Fix H2 database case-sensitive identifier issues + upgrades JUnit tests cases to version 5

[mches]

Impact

• Bug fix (non-breaking change which fixes expected existing functionality)
• Enhancement/New feature (adds functionality without impacting existing logic)
• Breaking change (fix or feature that would cause existing functionality to change)

Description

On H2,

1. Restore compatiblity with Liquibase 4.6.2 with respect to reserved words for H2 1.x.
2. Fold identifiers that must be quoted due to being an illegal identifier or reserved word to uppercase so that all names are in a consistent case.
3. Allow override of the defaults by way of objectQuotingStrategy on the change log or change set, liquibase.preserveSchemaCase configuration property, modifySql, or extension.

Fixes #2385
Fixes #3289

Things to be aware of

• Imported the JUnit 5 BOM for better coordination of dependency versions, as a separate commit.
• Corrected whitespace errors, e.g. tab indentation, in JUnit 4 database tests a separate commit.
  • IntelliJ IDEA ignores this commit for Git Blame by default.
  • It's also possible to ignore this commit in other tools by adding it to .git-blame-ignore-revs.
• Converted JUnit 4 database tests to JUnit 5, as a separate commit.
• Upgraded Groovy from 2.4 to 2.5, as a separate commit.
  • This allows use of the tap closure.
• Configured Maven to also execute tests named *Spec, as a seprate commit.
  • This allows H2DatabaseTest can H2DatabaseSpec to coexist, so that the existing test doesn't have to be renamed or converted, along with it's class hierachy, to free up the name.
• Established baseline specs for H2 and all other core databases, as a separate commit.
  • Tests were passing after each commit.
  • This makes it easier to see where potential bugs exist or when bugs are fixed or behavior changed.

Things to worry about

Some existing users' change sets or databases may require modification if they're targeting a long-lived embedded H2 database and have chosen names that require implicit quoting such as reserved words, e.g. user, group, groups, or value, or names with spaces.

Additional Context

Spring Boot users upgrading to 2.7.x and using jOOQ Open Source Edition often have to downgrade H2 to 1.4.x. See spring-projects/spring-boot#31168, https://github.com/spring-projects/spring-boot/wiki/Spring-Boot-2.7-Release-Notes#jooq.

If desired, I can create separate PRs for

1. Testing improvements
2. H2 1.x reserved words
3. Identifier case folding or preservation with respect to objectQuotingStrategy

[nvoxland]

Initial review:

Thanks for the fix, and the other general cleanup in there to, @mches . I appreciate the separate commits for each step you did.

The conversion to JUnit 5 tests is nice. We've generally been converting from junit 4 tests to spock tests, but that takes time for each existing test when there is just a small change you'd like to make. Having them be the nicer junit 5 will help the times when it's not a good time to redo the whole test, but still want to take advantage of nicer syntax.

I also like the separate reserved word list for v1 vs v2. That helps us support both versions, which we need to continue doing.

In your last commit, you override quoteObject() which I don't think is the right spot for the fix. That method is generally used as "quote the object" where the caller is expecting it to be quoted. I see it's missing javadoc specifying that, though... So it seems like the "should we quote?" logic needs to be pushed upstream to determine whether the quoteObject method should be called or not? I didn't step through where you are seeing the problem to know where that is not working, but is there a better place?

Things to be aware of:

• Nothing

[mches]

Thank you for your time and feedback @nvoxland. Of course you're right about the purpose of each of the methods. I took that to heart and came up with a better organization of the logic. Hope you like it.

[filipelautert]

This PR is restoring compatibility with H2 1.x reserved words, and as an "extra" it also upgrades JUnit tests cases to version 5.
Thanks for those fix and improvements @mches . Also, I added some "merge branch master" commits - please feel free to rebase to get rid of them .

[mches]

Resolved the conflict in liquibase-extension-testing/src/main/groovy/liquibase/extension/testing/command/CommandTests.groovy

[mches]

Rebased onto latest master

[filipelautert]

Thanks for the rebase! I'll get it merged today.

[nealeu]

How do we take this change out for a test ride before the release? Should I be able to get it from Github packages without an access token for your repo?

[filipelautert]

Hey @nealeu , are you able to download it from here -> https://github.com/liquibase/liquibase/suites/15586269682/artifacts/891121515 ?

[shsmysore]

My H2 database has a table with one of the column name - key. (lower case reserved keyword)
How can I stop liquibase from automatically converting it to uppercase?

I tried adding below line in the changeset, but its not working.

<property name="preserveSchemaCase" value="true"/>

[mches]

My H2 database has a table with one of the column name - key. (lower case reserved keyword) How can I stop liquibase from automatically converting it to uppercase?

I tried adding below line in the changeset, but its not working.

<property name="preserveSchemaCase" value="true"/>

Liquibase is uppercasing the reserved word key before quoting it in order to have it be treated case-insensitively. If you still want the key column to be lower-case, and therefore case-sensitive, then use something like this.

  <changeSet …>
    …
    <modifySql>
      <regExpReplace replace="&quot;KEY&quot;" with="&quot;key&quot;"/>
    </modifySql>
  </changeSet>

If you want to preserve the case of every identifier, then use someting like this.

  <changeSet … objectQuotingStrategy="QUOTE_ALL_OBJECTS">
    …
  </changeSet>

[shrralis]

@mches thanks for providing solutions!
The option with modifySql works great for me!
Anyway, it looks like it doesn't work for loadData, right?

[mches]

@mches thanks for providing solutions! The option with modifySql works great for me! Anyway, it looks like it doesn't work for loadData, right?

@shrralis Yes, whether a bug or undocumented limitation, Liquibase doesn't seem to apply SQL visitors like modifySql to prepared statements like it does to raw SQL statements. The loadData change often ends up implicitly using prepared statements.

Knowing that limitiation, you may be able to work around it with something like this, but in some cases prepared statements are required.

  <changeSet …>
    <loadData … usePreparedStatements="false"/>
    <modifySql>
      <regExpReplace replace="&quot;KEY&quot;" with="&quot;key&quot;"/>
    </modifySql>
  </changeSet>

You could also use object quoting strategy, but then also uppercase the CSV file's headers other than the illegal identifiers.

  <changeSet … objectQuotingStrategy="QUOTE_ALL_OBJECTS">
    <loadData …/>
  </changeSet>

key,FIRST,LAST
0,john,doe
1,eric,smith
2,cat,jones