LocalUsernamesToLowerCase: manage duplication automatically

When having two external-ids that are different just in their
case, the LocalUsernamesToLowerCase should automatically remove
one of them and only keep the lowercase version.

Example for the same account id:
  gerrit:MyUsername
  gerrit:MyUSERNAME

The above two external-ids would have caused
a DuplicateExternalIdKeyException which is not helpful as
both identities are the same and associated with the same account id
and have the same associated property values:
- email
- password

Automatically remove one of the two (or more) identities in
conflict and display a warning message explaining what has just
happened and how the situation has been managed transparently.

NOTE: The duplicates that are getting removed are identical
once they are converted to lowercase. The ones that are getting deleted
are not relevant: the only thing that is important is that only one of
them is kept, as all the others are identical.

To test this change:

Positive use-case:
--

Before the LocalUsernamesToLowerCase execution:

[externalId "gerrit:johndoe"]
    accountId = 1000000
    email = john@doe.org

[externalId "gerrit:JohnDoe"]
    accountId = 1000000
    email = JOHN@doe.org

After the LocalUsernamesToLowerCase execution, there is a single entry:

[externalId "gerrit:johndoe"]
    accountId = 1000000
    email = john@doe.org

OR

[externalId "gerrit:johndoe"]
    accountId = 1000000
    email = JOHN@doe.org

Negative use-cases:

[externalId "gerrit:johndoe"]
    accountId = 1000000

[externalId "gerrit:JohnDoe"]
    accountId = 1000000
    email = JOHN@doe.org

The LocalUsernamesToLowerCase execution fails with a
duplicated ids exception and no changes are made on the
external ids.

[externalId "gerrit:johndoe"]
    accountId = 1000000

[externalId "gerrit:JohnDoe"]
    accountId = 1000001

The LocalUsernamesToLowerCase execution fails with a
duplicated ids exception and no changes are made on the
external ids.

Bug: Issue 14645
Change-Id: I9832fbf0c8df947aab5494325169b9c726e40019

3 files changed, 39 insertions, 1 deletions

diff --git a/Documentation/pgm-LocalUsernamesToLowerCase.txt b/Documentation/pgm-LocalUsernamesToLowerCase.txt
index 53081a18ba..a526647624 100644
--- a/Documentation/pgm-LocalUsernamesToLowerCase.txt
+++ b/Documentation/pgm-LocalUsernamesToLowerCase.txt
@@ -28,10 +28,14 @@ Web UI.
 Please be aware that the conversion of the local usernames to lower
 case can't be undone.
 
-The program will produce errors if there are accounts that have the
+The program will produce errors if there are accounts with a different
+account-id or other properties (e.g. email, password) that have the
 same local username, but with different case. In this case the local
 username for these accounts is not converted to lower case.
 
+The program will automatically remove duplicates where the username
+differs only in case but all other attributes are identical.
+
 After all usernames have been migrated, the link:pgm-reindex.html[
 reindex] program is automatically invoked to reindex all accounts.
 
diff --git a/java/com/google/gerrit/pgm/BUILD b/java/com/google/gerrit/pgm/BUILD
index a57b37a505..2b1b83d643 100644
--- a/java/com/google/gerrit/pgm/BUILD
+++ b/java/com/google/gerrit/pgm/BUILD
@@ -46,6 +46,7 @@ java_library(
         "//lib:servlet-api-without-neverlink",
         "//lib/auto:auto-value",
         "//lib/auto:auto-value-annotations",
+        "//lib/commons:lang",
         "//lib/flogger:api",
         "//lib/guice",
         "//lib/guice:guice-assistedinject",
diff --git a/java/com/google/gerrit/pgm/LocalUsernamesToLowerCase.java b/java/com/google/gerrit/pgm/LocalUsernamesToLowerCase.java
index e6e091c360..8e2f70f5f0 100644
--- a/java/com/google/gerrit/pgm/LocalUsernamesToLowerCase.java
+++ b/java/com/google/gerrit/pgm/LocalUsernamesToLowerCase.java
@@ -36,6 +36,9 @@ import com.google.inject.Provider;
 import java.io.IOException;
 import java.util.Collection;
 import java.util.Locale;
+import java.util.Optional;
+import org.apache.commons.lang.StringUtils;
+import org.eclipse.jgit.errors.ConfigInvalidException;
 import org.eclipse.jgit.lib.ProgressMonitor;
 import org.eclipse.jgit.lib.Repository;
 import org.eclipse.jgit.lib.TextProgressMonitor;
@@ -108,8 +111,38 @@ public class LocalUsernamesToLowerCase extends SiteProgram {
                 extId.accountId(),
                 extId.email(),
                 extId.password());
+        replaceIfNotExists(extIdNotes, extId, extIdLowerCase);
+      }
+    }
+  }
+
+  private void replaceIfNotExists(
+      ExternalIdNotes extIdNotes, ExternalId extId, ExternalId extIdLowerCase) throws IOException {
+    try {
+      Optional<ExternalId> existingExternalId =
+          extIdNotes
+              .get(extIdLowerCase.key())
+              .filter(eid -> eid.accountId().equals(extIdLowerCase.accountId()))
+              .filter(eid -> StringUtils.equalsIgnoreCase(eid.email(), extId.email()))
+              .filter(eid -> StringUtils.equalsIgnoreCase(eid.password(), extId.password()));
+      if (existingExternalId.isPresent()) {
+        System.err.println(
+            "WARNING: external-id "
+                + extIdLowerCase
+                + " already exists with the same account-id "
+                + extId.accountId()
+                + " :"
+                + "removing the duplicate external-id "
+                + extId.key());
+        extIdNotes.delete(extId);
+      } else {
         extIdNotes.replace(extId, extIdLowerCase);
       }
+    } catch (ConfigInvalidException e) {
+      throw new IOException(
+          "Unable to parse external id definition when looking for current external-id "
+              + extIdLowerCase,
+          e);
     }
   }