On 22/04/14 21:09, Nicklas Avén wrote:
> Hallo
>
> I am struggling to find the best solution to ignore blank lines in
> csv-file when using file_fdw.
>
> A blank line makes the table unreadable.
>
> I would like to avoid manipulating the file directly and avoid the
> need to make a new corrected copy of the file.
>
> I am on Linux so I have found a solution when using COPY:
> COPY test_table from program 'sed ''/^ *$/d''
> /opt/builds/inotify_test/test.csv' with (format 'csv', header
> 'true');
>
> but since the "program" option not seems to be implemented in file_fdw
> I am still searching for a solution.
file_fdw uses the same mechanism internally as "COPY <table> FROM '/file.csv'";
I don't think there's currently a way for this mechanism to ignore blank
lines.
Unfortunately CSV is not exactly a well-defined standard, so it's debatable
whether it's worth modifying the mechanism to cope with this situation.
The closest thing to a standard, RFC 4180 ( http://tools.ietf.org/html/rfc4180 )
doesn't seem to have anything to say about them; on the other hand LibreOffice
Calc will happily import files with blank lines.
> I have also found in an email from 2011
> http://www.postgresql.org/message-id/4E699DE6.8010606@gmail.com
>
> that when force_not_null was implemented in file_fdw the patch also
> included "some cosmetic changes such as removing useless blank lines."
> But I do not find that blank lines is removed in general since I
> cannot read csv-files with blank lines, and I do not understand how
> the option "force_not_null" can do the trick since that is on the
> column level and not lines/row.
The "blank lines" referred to here are in the source code itself.
Regards
Ian Barwick
--
Ian Barwick http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
