Exporting Data


COPY ( <SELECT statement> ) TO '<file path>' [WITH (<property> = value, ...)];

<file path> must be a path on the server. This command exports the results of any SELECT statement to the file. There is a special mode when <file path> is empty. In that case, the server automatically generates a file in <MapD Directory>/mapd_export that is the client session id with the suffix .txt.

<property> in the optional WITH clause can be:

  • delimiter: a single-character string for the delimiter between column values. The default is ",", i.e., as a CSV file.
  • nulls: a string pattern indicating a field is NULL. The default is \N.
  • escape: a single-character string for escaping quotes. The default is the quote character itself.
  • quoted: 'true' or 'false' indicating whether all the column values should be output in quotes. The default is 'true'.
  • quote: a single-character string for quoting a column value. The default quote character is double quote ".
  • line_delimiter a single-character string for terminating each line. The default is "\n".
  • header: 'true' or 'false' indicating whether to output a header line for all the column names. The default is 'true'.

Note: A frequent error you might encounter when using the COPY TO command is “Query couldn’t keep the entire working set of columns in GPU Memory.” To avoid this error, you can restart your MapD server in CPU mode before using the COPY TO command. See Configuration.


COPY (SELECT * FROM tweets) TO '/tmp/tweets.csv';
COPY (SELECT * tweets ORDER BY tweet_time LIMIT 10000) TO
  '/tmp/tweets.tsv' WITH (delimiter = '\t', quoted = 'true', header = 'false');