Postgres COPY with multiple small chunks

Question

In postgres, we can do COPY to export the data into file. But all the data will be saved into a file. But I want to chunk that file into small files. We can do this after the export.

But is there a way to export it during the COPY process?

OR in psycopg2 can we do this?

• Vérace: You could provide the table DDL so that an optimal field can be chosen?

Answer 1

You can use a WHERE condition:

COPY (SELECT * FROM tab WHERE id % 3 = 0) TO '/dir/file1.csv' (FORMAT 'csv');
COPY (SELECT * FROM tab WHERE id % 3 = 1) TO '/dir/file2.csv' (FORMAT 'csv');
COPY (SELECT * FROM tab WHERE id % 3 = 2) TO '/dir/file3.csv' (FORMAT 'csv');

Here id stands for any numerical column. You can also use other data types and split up the data in other ways with an appropriate WHERE condition.

• TheDataGuy: WHERE id % 3 = 0 will this work only if i have the id column?
• Laurenz Albe: I don't quite understand your question.
• TheDataGuy: in the where condition you mentioned that id, if I don't have this column in the table, then it wont run right?
• Laurenz Albe: Ah, I see. I have added an explanation to the answer.

Answer 2

But is there a way to export it during the COPY process?

Streaming the COPY data into a program is supported through the PROGRAM clause.

The Unix command split can do the actual splitting. For instance:

COPY (<your query>) TO PROGRAM 'split -l 10000 -d - /path/file-';

This will split the output into files numbered sequentially that do not exceed 10000 lines each.

The PROGRAM clause is also supported by psql's \copy; in that case the program is executed client-side.

Caveat: if the export format is CSV and there are multi-line text fields, it comes with the risk of splitting a field across multiple files.

Answer 3

If you already have a big file and have enough disc space you can use the method below.

Suppose that you have created a big file with copy command;

COPY (SELECT ... FROM ... WHERE ...) TO '/path/to/file';

Split files into chunks (Linux only);

split --number=l/10 --numeric-suffixes /path/to/file file_