Postgres COPY FROM csv file- No such file or directory

Question

I'm trying to import a (rather large) .txt file into a table geonames in PostgreSQL 9.1. I'm in the /~ directory of my server, with a file named US.txt placed in that directory. I set the search_path variable to geochat, the name of the database I'm working in. I then enter this query:

COPY geonames
FROM 'US.txt',
DELIMITER E'\t',
NULL 'NULL');

I then receive this error:

ERROR: could not open file "US.txt" for reading: No such file or directory.

Do I have to type in \i US.txt or something similar first, or should it just get it from the present working directory?

Answer 1

A couple of misconceptions:

1.

I'm in the /~ directory of my server

There is no directory /~. It's either / (root directory) or ~ (home directory of current user). It's also irrelevant to the problem.

2.

I set the search_path variable to geochat, the name of the database I'm working in

The search_path has nothing to do with the name of the database. It's for schemas inside the current database. You probably need to reset this.

3.
You are required to use the absolute path for your file. As documented in the manual here:

filename

The absolute path name of the input or output file.

4.
DELIMITER: just noise.

The default is a tab character in text format

5.
NULL: It's rather uncommon to use the actual string 'NULL' for a NULL value. Are you sure?

The default is \N (backslash-N) in text format, and an unquoted empty string in CSV format.

My guess (after resetting search_path - or you schema-qualify the table name):

COPY geonames FROM '/path/to/file/US.txt';

Answer 2

Maybe a bit late, but hopefully useful:

Use \copy instead

https://wiki.postgresql.org/wiki/COPY

jvdw

Answer 3

The paths are relative to the PostgreSQL server, not the psql client.

Assuming you are running PostgreSQL 9.4, you can put US.txt in the directory /var/lib/postgresql/9.4/main/.

Answer 4

Another option is to pipe it in from stdin:

cat US.txt | psql -c "copy geonames from STDIN WITH (FORMAT csv);"

Answer 5

if you're running your COPY command from a script, you can have a step in the script that creates the COPY command with the correct absolute path.

MYPWD=$(pwd)
echo "COPY geonames FROM '$MYPWD/US.txt', DELIMITER E'\t';"
MYPWD=

you can then run this portion into a file and execute it

./step_to_create_COPY_with_abs_path.sh >COPY_abs_path.sql
psql -f COPY_abs_path.sql -d your_db_name