Here's an annoying one. The quirky but uber-fast Text::CSV_XS module can handle newlines embedded in fields if you turn on binary mode (binary => 1). MySQL's SELECT INTO OUTFILE and LOAD DATA INFILE can also handle embedded newlines. However, MySQL wants to prefix them with the escape character (a '\' in my case), and Text::CSV_XS chokes on that.

After fiddling with the options for both, I can't find a way to make them agree. It looks like my only option might be to s{\\n}{\n}g on the way out of MySQL and then reverse that on the way back in after processing with Text::CSV_XS. That is likely to hurt performance, since I am using the getline() and print() methods in Text::CSV_XS and I would have to switch to reading and printing the lines myself. (Reading the lines and dealing with embedded newlines myself? Ick!)

If anyone has any suggestions, I would appreciate them. Here's a (reduced) sample of the code I'm using:

use Text::CSV_XS;
use IO::Handle;
use Carp qw(croak);

my $file = $ARGV[0];

my $parser =
  Text::CSV_XS->new({
                     binary       => 1,
                     always_quote => 1,
                     escape_char  => '\\', #' help emacs
                    });

# open input and output temp files
open(my $in_fh, '<', $file)
  or croak("Unable to open input file $!");

my $line_num = 0;

# Using while(1) here and exiting via last() so we can avoid testing t
+o see
# if it's the last line (with eof) more than once per line.
while (not eof($in_fh)) {
    $line_num++;

    # parse the row and handle any errors
    my $in_values = $parser->getline($in_fh);
    croak("Unable to parse line $line_num of file $file: " . $parser->
+error_input())
      if !$in_values;
}
close $in_fh
  or croak("Error closing file $file: $!");
[download]

And the SQL looks like this:

SELECT blah blah blah
INTO OUTFILE ?
FIELDS TERMINATED BY ','
ESCAPED BY '\\'
ENCLOSED BY '"'
LINES TERMINATED BY '\n'
[download]