A Short i18n/l10n Example in Ada

This started with me trying, in probably a ham-fisted way, to apply the recommendations in a gemlog post I liked.

Any mistakes in the following are my fault.

First, I started with the usual "alr init", followed by "alr with intl", then edited the main .adb file to be the following:

with Intl;
with Ada.Strings.Wide_Bounded;
with Ada.Strings.UTF_Encoding.Wide_Strings;
with Ada.Wide_Text_IO;
procedure Hello is
   PROG_NAME : constant String := "hello";
   function Gettext (Message : in String) return String is (Intl.Gettext (Message));
   package Dyn is new Ada.Strings.Wide_Bounded.Generic_Bounded_Length (Max => 100);
   Message : Dyn.Bounded_Wide_String;
begin
   Intl.Initialize (PROG_NAME, "/home/user/locale");
   Message := Dyn.To_Bounded_Wide_String (Source => Ada.Strings.UTF_Encoding.Wide_Strings.Decode (Gettext ("Hello world!")));
   Ada.Wide_Text_IO.Put_Line (Dyn.To_Wide_String (Message));
end Hello;

In my opinion, even though Ada can be verbose this is reasonably short and readable. Some salient points:

Need to use Wide_ variants of packages. I chose this instead of Wide_Wide_ out of pure stubborness--I care a lot about modern East Asian languages where a big effort was made with Han unification to fit in the Basic Multilingual Plane, and not at all about emojis.
I had to use Gettext instead of the shorter "-" to work with the xgettext utility (see below).
The use of the `Message` variable is artificial, just for experimentation. The result of `Decode` could have been passed directly to `Put_Line`.

Now the rest is mostly done at a shell prompt:

xgettext --from-code=UTF-8 --keyword=Gettext --add-comments -C -o hello.pot src/hello.adb
msginit --input=hello.pot --locale=ja --output=ja.po

Note that for longer Ada programs you might have unmatched single quotes, e.g. for attributes like `'Value` and so have to add comments like `-- '` just to keep xgettext happy. Now change the encoding in ja.po to UTF-8 if necessary, and replace the empty string translation with "こんにちは世界！". Next, some more shell commands:

mkdir -p $HOME/locale/ja/LC_MESSAGES
msgfmt --output-file=$HOME/locale/ja/LC_MESSAGES/hello.mo ja.po
export LANGUAGE=ja
./bin/hello

And it does what you expect. If LANGUAGE is unset, or no translation exists, the fallback (English) is used. All-in-all, pretty painless. I'm interested in language processors, so have some advice there:

Don't internationalize keywords & variable/constant names. France tried this, it's pointless since programming languages are closer to Maths than English anyway.
It's a tough call, but I think comments should be in English if you want anyone else to help with maintenance. In an ideal world everyone would be a polyglot, but ...
Signon banners, progress messages, etc. should be internationalized.
Error messages should be internationalized.

I've been doing other stuff, but it seemed better captured in PRs, defect reports, etc. instead of gemlog entries. I wholeheartedly recommend the Distributed Systems Annex!

Back to my gemlog