Re: UTF-8 representation question

First, the code you posted and the code you ran is actually different. There's only one byte in the string assigned to $test in the code you posted. There was two in the code you ran.

Second, you'll find Devel::Peek's Dump more useful at debugging this kind of problem.

                                                   Internally
>perl -MDevel::Peek -e"Dump(qq{\x{C2AE}});"     __ encoded
SV = PV(0x23734c) at 0x235fc0                  /   as UTF-8
  REFCNT = 1                                  /
  FLAGS = (PADBUSY,PADTMP,POK,READONLY,pPOK,UTF8)
  PV = 0x182f574 "\354\212\256"\0 [UTF8 "\x{c2ae}"]
  CUR = 3         -----v------           ---v----
  LEN = 4               \                    \____ String
                         \
                          \_______________________ Internal
                                                   encoding
                                                   (bytes)
>perl -MDevel::Peek -e"Dump('abc');"
SV = PV(0x236e00) at 0x182f778
  REFCNT = 1
  FLAGS = (PADBUSY,PADTMP,POK,READONLY,pPOK)
  PV = 0x182f56c "abc"\0
  CUR = 3         -v-         Bytes treated as
  LEN = 4           \________ iso-latin-1 by
                              perl functions
[download]

Can anyone tell me why the following code:

It all boils down to the following: Until you tell it otherwise by using "use utf8;", perl treats source code as iso-latin-1. You created an UTF-8 source file, and failed to notify perl.

$test is assigned two bytes or two iso-latin-1 chars. (Same thing, the difference is in how its used.) Adding "use utf8;" before the constant will cause the constant to be decoded from UTF-8.

When you call encode('UTF-8'), you're encoding characters that you've never decoded, producing junk.

On a related note, what is the difference between:
$test = "\x{05D0}\x{20AC}";
and
$test = "\x05\xD0\x20\xAC";

The first is a string of two UNICODE characters (internally encoded as UTF-8).
The second is a string of four bytes or four iso-latin-1 characters.

>perl -MDevel::Peek -e"Dump(qq{\x{05D0}\x{20AC}});"
SV = PV(0x237354) at 0x235fc8
  REFCNT = 1
  FLAGS = (PADBUSY,PADTMP,POK,READONLY,pPOK,UTF8)
  PV = 0x18307bc "\327\220\342\202\254"\0 [UTF8 "\x{5d0}\x{20ac}"]
  CUR = 5
  LEN = 8

>perl -MDevel::Peek -e"Dump(qq{\x05\xD0\x20\xAC});"
SV = PV(0x237354) at 0x235fc8
  REFCNT = 1
  FLAGS = (PADBUSY,PADTMP,POK,READONLY,pPOK)
  PV = 0x18307bc "\5\320 \254"\0
  CUR = 4
  LEN = 8
[download]

Comment on Re: UTF-8 representation question Select or Download Code

Replies are listed 'Best First'.
Re^2: UTF-8 representation question by bpa (Novice) on Sep 04, 2008 at 03:22 UTC
Thank you, ikegami, for your extremely informative reply. I see your point about not notifying perl I was using UTf-8. When I include the 'use utf8' (without the semicolon) in front of my $test assignment line, the length() function still tells me that the length of $test is 2 chars, but it does print out correctly to the shell as the registered trademark symbol, instead of junk. If I insert the semicolon, it prints junk. Is it possible that my shell is interpreting the output differently than perl? I suppose I should read up on Perl's inner workings. One minor note however, the code I posted was in fact the code that I ran, so I am still confused on that point. I appreciate your help.	[reply]
Re^3: UTF-8 representation question by ikegami (Patriarch) on Sep 04, 2008 at 05:38 UTC
Is it possible that my shell is interpreting the output differently than perl? It's all bytes until you pass them to something that cares, such as `lc` or your shell. Each of those "somethings" will interpret the bytes as it sees fit. One minor note however, the code I posted was in fact the code that I ran, so I am still confused on that point. I appreciate your help. There are a number of stetps between the file and Perlmonks where the bytes could have been substituted, plus PerlMonks itself and my browser. If I insert the semicolon, it prints junk. So far, we've only covered decoding the source. Sounds like you didn't properly encode the data while printing it. One way: `#!/usr/bin/perl use strict; use warnings; # Decode source from UTF-8. use utf8; # Decode STDIN as per locale. # Encode STDOUT & STDERR as per locale. use open qw( :std :locale ); my $test = '...'; print($test); Dump($test);` [download] Or if you want to decode/encode your input/output using a specific encoding, you can do it as follows: `#!/usr/bin/perl use strict; use warnings; # Decode source from UTF-8. use utf8; # Expect UTF-8 from STDIN. # Send UTF-8 to STDOUT & STDERR. BEGIN { binmode STDIN, ':encoding(UTF-8)' or die; binmode STDOUT, ':encoding(UTF-8)' or die; binmode STDERR, ':encoding(UTF-8)' or die; } my $test = '...'; print($test); Dump($test);` [download] Feel free to replace "..." with characters of your choice. If you're still having problem, please provide the `Dump` output, a description of what you see from the `print` (primarily the number of characters you see), how many characters you are expecting to see, and which of the two programs produced the output.	[reply] [d/l] [select]