leaves12138 commented on code in PR #14:
URL: 
https://github.com/apache/terraform-provider-paimon/pull/14#discussion_r3893393841


##########
internal/provider/table_schema_changes.go:
##########
@@ -0,0 +1,306 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information regarding
+// copyright ownership. The ASF licenses this file to You under the Apache
+// License, Version 2.0 (the "License"); you may not use this file except in
+// compliance with the License. You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package provider
+
+import (
+       "errors"
+       "fmt"
+       "slices"
+       "strings"
+
+       "github.com/apache/terraform-provider-paimon/internal/client"
+)
+
+func assignTemporaryIDsToNewFields(before []client.Field, planned 
[]tableFieldModel, after []client.Field) error {
+       if len(planned) != len(after) {
+               return errors.New("the planned field models and converted 
schema have different lengths")
+       }
+
+       used := make(map[int]struct{})
+       next := 0
+       reserve := func(id int) {
+               used[id] = struct{}{}
+               if id >= next {
+                       next = id + 1
+               }
+       }
+       for _, field := range before {
+               reserve(field.ID)
+               for _, nestedID := range field.NestedFieldIDs {
+                       reserve(nestedID)
+               }
+       }
+       for index, field := range planned {
+               if !field.ID.IsNull() && !field.ID.IsUnknown() {
+                       reserve(int(field.ID.ValueInt64()))
+               }
+               for _, nestedID := range after[index].NestedFieldIDs {
+                       reserve(nestedID)
+               }
+       }
+
+       for index, field := range planned {
+               if !field.ID.IsNull() && !field.ID.IsUnknown() {
+                       continue
+               }
+               for {
+                       if _, exists := used[next]; !exists {
+                               break
+                       }
+                       next++
+               }
+               if next > maxPaimonFieldID {
+                       return errors.New("no Paimon field IDs remain for a 
newly added field")
+               }
+               after[index].ID = next
+               reserve(next)
+       }
+
+       return nil
+}
+
+func tableFieldSchemaChanges(before, after []client.Field, addBeforePartition 
bool, partitionKeys []string) ([]client.SchemaChange, error) {
+       beforeByID, err := fieldsByID(before)
+       if err != nil {
+               return nil, err
+       }
+       afterByID, err := fieldsByID(after)
+       if err != nil {
+               return nil, err
+       }
+
+       changes := make([]client.SchemaChange, 0)
+       for _, previous := range before {
+               planned, exists := afterByID[previous.ID]
+               if !exists {
+                       continue
+               }
+               previousType, previousNullable := splitFieldType(previous.Type)
+               plannedType, plannedNullable := splitFieldType(planned.Type)
+               path := []string{previous.Name}
+               if !client.EquivalentDataTypes(previousType, plannedType) {
+                       plannedTypeField := planned
+                       plannedTypeField.Type = plannedType
+                       changes = append(changes, client.SchemaChange{
+                               "action":          "updateColumnType",
+                               "fieldNames":      path,
+                               "newDataType":     
schemaChangeDataType(plannedTypeField, before, previous.ID),
+                               "keepNullability": true,
+                       })
+               }
+               if previousNullable != plannedNullable {
+                       changes = append(changes, client.SchemaChange{
+                               "action":         "updateColumnNullability",
+                               "fieldNames":     path,
+                               "newNullability": plannedNullable,
+                       })
+               }
+               if !stringPointersEqual(previous.Description, 
planned.Description) {
+                       changes = append(changes, client.SchemaChange{
+                               "action":     "updateColumnComment",
+                               "fieldNames": path,
+                               "newComment": planned.Description,
+                       })
+               }
+               if !stringPointersEqual(previous.DefaultValue, 
planned.DefaultValue) {
+                       changes = append(changes, client.SchemaChange{
+                               "action":          "updateColumnDefaultValue",
+                               "fieldNames":      path,
+                               "newDefaultValue": planned.DefaultValue,
+                       })
+               }
+       }
+
+       for _, previous := range before {
+               if _, exists := afterByID[previous.ID]; !exists {
+                       changes = append(changes, client.SchemaChange{"action": 
"dropColumn", "fieldNames": []string{previous.Name}})
+               }
+       }
+
+       retainedNames := make(map[string]int)
+       pendingRenames := make(map[string]client.Field)
+       for _, previous := range before {
+               if planned, exists := afterByID[previous.ID]; exists {
+                       retainedNames[previous.Name] = planned.ID
+                       if previous.Name != planned.Name {
+                               pendingRenames[previous.Name] = planned
+                       }
+               }
+       }
+       for len(pendingRenames) > 0 {
+               progress := false
+               for _, previous := range before {
+                       planned, pending := pendingRenames[previous.Name]
+                       if !pending {
+                               continue
+                       }
+                       if conflictingID, conflict := 
retainedNames[planned.Name]; conflict && conflictingID != planned.ID {
+                               continue
+                       }
+                       changes = append(changes, client.SchemaChange{
+                               "action":     "renameColumn",
+                               "fieldNames": []string{previous.Name},
+                               "newName":    planned.Name,
+                       })
+                       delete(retainedNames, previous.Name)
+                       retainedNames[planned.Name] = planned.ID
+                       delete(pendingRenames, previous.Name)
+                       progress = true
+               }
+               if !progress {
+                       return nil, errors.New("cannot apply a cycle of table 
field renames in one apply; use an intermediate name")
+               }
+       }
+
+       for _, planned := range after {
+               if _, exists := beforeByID[planned.ID]; exists {
+                       continue
+               }
+               changes = append(changes, client.SchemaChange{
+                       "action":     "addColumn",
+                       "fieldNames": []string{planned.Name},
+                       "dataType":   schemaChangeDataType(planned, before, -1),
+                       "comment":    planned.Description,
+               })
+               if planned.DefaultValue != nil {
+                       changes = append(changes, client.SchemaChange{
+                               "action":          "updateColumnDefaultValue",
+                               "fieldNames":      []string{planned.Name},
+                               "newDefaultValue": planned.DefaultValue,
+                       })
+               }
+       }
+
+       currentOrder := make([]string, 0, len(after))
+       for _, previous := range before {
+               if planned, exists := afterByID[previous.ID]; exists {
+                       currentOrder = append(currentOrder, planned.Name)
+               }
+       }
+       partitionKeySet := make(map[string]struct{}, len(partitionKeys))
+       for _, name := range partitionKeys {
+               partitionKeySet[name] = struct{}{}
+       }
+       for _, planned := range after {
+               if _, exists := beforeByID[planned.ID]; !exists {
+                       insertAt := len(currentOrder)
+                       if addBeforePartition {
+                               for index, name := range currentOrder {
+                                       if _, partitionKey := 
partitionKeySet[name]; partitionKey {
+                                               insertAt = index
+
+                                               break
+                                       }
+                               }
+                       }
+                       currentOrder = insertString(currentOrder, insertAt, 
planned.Name)
+               }
+       }
+       for index, planned := range after {
+               if _, retained := beforeByID[planned.ID]; retained {
+                       continue
+               }
+               currentIndex := slices.Index(currentOrder, planned.Name)
+               if currentIndex == index {
+                       continue
+               }
+               changes = append(changes, columnPositionChange(after, index))
+               currentOrder = moveString(currentOrder, currentIndex, index)

Review Comment:
   `moveString` models this operation as a move to a numeric index, but Paimon 
applies the emitted `AFTER` move relative to the reference field's current 
position. Those are not equivalent when the reference field has not been 
reordered yet. For example, start with fields `[a,b,c,d]`, partition keys 
`[c,d]`, and `add-column-before-partition=true`; plan `[a,b,d,x,c,y]` where `x` 
and `y` are new. The two adds produce `[a,b,x,y,c,d]`, and this code emits `x 
AFTER d`, `y AFTER c`, then `d AFTER b`. Paimon ends at `[a,b,d,c,y,x]`, not 
the planned order, so reconciliation fails. Please simulate each emitted move 
using its actual FIRST/AFTER reference semantics (or order moves so the 
reference prefix is already stable), and cover two additions combined with 
retained-field reordering.



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to