Incorrect Translations

[rposborne]

So thank you for this great gem. We have recently swapped out the countries gem to leverage the i18n_data and we have found some issues, that are rooted in some bad translations. What is the best ways to get these corrected? As the debian repo's aren't exactly "approachable".

See countries/countries#243 (comment)

[grosser]

adding modifications to the dump rake task should be a simple way of doing it, I don't think anyone is using the live-data anyway ...

[rposborne]

Certainly so I did a quick run through on our side, and it looks to be a bit more widespread then i thought. Mainly the XH translation is covered in duplicates.

Basically any name that is used twice for a given translation is a smell wouldn't you say?

[grosser]

yep, sounds good

On Thu, Mar 12, 2015 at 1:31 PM, rpo [email protected] wrote:

Certainly so I did a quick run through on our side, and it looks to be a
bit more widespread then i thought. Mainly the XH translation is covered in
duplicates.

Basically any name that is used twice for a given translation is a smell
wouldn't you say?

—
Reply to this email directly or view it on GitHub
#12 (comment).

[tilo]

here are two of the issues -- they are both in the XH file!!
Someone messed-up, and probably checked-in half translated files???

$ grep BZ cache/file_data_provider/countries* | grep Belg
cache/file_data_provider/countries-XH.txt:BZ;;Belgium

$ grep UG cache/file_data_provider/countries* | grep Cana
cache/file_data_provider/countries-XH.txt:UG;;Canada

let's see..

$ git blame cache/file_data_provider/countries-XH.txt | grep Cana
a63d006 (grosser 2012-06-20 09:31:15 -0700 40) CA;;Canada
f875daf (grosser 2009-01-18 12:20:41 +0100 184) RW;;Canada
a63d006 (grosser 2012-06-20 09:31:15 -0700 232) UG;;Canada

$ git blame cache/file_data_provider/countries-XH.txt | grep Belg
a63d006 (grosser 2012-06-20 09:31:15 -0700 22) BE;;Belgium
a63d006 (grosser 2012-06-20 09:31:15 -0700 23) BZ;;Belgium

So is XH an auto-generated file?

Would it be a safe heuristic/test to say that the same string for a country can not map to more than one Alpha2 code? At least for some languages?

[grosser]

idk .. just hit regenerate and see if that fixes it ?

On Thu, Mar 12, 2015 at 10:07 PM, Tilo [email protected] wrote:

What's XH anyhow? it's user assigned..

—
Reply to this email directly or view it on GitHub
#12 (comment).

[rposborne]

so I can confirm that the bugs are upstream in the pkg_isocodes. My favorite is the US virgin Islands are called "Finland"

Canda
http://anonscm.debian.org/cgit/pkg-isocodes/iso-codes.git/tree/iso_3166/xh.po#n1347

Finland
http://anonscm.debian.org/cgit/pkg-isocodes/iso-codes.git/tree/iso_3166/xh.po#n1783

The XH language appears to be south african but there is no way this translation file is "quality". http://en.wikipedia.org/wiki/Xhosa_language

[grosser]

maybe using cldr as datasource would give better results :(

On Fri, Mar 13, 2015 at 6:43 AM, rpo [email protected] wrote:

so I can confirm that the bugs are upstream in the pkg_isocodes. My
favorite is the US virgin Islands are called "Finland"

Canda

http://anonscm.debian.org/cgit/pkg-isocodes/iso-codes.git/tree/iso_3166/xh.po#n1347

Finland

http://anonscm.debian.org/cgit/pkg-isocodes/iso-codes.git/tree/iso_3166/xh.po#n1783

The XH language appears to be south african but there is no way this
translation file is "quality". http://en.wikipedia.org/wiki/Xhosa_language

—
Reply to this email directly or view it on GitHub
#12 (comment).

[rposborne]

Maybe so. Regardless CLDR seems much more approachable then the debian source does. It comes down in JSON which is nice 😄

[rposborne]

@grosser i would almost kill the XH translation all together... it doesn't even look like anything is translated.

[grosser]

go crazy :)

On Tue, Mar 17, 2015 at 1:19 AM, rpo [email protected] wrote:

@grosser https://github.com/grosser i would almost kill the XH
translation all together... it doesn't even look like anything is
translated.

—
Reply to this email directly or view it on GitHub
#12 (comment).