Unpunycode search keywords and short names.

components/search_engines/template_url.cc

Index: components/search_engines/template_url.cc
diff --git a/components/search_engines/template_url.cc b/components/search_engines/template_url.cc
index 8202fa4483c0d7d0569b9bf5da797ff44b1c2f26..9cb82c82d3a768f9261d796b3ce9214cf67b5e52 100644
--- a/components/search_engines/template_url.cc
+++ b/components/search_engines/template_url.cc
@@ -1217,7 +1217,9 @@ base::string16 TemplateURL::GenerateKeyword(const GURL& url) {
   // Special case: if the host was exactly "www." (not sure this can happen but
   // perhaps with some weird intranet and custom DNS server?), ensure we at
   // least don't return the empty string.
-  base::string16 keyword(net::StripWWWFromHost(url));
+  // We do not want any punycode in any language.

[Matt Giuca, 2015/07/17 04:13:11]

This comment should be more detailed and should refer to IDN, not Punycode. How
about:
"The URL host may be an internationalized domain name (IDN). If so, decode it
into a regular Unicode string."

You should not be using the term "Punycode" here. Punycode is a sub-step in the
IDN encoding/decoding algorithm. Roughly, the IDN encoding algorithm is this:
1. Split the domain name by '.' character.
2. For any segment that contains non-ASCII characters, Punycode-encode it and
prepend "xn--".
3. Join the segments by '.' characters.

So what you are doing here is not Punycode-decoding, but IDN-decoding.

Also I'd like to see it separated from the above comment about stripping "www".
Maybe break this out like so:

base::string16 keyword = url.host();
// <comment about decoding IDN>
keyword = net::IDNToUnicode(keyword, std::string());
// <comment about stripping "www">
keyword = net::StripWWW(keyword);

[alshabalin, 2015/07/20 14:06:37]

Done.

+  base::string16 keyword(
+      net::StripWWW(net::IDNToUnicode(url.host(), std::string())));
   return keyword.empty() ? base::ASCIIToUTF16("www") : keyword;
 }