MDEV-38583 : Fix TO_DATE Urdu parsing mismatch with Oracle

[Mahmoud-kh1]

The problem :
previously, TO_DATE() would return NULL when using Urdu month or day names

The bug :
I found the bug is a mismatch for certain characters in days and months

Key changes :

• fixed the harded coded words by replacing all with standard urdu and shows the hexa to ensure it's correct
• added test to verify the query run without any problem and hexa based testing to ensure there is no typo or mismatching characters.

Now it works as intended

issue ID:
MDEV-38583

[gkodinov]

Thank you for your contribution. This is a preliminary review.
Couple of things:

• Please add a better commit message that according to the coding standard: describes what the issue is, how it was fixed and how it was tested.
• I do not see the CLA bot confirmation. Please rectify this
• I would add some HEX codes into the tests to make sure the relevant encoded texts (day names) are spelled correctly.

[Mahmoud-kh1]

Thank you for your contribution. This is a preliminary review. Couple of things:

* Please add a better commit message that according to the coding standard: describes what the issue is, how it was fixed and how it was tested.

* I do not see the CLA bot confirmation. Please rectify this

* I would add some HEX codes into the tests to make sure the relevant encoded texts (day names) are spelled correctly.

I fixed the tests and description of pr and commit message , for the CLA bot I don't know what's the problem exactly but its check passes here

[gkodinov]

Thank you! LGTM. Please stand by for the final review.

[abarkov]

The patch looks correct for me.
The new locale data matches Unicode Locale Data at:
https://github.com/unicode-org/cldr/blob/main/common/main/ur.xml

Can you please backport the patch to 12.3, so we merge it into the earliest version which supports TO_DATE() - since 12.3.1.

Another option would be to backport to 10.6, but this will need rewritting tests using STR_TO_DATE() instead of TO_DATE().

Thanks.

[Mahmoud-kh1]

The patch looks correct for me. The new locale data matches Unicode Locale Data at: https://github.com/unicode-org/cldr/blob/main/common/main/ur.xml

Can you please backport the patch to 12.3, so we merge it into the earliest version which supports TO_DATE() - since 12.3.1.

Another option would be to backport to 10.6, but this will need rewritting tests using STR_TO_DATE() instead of TO_DATE().

Thanks.

done , I backport it to 12.3

[vuvova]

Thank you, @Mahmoud-kh1. Your PR made us think what other languages might be incorrect and we've done a complete comparison with https://unicode.org/cldr/charts/48/summary/root.html and hopefully fixed everything that was wrong.
See for example 1a88fa7.

As it fixed everything, it includes your ur_PK corrections too. Just to be extra sure, I've verified that https://unicode.org/cldr/charts/48/summary/ur.html matches your changes.

But it means we'll push the commit that fixes everything, not your PR, unfortunately. Thanks for the work though, it prompted us to do this.

[Mahmoud-kh1]

Thank you, @Mahmoud-kh1. Your PR made us think what other languages might be incorrect and we've done a complete comparison with https://unicode.org/cldr/charts/48/summary/root.html and hopefully fixed everything that was wrong. See for example 1a88fa7.

As it fixed everything, it includes your ur_PK corrections too. Just to be extra sure, I've verified that https://unicode.org/cldr/charts/48/summary/ur.html matches your changes.

But it means we'll push the commit that fixes everything, not your PR, unfortunately. Thanks for the work though, it prompted us to do this.

Thanks, honestly knowing that my PR prompted you is more valuable to me than the merge itself.