Canonicalizing Strings with Unicode Combined Characters (Perl Cookbook, 2nd Edition)

1.9.1. Problem

You have two strings that look the same when you print them out, but they don't test as string equal and sometimes even have different lengths. How can you get Perl to consider them the same strings?

1.9.2. Solution

When you have otherwise equivalent strings, at least some of which contain Unicode combining character sequences, instead of comparing them directly, compare the results of running them through the NFD( ) function from the Unicode::Normalize module.

use Unicode::Normalize;
$s1 = "fa\x{E7}ade";                
$s2 = "fac\x{0327}ade";                
if (NFD($s1) eq NFD($s2)) { print "Yup!\n" }

1.9.3. Discussion

The same character sequence can sometimes be specified in multiple ways. Sometimes this is because of legacy encodings, such as the letters from Latin1 that contain diacritical marks. These can be specified directly with a single character (like U+00E7, LATIN SMALL LETTER C WITH CEDILLA) or indirectly via the base character (like U+0063, LATIN SMALL LETTER C) followed by a combining character (U+0327, COMBINING CEDILLA).

Another possibility is that you have two or more marks following a base character, but the order of those marks varies in your data. Imagine you wanted the letter "c" to have both a cedilla and a caron on top of it in order to print a . That could be specified in any of these ways:

$string = v231.780;
#   LATIN SMALL LETTER C WITH CEDILLA
#   COMBINING CARON

$string = v99.807.780;
#         LATIN SMALL LETTER C
#         COMBINING CARON
#         COMBINING CEDILLA

$string = v99.780.807
#         LATIN SMALL LETTER C
#         COMBINING CEDILLA
#         COMBINING CARON

The normalization functions rearrange those into a reliable ordering. Several are provided, including NFD( ) for canonical decomposition and NFC( ) for canonical decomposition followed by canonical composition. No matter which of these three ways you used to specify your , the NFD version is v99.807.780, whereas the NFC version is v231.780.

Sometimes you may prefer NFKD( ) and NFKC( ), which are like the previous two functions except that they perform compatible decomposition, which for NFKC( ) is then followed by canonical composition. For example, \x{FB00} is the double-f ligature. Its NFD and NFC forms are the same thing, "\x{FB00}", but its NFKD and NFKC forms return a two-character string, "\x{66}\x{66}".

1.9. Canonicalizing Strings with Unicode Combined Characters

1.9.1. Problem

1.9.2. Solution

1.9.3. Discussion

1.9.4. See Also