxah
-rw------- 1 postgres postgres 400 May 1 19:08 /tmp/xxxai
-rw------- 1 postgres postgres 401 May 1 19:08 /tmp/xxxaj
Each of those contains 100 lines.
Torsten
