http://www.perlmonks.org?node_id=11104489


in reply to Re^5: Text::CSV encoding parse()
in thread Text::CSV encoding parse()

Hello haukex, thanks a million for your advice, but I have to confess this is way beyond my understanding or abilities in many ways, so I think I'm going to have to live with it. FWIW every time encoding problems come up I get lost in the weeds. (I'm not a developer, just a Perl hack, don't grok hexdump or Devel::Peek etc.)

All I know in this case is I can print my @sorted_array rows to a flat file (opened with Notepad++) or to a web page using CGI, and those characters look fine. It's only when I use Text:CSV that something goes haywire.

Anyway I appreciate your patience and help, sorry for the trouble.

Replies are listed 'Best First'.
Re^7: Text::CSV encoding parse()
by haukex (Bishop) on Aug 15, 2019 at 06:43 UTC

    Try taking the following code and replacing my $data with your code that fetches the data from the database (as short as possible), and post both your code and the output here. <update> Note that PerlMonks does not handle Unicode inside of <code> tags well, so tell us if you've got any Unicode in there. </update>

    #!/usr/bin/env perl
use warnings;
use 5.012;
use utf8; # Perl script file is encoded as UTF-8
use open qw/:utf8 :std/; # reopen STDIN/OUT/ERR as UTF-8
use Text::CSV;
use CGI qw/escapeHTML/;
use CGI::Carp qw/fatalsToBrowser warningsToBrowser/; # for debug ONLY!
use Data::Dumper;
$Data::Dumper::Useqq=1;

my $cgi = CGI->new;
print $cgi->header(-charset=>'UTF-8');
print $cgi->start_html(-title=>'Example', -encoding=>'UTF-8');
warningsToBrowser(1);

my $data = "Euro symbol: € | I \N{U+2764}\N{U+FE0F} \N{U+1F42A}";

print $cgi->pre(escapeHTML( Dumper( $data ) # debugging
	."UTF-8 flag is ".( utf8::is_utf8( $data )?'on':'off' ) ));

print $cgi->p(escapeHTML( $data ));

my $csv = Text::CSV->new ({ binary => 1, sep_char => "|" });
$csv->parse($data);
my ($c1,$c2) = $csv->fields;
print $cgi->p(escapeHTML( "After Text::CSV: ".$c1." | ".$c2 ));

print $cgi->end_html;

    (Note: It is better to use utf8::is_utf8() only for debugging.) The output you should see in the browser from the above:

    $VAR1 = "Euro symbol: \x{20ac} | I \x{2764}\x{fe0f} \x{1f42a}";
UTF-8 flag is on

    Euro symbol: € | I ❤️ 🐪

    After Text::CSV: Euro symbol: € | I ❤️ 🐪

[reply]
[d/l]
[select]

      Using the perl internals through Data::Peek's DPeek, you'll see both versions if UTF-8 is in effect without the fragile use of utf8 function calls. It also shows the importance of using utf8 in your example code.

      $ perl        -MData::Peek -wE'my $data = "Euro symbol: € | I \N{U+276
+4}\N{U+FE0F} \N{U+1F42A}"; DPeek $data'
PV("Euro symbol: \303\242\302\202\302\254 | I \342\235\244\357\270\217
+ \360\237\220\252"\0) [UTF8 "Euro symbol: \x{e2}\x{82}\x{ac} | I \x{2
+764}\x{fe0f} \x{1f42a}"]

$ perl -Mutf8 -MData::Peek -wE'my $data = "Euro symbol: € | I \N{U+276
+4}\N{U+FE0F} \N{U+1F42A}"; DPeek $data'
PV("Euro symbol: \342\202\254 | I \342\235\244\357\270\217 \360\237\22
+0\252"\0) [UTF8 "Euro symbol: \x{20ac} | I \x{2764}\x{fe0f} \x{1f42a}
+"]

$ perl -Mutf8 -MData::Peek -wE'my $data = "Euro symbol: \xe2\x82\xac |
+ I \xe2\x9d\xa4\xef\xb8\x8f \xf0\x9f\x90\xaa"; DPeek $data'
PV("Euro symbol: \342\202\254 | I \342\235\244\357\270\217 \360\237\22
+0\252"\0)

      [download]

      Enjoy, Have FUN! H.Merijn
[reply]
[d/l]
[select]
        without the fragile use of utf8 function calls

        Although you're certainly right that several of the functions from utf8:: should be used with extreme caution (or not at all), AFAIK using is_utf8 to check on the flag for debugging (only!) should be fine. I was just trying to provide a slightly "nicer" debugging output because slugger415 said "I'm not a developer, just a Perl hack, don't grok hexdump or Devel::Peek etc.".

        slugger415: I just wanted to add that I wasn't necessarily suggesting you should try to understand the output, it's also very useful information for us to help you debug.

[reply]
[d/l]
[select]

      Hello, thank you for the test code. Yes I see the same in the browser with your sample code. I've substituted my own string for yours and here's what I see now in the browser:

      $VAR1 = "/search/\x{bf}Cuales son las partes de una cadena de conexi\x
+{f3}n??scope=SSGU8G_12.1.0|/com.ibm.jdbc_pg.doc/ids_jdbc_011.htm|0|1|
+1|0\n";
UTF-8 flag is on

/search/¿Cuales son las partes de una cadena de conexión??scope=SSGU8G
+_12.1.0|/com.ibm.jdbc_pg.doc/ids_jdbc_011.htm|0|1|1|0

After Text::CSV: /search/¿Cuales son las partes de una cadena de conex
+ión??scope=SSGU8G_12.1.0 | /com.ibm.jdbc_pg.doc/ids_jdbc_011.htm

      [download]

      BTW I'm assigning $data by reading a stdout output file that is returned from the db via the tool I'm using for the SQL query.

      my($data);

open my $fh, "<:encoding(utf8)", $file || die("cannot open $file file\
+n");
while(<$fh>){
 if($_ =~ /Cuales/){
  $data = $_;
  print $_;
 } 
}
close($fh);

      [download]

      I've also tried opening the file without the :encoding and get the same result. 

      open(R, "$resultsFile") || die("cannot open results file $resultsFile 
+for reading.\n");

      [download]

      Hope this is helpful.

[reply]
[d/l]
[select]
        I've substituted my own string for yours and here's what I see now in the browser: 
        $VAR1 = "/search/\x{bf}Cuales son las partes de una cadena de conexi\x{f3}n??scope=SSGU8G_12.1.0|/com.ibm.jdbc_pg.doc/ids_jdbc_011.htm|0|1|1|0\n";
UTF-8 flag is on

/search/¿Cuales son las partes de una cadena de conexión??scope=SSGU8G_12.1.0|/com.ibm.jdbc_pg.doc/ids_jdbc_011.htm|0|1|1|0

After Text::CSV: /search/¿Cuales son las partes de una cadena de conexión??scope=SSGU8G_12.1.0 | /com.ibm.jdbc_pg.doc/ids_jdbc_011.htm

        \x{bf} or U+00BF is the inverted question mark and U+00F3 is “ó”, so AFAICT that looks correct, doesn't it?

        BTW I'm assigning $data by reading a stdout output file that is returned from the db via the tool I'm using for the SQL query.

        Ok, so do I understand correctly that you've tested the code you showed within the script I posted?

        (BTW, why not use DBI to connect directly to the database instead of going through a file?)

        I've also tried opening the file without the :encoding and get the same result.

        Well, if you've got the use open qw/:std :utf8/; that I suggested at the top of the script, that also sets utf8 as the default encoding layer. As long as you're not getting warnings such as "Malformed UTF-8 character: \xbf" or "utf8 "\xBF" does not map to Unicode" (check the web server's error logs and/or run the script from the command line), then it would seem your file is probably encoded as UTF-8, and it would seem that everything is correct...

        So are you still seeing the same problem as before, or is it working now?

[reply]
[d/l]
[select]
Re^7: Text::CSV encoding parse()
by jcb (Parson) on Aug 14, 2019 at 23:16 UTC

    I am sorry, but you will need to learn to use hex dumps. They are the only reliable way to solve this type of encoding problem, where the "bytes on the wire" are what matters. Have you tried using Wireshark to observe the actual network traffic? It offers a convenient hex dump view of each packet.

[reply]

In Section Seekers of Perl Wisdom