Issue #2814 : FORMAT date locale, log effective regional settings

ValueMetaBase copied Locale.getDefault() into every field, which is the
interface language once DISPLAY and FORMAT are split. New fields now take
the FORMAT category so month names follow regional settings, not the GUI
language.

hop-gui, hop-run and hop-server log the effective language, FORMAT locale
and timezone at start so a machine-local mismatch is visible.

Docs: field-level symbols still win; default OS source still differs
across machines unless Override is set; Beam/Spark workers do not load
hop-config.json.

Author: mattcasters

core/src/main/java/org/apache/hop/core/row/value/ValueMetaBase.java

@@ -316,7 +316,10 @@ protected ValueMetaBase(
     this.decimalSymbol = "" + Const.getDefaultDecimalSeparator();
     this.groupingSymbol = "" + Const.getDefaultGroupingSeparator();
     this.currencySymbol = "" + Const.getDefaultCurrencySymbol();
-    this.dateFormatLocale = Locale.getDefault();
+    // FORMAT, not Locale.getDefault(): the latter is the interface language once DISPLAY and
+    // FORMAT are split, and a field with no explicit date locale must follow the regional
+    // settings rather than the GUI language.
+    this.dateFormatLocale = Locale.getDefault(Locale.Category.FORMAT);
     this.collatorDisabled = true;
     this.collatorLocale = Locale.getDefault();
     this.collator = Collator.getInstance(this.collatorLocale);

core/src/main/java/org/apache/hop/i18n/RegionalSettings.java

@@ -19,7 +19,9 @@
 
 import java.util.Arrays;
 import java.util.Locale;
+import java.util.TimeZone;
 import org.apache.hop.core.config.HopConfig;
+import org.apache.hop.core.logging.ILogChannel;
 import org.apache.hop.core.logging.LogChannel;
 import org.apache.hop.core.util.EnvUtil;
 import org.apache.hop.core.util.Utils;
@@ -141,6 +143,7 @@ public void applyHeadless() {
       return;
     }
     Locale.setDefault(Locale.Category.FORMAT, formatLocale);
+    logEffective(LogChannel.GENERAL, "installation:" + source.name());
   }
 
   /**
@@ -160,6 +163,31 @@ public void applyGui() {
       return;
     }
     Locale.setDefault(Locale.Category.FORMAT, formatLocale);
+    logEffective(LogChannel.GENERAL, "installation:" + source.name());
+  }
+
+  /**
+   * Writes the language, FORMAT locale and default timezone currently in effect. hop-gui, hop-run
+   * and hop-server all log this so a machine-local mismatch is visible without inspecting
+   * configuration files.
+   *
+   * @param log channel to write to; {@link LogChannel#GENERAL} when none is available yet
+   * @param sourceDescription where the FORMAT locale came from, for example {@code
+   *     installation:CUSTOM}
+   */
+  public static void logEffective(ILogChannel log, String sourceDescription) {
+    if (log == null) {
+      return;
+    }
+    log.logBasic(
+        "Regional settings: language="
+            + Locale.getDefault()
+            + " format="
+            + Locale.getDefault(Locale.Category.FORMAT)
+            + " timezone="
+            + TimeZone.getDefault().getID()
+            + " source="
+            + sourceDescription);
   }
 
   private static boolean isUsable(Locale locale) {

core/src/test/java/org/apache/hop/core/row/value/ValueMetaDateFormatLocaleTest.java

@@ -84,6 +84,20 @@ void withoutAnExplicitDateLocaleTheRegionalSettingsWin() {
     assertEquals("gennaio 2025", valueMeta.getDateFormat().format(DATE));
   }
 
+  /**
+   * A newly constructed field must pick FORMAT as its date locale. Using DISPLAY (the interface
+   * language) here would format month names in English while numbers followed Italian separators.
+   */
+  @Test
+  void constructorDateLocaleFollowsTheRegionalSettingsNotTheLanguage() {
+    languageAndRegionalSettings(Locale.US, Locale.ITALY);
+
+    ValueMetaDate valueMeta = new ValueMetaDate("d");
+    valueMeta.setConversionMask(MASK);
+
+    assertEquals("gennaio 2025", valueMeta.getDateFormat().format(DATE));
+  }
+
   @Test
   void timestampExplicitDateLocaleIsHonouredEvenWhenItEqualsTheInterfaceLanguage() {
     languageAndRegionalSettings(Locale.ITALY, Locale.US);
@@ -105,4 +119,14 @@ void timestampWithoutAnExplicitDateLocaleFollowsTheRegionalSettings() {
 
     assertEquals("gennaio 2025", valueMeta.getDateFormat().format(DATE));
   }
+
+  @Test
+  void timestampConstructorDateLocaleFollowsTheRegionalSettingsNotTheLanguage() {
+    languageAndRegionalSettings(Locale.US, Locale.ITALY);
+
+    ValueMetaTimestamp valueMeta = new ValueMetaTimestamp("t");
+    valueMeta.setConversionMask(MASK);
+
+    assertEquals("gennaio 2025", valueMeta.getDateFormat().format(DATE));
+  }
 }

docs/hop-user-manual/modules/ROOT/pages/hop-gui/perspective-configuration.adoc

@@ -141,7 +141,9 @@ a|use an explicitly chosen locale for number, currency and date formats, regardl
 
 Two preview panels below the options show sample dates, times, numbers and currency for the current selection, updated live as options change, before anything is saved.
 
-Whichever source is in effect, operating system or an explicit override, applies both to Hop GUI and to headless runs: `hop-run` and `hop-server` read the regional configuration of the machine they run on and honor it the same way the GUI does.
+Whichever source is in effect, operating system or an explicit override, applies both to Hop GUI and to headless runs: `hop-run` and `hop-server` read the same `hop-config.json` and honor it the same way the GUI does.
+
+At start, hop-gui, `hop-run` and `hop-server` log the effective settings, for example `Regional settings: language=en_US format=it_IT timezone=Europe/Rome source=installation:CUSTOM`. Use that line to confirm the process you are looking at is formatting the way you think it is.
 
 NOTE: Because this setting applies to the whole running process, in Hop Web it belongs to the installation rather than to each connected user. This is already true of the preferred language today, so it is not a new limitation introduced by this feature. Saving this tab applies the preferred language to the whole running process as well, so in Hop Web it affects every connected session until the server restarts.
 
@@ -174,7 +176,15 @@ On an existing installation whose configuration is left untouched, nothing chang
 
 The reason to untick both options is reproducibility against the preferred language: it makes Hop GUI, `hop-run` and `hop-server` agree on that language, so a number converted to a string reads the same in the editor and in production, and any hash computed over such a string stays stable between them.
 
-The setting belongs to the Hop installation, not to a project or a pipeline, and it applies to the whole running process rather than to an individual project or session. In Hop Web that means the whole server, and therefore every connected session, shares one set of regional settings, updated live whenever the configuration tab is saved.
+Resolution order, most specific first:
+
+. Field-level decimal, grouping, currency, date locale and date timezone on a transform field. These always win.
+. This installation tab (operating system, an explicit override, or the preferred language).
+. The JVM / operating system default.
+
+The default source is the operating system. Hop GUI, `hop-run` and `hop-server` then agree *on the same machine*, but a laptop on `nl_BE` and a hop-server on `en_US` still format blank fields differently. *Override regional settings* (and copying that `hop-config.json`) is the way to pin a locale for an installation. Transform field symbols remain the way to handle a file that does not match that locale.
+
+Distributed Beam and Spark workers never load `hop-config.json`. They format with their own JVM locale and timezone unless the pipeline sets field-level symbols.
 
 === Plugins