fix: convert duplicate YAML merge key panic to diagnostic error (#5188)

Better error for #5182

Author: denik

acceptance/bundle/validate/duplicate_yaml_merge_key/databricks.yml

@@ -0,0 +1,20 @@
+bundle:
+  name: test-bundle
+
+definitions:
+  cluster1: &cluster1
+    num_workers: 1
+  cluster2: &cluster2
+    spark_version: "13.3.x-scala2.12"
+
+resources:
+  jobs:
+    my_job:
+      name: "test job"
+      tasks:
+        - task_key: "main"
+          new_cluster:
+            <<: *cluster1
+            <<: *cluster2
+          notebook_task:
+            notebook_path: "/notebook"

acceptance/bundle/validate/duplicate_yaml_merge_key/out.test.toml

@@ -0,0 +1,20 @@
+
+>>> [CLI] bundle validate
+Error: duplicate YAML merge key ('<<') is not allowed; to merge multiple maps, use a sequence: '<<: [*anchor1, *anchor2]'
+  in databricks.yml:18:13
+
+
+Found 1 error
+
+>>> [CLI] bundle validate
+Name: test-bundle
+Target: default
+Workspace:
+  User: [USERNAME]
+  Path: /Workspace/Users/[USERNAME]/.bundle/test-bundle/default
+
+Validation OK!
+{
+  "num_workers": 1,
+  "spark_version": "13.3.x-scala2.12"
+}

acceptance/bundle/validate/duplicate_yaml_merge_key/output.txt

@@ -0,0 +1,8 @@
+musterr trace $CLI bundle validate
+
+update_file.py databricks.yml "            <<: *cluster1
+            <<: *cluster2" "            <<: [*cluster1, *cluster2]"
+
+trace $CLI bundle validate
+
+$CLI bundle validate -o json | jq '.resources.jobs.my_job.tasks[0].new_cluster'

acceptance/bundle/validate/duplicate_yaml_merge_key/script

@@ -3,6 +3,7 @@ package config
 import (
 	"bytes"
 	"context"
+	"errors"
 	"fmt"
 	"os"
 	"reflect"
@@ -106,6 +107,14 @@ func LoadFromBytes(path string, raw []byte) (*Root, diag.Diagnostics) {
 	// Load configuration tree from YAML.
 	v, err := yamlloader.LoadYAML(path, bytes.NewBuffer(raw))
 	if err != nil {
+		var le *yamlloader.LocationError
+		if errors.As(err, &le) {
+			return nil, diag.Diagnostics{{
+				Severity:  diag.Error,
+				Summary:   le.Summary,
+				Locations: []dyn.Location{le.Loc},
+			}}
+		}
 		return nil, diag.Errorf("failed to load %s: %v", path, err)
 	}
 

bundle/config/root.go

@@ -10,6 +10,17 @@ import (
 	"go.yaml.in/yaml/v3"
 )
 
+// LocationError is an error with a YAML source location that can be displayed
+// to the user with a file path, line, and column number.
+type LocationError struct {
+	Loc     dyn.Location
+	Summary string
+}
+
+func (e *LocationError) Error() string {
+	return fmt.Sprintf("yaml (%s): %s", e.Loc, e.Summary)
+}
+
 type loader struct {
 	path string
 }
@@ -110,7 +121,12 @@ func (d *loader) loadMapping(node *yaml.Node, loc dyn.Location) (dyn.Value, erro
 			// However, when used as a key, it is treated as the string "null".
 		case "!!merge":
 			if merge != nil {
-				panic("merge node already set")
+				// The YAML merge key spec allows a single '<<' key per mapping.
+				// To merge multiple maps, use a sequence: '<<: [*anchor1, *anchor2]'.
+				return dyn.InvalidValue, &LocationError{
+					Loc:     d.location(key),
+					Summary: "duplicate YAML merge key ('<<') is not allowed; to merge multiple maps, use a sequence: '<<: [*anchor1, *anchor2]'",
+				}
 			}
 			merge = val
 			continue