Changes for Postgresql Copy Export:

[RobertBHamilton]

1. Support for empty null string --null-string ''
2. Support for non-xml delim --fields-terminated-by'\0x1c'
3. Added line buffering perf --batch
4. optional TEXT mode instead of CSV -Dpostgresql.format.text=true
5. Support for postgres Version 8x -Dpostgresql.targetdb.ver=8
6. Optional disable escape sequences -Dpostgresql.input.israw=true

[RobertBHamilton]

This is in process. These changes were motivated by a project at GM to move significant data sets to Greenplum. Because Greenplum is currently still in version 8x of postgress we added support for the 8x syntax of the copy command. We also noticed that implementing line buffering in the mapper would significantly enhance performance. Also we conventionally use a char for field delim which happens to be invalid XML char so we added support for non-xml delims with direct mode.
TODO: put the same changes to the 9.x COPY syntax.

[szvasas]

Hi Robert,

Thanks for sharing these improvements!
As you described in your email it seems that there are a few different improvements covered in this PR so I would like to ask you to split these changes into separate patches and raise separate Sqoop issues for them on JIRA (https://issues.apache.org/jira/projects/SQOOP/issues).

I haven't done an in-depth review yet but I noticed that the indentation and the formatting you used is quite different from the surrounding code so please try to follow those conventions, e.g.:

• Add spaces before and after '=' character (e.g. bufferMode = false instead of bufferMode=false)

• The content of a block should go in a new line, e.g.

public void clear() {
  sb.setLength(0);
}

instead of
public void clear(){ sb.setLength(0); }

Apart from this I have seen a few commented out code lines, please remove those as well.

Regards,
Szabolcs

[Fokko]

Rebasing onto master will fix the CI again.