42

I've been using the psql Postgres terminal to import CSV files into tables using the following

COPY tbname FROM
'/tmp/the_file.csv'
delimiter '|' csv;

which works fine except that I have to be logged into the psql terminal to run it.

I would like to know if anyone knows of a way to do a command similar to this from the Linux shell command line similar to how Postgres allows a shell command like bellow

/opt/postgresql/bin/pg_dump dbname > /tmp/dbname.sql

This allows the dumping of a database from the Linux shell without being logged into psql terminal.

TheLovelySausage
  • 3,017
  • 5
  • 45
  • 80

4 Answers4

50

The solution in the accepted answer will only work on the server and when the user executing the query will have permissions to read the file as explained in this SO answer.

Otherwise, a more flexible approach is to replace the SQL's COPY command with the psql's "meta-command" called \copy which which takes all the same options as the "real" COPY, but is run inside the client (with no need for ; at the end):

psql -c "\copy tbname FROM '/tmp/the_file.csv' delimiter '|' csv"

As per docs, the \copy command:

Performs a frontend (client) copy. This is an operation that runs an SQL COPY command, but instead of the server reading or writing the specified file, psql reads or writes the file and routes the data between the server and the local file system. This means that file accessibility and privileges are those of the local user, not the server, and no SQL superuser privileges are required.

In addition, if the the_file.csv contains the header in the first line, it can be recognized by adding header at the end of the above command:

psql -c "\copy tbname FROM '/tmp/the_file.csv' delimiter '|' csv header"
Community
  • 1
  • 1
Dmitri Zaitsev
  • 11,773
  • 9
  • 61
  • 103
  • 2
    This is the best answer because production databases should certainly not allow users to login and execute locally, and exposing a file system share on the database is also a bad idea. COPY can be safely considered when the file is on a network share accessible both from database and user. In most cases, this \copy is a better way. Thanks! – Brian Bulkowski Mar 15 '20 at 22:59
40

As stated in The PostgreSQL Documentation (II. PostgreSQL Client Applications - psql) you can pass a command to psql (PostgreSQL interactive terminal) with the switch -c. Your options are:

1, Client-side CSV: \copy meta-command

perform the SQL COPY command but the file is read on the client and the content routed to the server.

psql -c "\copy tbname FROM '/tmp/the_file.csv' delimiter '|' csv"

(client-side option originally mentioned in this answer)

2. Server-side CSV: SQL COPY command

reads the file on the server (current user needs to have the necessary permissions):

psql -c "COPY tbname FROM '/tmp/the_file.csv' delimiter '|' csv;"

the DB roles needed for reading the file on the server:

COPY naming a file or command is only allowed to database superusers or users who are granted one of the default roles pg_read_server_files, pg_write_server_files, or pg_execute_server_program

also the PostgreSQL server process needs to have access to the file.

Simo Kivistö
  • 3,553
  • 3
  • 33
  • 38
11

The most flexible way is to use a shell HERE document, which allows you to use shell variables inside your query, even inside (double or single) quotes:

#!/bin/sh

THE_USER=moi
THE_DB=stuff
THE_TABLE=personnel
PSQL=/opt/postgresql/bin/psql
THE_DIR=/tmp
THE_FILE=the_file.csv

${PSQL} -U ${THE_USER} ${THE_DB} <<OMG
COPY ${THE_TABLE} FROM '${THE_DIR}/${THE_FILE}' delimiter '|' csv;
OMG
wildplasser
  • 38,231
  • 6
  • 56
  • 94
11

To complete the previous answer, I would suggest:

psql -d your_dbname --user=db_username -c "COPY tbname FROM '/tmp/the_file.csv' delimiter '|' csv;"
Community
  • 1
  • 1
Andrea Araldo
  • 923
  • 10
  • 16