join: compare a missing join field as an empty one

src/uu/join/src/join.rs

@@ -320,25 +320,20 @@ impl<Sep: Separator> Input<Sep> {
         }
     }
 
+    /// Compare two join fields. Like `sort -k`, a field missing from a line
+    /// that is too short compares as an empty field.
     fn compare(&self, field1: Option<&[u8]>, field2: Option<&[u8]>) -> Ordering {
-        if let (Some(field1), Some(field2)) = (field1, field2) {
-            if self.ignore_case {
-                let field1 = CaseInsensitiveSlice { v: field1 };
-                let field2 = CaseInsensitiveSlice { v: field2 };
-                field1.cmp(&field2)
-            } else if self.use_locale {
-                locale_cmp(field1, field2)
-            } else {
-                field1.cmp(field2)
-            }
+        let field1 = field1.unwrap_or_default();
+        let field2 = field2.unwrap_or_default();
+
+        if self.ignore_case {
+            let field1 = CaseInsensitiveSlice { v: field1 };
+            let field2 = CaseInsensitiveSlice { v: field2 };
+            field1.cmp(&field2)
+        } else if self.use_locale {
+            locale_cmp(field1, field2)
         } else {
-            match field1 {
-                Some(_) => Ordering::Greater,
-                None => match field2 {
-                    Some(_) => Ordering::Less,
-                    None => Ordering::Equal,
-                },
-            }
+            field1.cmp(field2)
         }
     }
 }

tests/by-util/test_join.rs

@@ -911,3 +911,29 @@ fn test_operand_count() {
         .stderr_contains("join: extra operand 'fields_3.txt'\n")
         .stderr_contains("--help");
 }
+
+#[test]
+fn test_missing_key_is_empty_key() {
+    // Like `sort -k`, a line too short to have the join field has an empty key.
+    let ts = TestScenario::new(util_name!());
+    let at = &ts.fixtures;
+    at.write("short", "x,\ny\nbob,1002\n");
+    at.write("keys", "1002\n");
+    ts.ucmd()
+        .args(&["-t", ",", "-1", "2", "--check-order", "short", "keys"])
+        .succeeds()
+        .stdout_only("1002,bob\n");
+
+    // ...and pairs with lines whose join field is empty.
+    ts.ucmd()
+        .args(&["-t", ",", "-j", "2", "short", "short"])
+        .succeeds()
+        .stdout_only(",x,x\n,x,y\n,y,x\n,y,y\n1002,bob,bob\n");
+
+    // Same with the default blank separator and a trailing blank.
+    at.write("blank", "a \nb\nbob 1002\n");
+    ts.ucmd()
+        .args(&["-1", "2", "--check-order", "blank", "keys"])
+        .succeeds()
+        .stdout_only("1002 bob\n");
+}