Add German validation messages

[rustermi]

Summary
Similar to the addition of Chinese i18n messages, I have added German message strings.

Drawbacks
For simplicity reasons, I have reused the Issue471Test. Using a single class to test different languages would be a hassle due to the ResourceBundle being statically initialised. I hope this suffices as I did not want to add a Powermock dependency or do similar reflective classloader manipulation.

[stevehu]

@rustermi Thanks a lot for your help.

[AndreasALoew]

Sorry, but I seem to have run into a scenario where the translation is incomplete:

"requestBody.data darf nur für ein einziges Schema but more than one schemas {{"$ref":"#/components/schemas/tEnS"}{"$ref":"#/components/schemas/vEnS"}{"$ref":"#/components/schemas/vEnSAndTens"}} are valid gültig sein"

This is NOT a bug in the translated messages, but rather the code location raising this validation message makes improper use of the parameter to augment the message text, which then leads to this mixture of German and English...

Can you please look into fixing this? Shall I open a new issue for this?

Thanks a million! 😄

[stevehu]

@AndreasALoew If you know how to get it fixed, please open a PR and let @rustermi review it. Thanks a lot for raising the issue.

[AndreasALoew]

@stevehu Unfortunately I don't 😢 @rustermi Can you help?

My assumption is that the place in the codebase where the above exception is being thrown, adds exception text "but more than one schemas" (...) "are valid" to the exception parameters instead of treating this text as part of the message resource text that is to be translated

Does that help?

[rustermi]

Good find @AndreasALoew.

I currently don't have the time to further investigate but at first glance, it seems straight forward:
OneOfValidator line 287 adds the untranslated message "but more than one schemas {%s} are valid ".

We need a new message string for that in

• jsv-messages.properties
• jsv-messages_DE.properties
• jsv-messages_zh_CN.properties

Just add a translation for messages that you speak. ResourceBundle will default to the English translation in jsv-messages.properties if there is for example no Chinese translation.

Let's say, we call this message

oneOfMultipleSchemas = but more than one schemas {0} are valid 

Then, in OneOfValidator use I18nSupport.getString("oneOfMultipleSchemas"). Pay attention to the space in the end. It will potentially be removed/ignored in the messages.properties file.

Maybe this information is sufficient for you to create a PR?

[AndreasALoew]

Many thanks @rustermi - from the Java codebase perspective, it now looks feasible to me.

I have to admit that this will be the very first PR that I ever raised on GitHub (we're using GitLab internally at my customer), but I hope I'll find my way through (otherwise, I'll get back here and ask...).

So stay tuned, if all goes well, you should find a PR here early next week...

[AndreasALoew]

Another case of incomplete German translation:

"requestBody.data.zeitraum.vonZeitpunkt: X is an invalid date-time"

When I look into this over the weekend, I try to find the matching code location as well...

[rustermi]

@AndreasALoew thanks for your efforts. You can just paste the untranslated snippet into the Github search bar on the top left and search for code appearances in this repository. You will then find that
ValidatorTypeCode line 65 is untranslated as well.

[AndreasALoew]

I've completed a "potential" patch/candidate pull request, but as I cannot get Maven tests to complete without numerous errors, I have so far refrained from creating the PR.

The issue is that I seem unable to find out how to make the Maven surefire plugin use a locale of "en_US" to run the tests (and not my default locale of de_DE)...

@rustermi can you please advise how I can successfully switch my Maven and/or JVM locale to make the Maven tests pass?

[rustermi]

@AndreasALoew you can use Locale.setDefault to set your JVM locale. However be aware that this project initialises ResourceBundle statically, meaning that once it has been initialised, you can no longer switch the Locale. More specifically that means if you setDefault after ResourceBundle has been created, your changes will not have any effect.

Issue471Test illustrates this perfectly. You can enable a single test and it should succeed. However, you cannot enable more than one because the Locale will only be set once.

[AndreasALoew]

So does this mean that - as my main work computer runs with a German Windows OS Locale, I don't have any other chance passing the full testsuites of json-schema-validator other than by modifying test sources directly?

Unfortunately, I am far too busy in my current customer project in order to even think about making more substantial contributions, but I'd rather hoped that it would be possible to externally configure my Maven environment (and/or the main Maven POM, but even this would not be an elegant way) in order to run/pass all tests with a default Locale of en_US that can be set by JVM and/or Maven system properties?

In case this indeed is not possible, would you accept a PR that has not even been verified by running the Maven surefire tests?

[rustermi]

The tests that verify i18n should not rely on the system locale or else they would fail on some systems and succeed on others. Instead, the i18n tests should programmatically define a locale via Locale.setDefault first.

I have not tried to manipulate my system locale on maven level but this StackOverflow post seems to illustrate exactly this; once via pom and once via an environment variable.

You can open up a PR and assign me. If I don't get the tests running myself then, I wouldn't approve though.

[AndreasALoew]

@rustermi I've tried it in several ways including the ones mentioned in the StackOverflow post you mentioned, but I simply cannot get the standard "mvn clean install" surefire tests to assume any other locale than German, and so a huge number fails...

Either there is something really strange in my environment that does not allow Locales to be overridden (but I would have no idea what this could be...), or I fear that the whole "json-schema-validator" testing needs to be reworked to support running from clients with a default locale that is different from en_US... 😢

So if you also don't know of a specific change in Maven settings and/or MAVEN_OPTS, JVM system properties etc. I fear that the only way to move forward is that I create an untested PR (or attach an old-style unified patch, because it's just very few lines...).

Please advise!

[rustermi]

Do so, I'm working with an English locale anyways. If tests run green for me, your changes look okay and I understand why it would fail on your machine, I'd approve it.

To help, it would be great if you could post your (failing test) output together with your PR so that I can verify what's going on.

[AndreasALoew]

ALL CLEAR: Successfully uncovered and properly resolved the true root cause of the Locale/testing issues -> see PR #536

[rishabh413]

Hi, I am importing this library for my project and I am trying to understand why System.exit() calls were added in this commit. In general, a library is not expected to have JVM exit() calls. Could you help understand the requirements here?

[stevehu]

I did a full-text search and couldn't find the exit() you mentioned. Could you please point me to the line of code? Thanks.

[rishabh413]

Thanks for looking into it.

1. please take a look at search results here:
   

2. The code is in I18nSupport class. Reference code links: 1 2

3. These were added in Add German validation messages #511

[stevehu]

@rishabh413 Thanks for raising it. I have fixed it with the following issue.

#708