Re: importing a messy text file

[Rob Sargentg <robjsargent@xxxxxxxxx>]

On 04/30/2014 03:50 AM, Willy-Bas Loos wrote:

On Wed, Apr 30, 2014 at 11:06 AM, Rémi Cura <remi.cura@xxxxxxxxx> wrote:

with latest version 9.3 you can use a copy from with a programm as argument.

You can use the unix | pipe for most of the stuff.

If you have an older version you can use the mkfifo command to "trick" postgres into believing he will copy from a file.

Hi, you're right i can see the new feature in the docs.

But since i am working from the (bash) command line already, i don't see any use for this.

I already pass the data like this:
cat <file> |psql -c "copy <table> from stdin NULL ''" <db> > <file>.log 2>&1

It's especially handy if you want to stay in sql(transaction), i gues.

Cheers,

--

Willy-Bas Loos

Unless you know this is the only problem (extra/missing tab) I would triage the file with a scripting language. One favourite check is

awk '{a[NF]++}END{print "Field count: Record count";for (i in a){printf "%11d : %d", i, a[i]}}

If you only have the tab problem you will get two lines of output with field count = N, N-1.  Take care in setting the field separator.