Migrate: Duplicate entries in url_alias after update migrations

Question

I have term and node migrations from wprdpress to drupal 8 (with pathauto module enabled). I set the url alias in the migration templates like:

process:
  path/alias: source_path
  path/pathauto:
    plugin: default_value
    default_value: 0

Setting the path that way works fine on importing content with drush mi term. But when I update the terms with drush mi term --update a new (the same) path gets inserted into the url_alias table. For example the term with tid 1 gets imported once and updated 2 times there will be the following aliases in url_alias table:

select * from url_alias where source = '/taxonomy/term/1';
+------+------------------+---------------+----------+
| pid  | source           | alias         | langcode |
+------+------------------+---------------+----------+
| 1514 | /taxonomy/term/1 | /term-1-alias | de       |
| 2731 | /taxonomy/term/1 | /term-1-alias | de       |
| 3632 | /taxonomy/term/1 | /term-1-alias | de       |
+------+------------------+---------------+----------+

BTW: I have pathauto module enabled, but uninstalling it results in the same behavior.

Answer 1

Looks like a bug in the core path module: https://www.drupal.org/node/2350135

Answer 2

I made the following workaround to avoid the problem:

create a custom process plugin that checks if an url_alias record was already inserted for the $value / node combination. If so, it returns NULL and if not it returns $value:

<?php
/**
 * @file
 * Contains \Drupal\your_migrate_module\Plugin\migrate\process\EmptyIfUrlAliasExists.
 *
 */

namespace Drupal\your_migrate_module\Plugin\migrate\process;

use Drupal\Core\Database\Database;
use Drupal\migrate\Annotation\MigrateProcessPlugin;
use Drupal\migrate\MigrateExecutableInterface;
use Drupal\migrate\ProcessPluginBase;
use Drupal\migrate\Row;

/**
 * This process plugin can be used for path/alias fields as long as the following issues are not resolved. It is a workaround for:
 *
 * https://www.drupal.org/node/2350135#comment-9476629
 * https://drupal.stackexchange.com/questions/238393/migrate-duplicate-entries-in-url-alias-after-update-migrations)
 *
 * @MigrateProcessPlugin(
 *   id = "empty_if_url_alias_exists",
 * )
 */
class EmptyIfUrlAliasExists extends ProcessPluginBase
{
    /**
     * {@inheritdoc}
     */
    public function transform(
        $value,
        MigrateExecutableInterface $migrate_executable,
        Row $row,
        $destination_property
    ) {
        // Retrieves a \Drupal\Core\Database\Connection which is a PDO instance
        $db = Database::getConnection();

        $sth = $db->select('url_alias', 'u')
            ->fields('u', ['pid']);
        $and = $sth->andConditionGroup()
            ->condition('u.source', '/node/' . $row->getIdMap()['destid1'])
            ->condition('u.alias', $value);
        $sth->condition($and);
        $data = $sth->execute();

        $results = $data->fetch(\PDO::FETCH_NUM);

        //when no url_alias record found, return the url, so it can be added.
        if ($results === false || count($results) === 0) {
            return $value;
        }
        //when an url_alias record is already present, return null, so the migration does not add it again
        return null;
    }
}

Then in the process part of my node migration yml, I use the following:

process:
  'path/pathauto': 
     plugin: default_value
     default_value: 0 # Disable pathauto.
  'path/alias':
    -
      plugin: empty_if_url_alias_exists
      source: public_url
    -
      plugin: skip_on_empty
      method: process