Drupal Search API Patterns

This skill documents Search API configuration patterns, boosting strategies, and custom processor development for Drupal sites.

Index Configuration

Field Type Requirements

Boolean Fields for Boosting

• Boolean fields work best with custom boost processors
• Integer fields can cause issues with boolean-based boost logic
• Always configure boolean fields as type: boolean in index settings

field_settings:
  field_featured:
    label: Featured
    datasource_id: 'entity:node'
    property_path: field_featured
    type: boolean  # NOT integer
    boost: 8.0

Configuration Command

ddev drush config:set search_api.index.{index_name} \
  field_settings.field_featured.type boolean

Numeric Fields for Engagement Boosting

Flag Count Fields

• Use type: integer for count fields
• Managed by flag_search_api module
• Automatically indexed and updated

field_settings:
  flag_bookmark_count:
    label: 'Bookmark count'
    property_path: flag_bookmark_count
    type: integer
  flag_favorite_count:
    label: 'Favorite count'
    property_path: flag_favorite_count
    type: integer

Boost Processors

Custom Boolean Boost Processor

When to Use

• Boolean field boosting (featured flags, promoted content)
• Needs 100% control over boost logic
• Complex conditional boosting

Pattern: FeaturedContentBoost.php

<?php

namespace Drupal\custom_search\Plugin\search_api\processor;

use Drupal\search_api\Item\ItemInterface;
use Drupal\search_api\Processor\ProcessorPluginBase;

/**
 * @SearchApiProcessor(
 *   id = "featured_content_boost",
 *   label = @Translation("Featured content boost"),
 *   description = @Translation("Adds a boost to indexed items marked as featured."),
 *   stages = {
 *     "preprocess_index" = 0,
 *   },
 *   locked = false,
 *   hidden = false,
 * )
 */
class FeaturedContentBoost extends ProcessorPluginBase {

  /**
   * {@inheritdoc}
   */
  public function preprocessIndexItems(array $items) {
    /** @var \Drupal\search_api\Item\ItemInterface $item */
    foreach ($items as $item) {
      try {
        $entity = $item->getOriginalObject()->getValue();

        // Check if entity has featured field and it's set to TRUE.
        if ($entity->hasField('field_featured') &&
            !$entity->get('field_featured')->isEmpty() &&
            $entity->get('field_featured')->value == 1) {
          $old_boost = $item->getBoost();
          // Apply 2x boost to featured content.
          $item->setBoost($old_boost * 2.0);
        }
      }
      catch (\Exception $e) {
        // Skip items that can't be loaded.
        continue;
      }
    }
  }
}

Key Points

• Use multiplicative boost: $item->setBoost($old_boost * boost_factor)
• Pattern from search_api_boolean_field_boost module
• Boost at index time via preprocess_index stage
• Always get old boost first to preserve other boosts

Recommended Boost Factors

• Featured content: 2.0x (modest but effective)
• Featured content: 1.5x - 3.0x
• Premium content: 1.5x - 2.0x

Number Field Boost Processor

When to Use

• Engagement metrics (likes, bookmarks, views)
• Numeric quality scores
• Time-based decay factors

Configuration Pattern

processor_settings:
  number_field_boost:
    weights:
      preprocess_index: 0
    boosts:
      flag_bookmark_count:
        boost_factor: 0.01
        aggregation: max
      flag_favorite_count:
        boost_factor: 0.1
        aggregation: max

How It Works

• Adds to boost based on field value
• Formula: boost += (field_value * boost_factor)
• Example: 138 bookmarks x 0.01 = +1.38 boost

Configuration Commands

# Set bookmark count boost (0.01 per bookmark)
ddev drush config:set search_api.index.{index_name} \
  processor_settings.number_field_boost.boosts.flag_bookmark_count.boost_factor 0.01

# Set favorite count boost (0.1 per favorite)
ddev drush config:set search_api.index.{index_name} \
  processor_settings.number_field_boost.boosts.flag_favorite_count.boost_factor 0.1

Recommended Boost Factors

• Bookmark count: 0.01 - 0.05 (for counts in 10-1000 range)
• Favorite count: 0.1 - 0.5 (for counts in 1-50 range)
• View count: 0.001 - 0.01 (for counts in 100-10000 range)

Processor Conflicts

Problem: Multiple Processors on Same Field

If multiple processors target the same field, they can conflict:

• One overrides the other
• Boosts don't combine as expected
• Unpredictable results

Solution: Remove Conflicting Configuration

# Remove field from number_field_boost if using custom processor
ddev drush config:set search_api.index.{index_name} \
  processor_settings.number_field_boost.boosts.field_featured null

Best Practice

• Use custom processor for boolean/complex logic
• Use number_field_boost for simple numeric boosts
• Don't configure same field in multiple processors

Configuration Management

Direct Config Updates with PHP

When drush config:set fails or produces unexpected results (e.g., side effects on unrelated config), use PHP to directly update active configuration:

ddev drush php:eval "
\$config = \Drupal::configFactory()->getEditable('search_api.index.{index_name}');

// Set field type
\$config->set('field_settings.field_featured.type', 'boolean');

// Remove from number_field_boost
\$boosts = \$config->get('processor_settings.number_field_boost.boosts');
unset(\$boosts['field_featured']);
\$config->set('processor_settings.number_field_boost.boosts', \$boosts);

// Set boost factors
\$config->set('processor_settings.number_field_boost.boosts.flag_bookmark_count.boost_factor', 0.01);
\$config->set('processor_settings.number_field_boost.boosts.flag_favorite_count.boost_factor', 0.1);

\$config->save();
echo \"Configuration updated\n\";
"

When to Use PHP Instead of drush config:set:

• When config:set creates unwanted side effects (e.g., changing server: {server_name} to server: null)
• When removing array keys (unset pattern works better than null)
• When making multiple related changes atomically
• When field type changes cause Drupal to fallback to generic types

After PHP Config Updates:

1. Verify changes: ddev drush config:get search_api.index.{index_name} field_settings.field_featured
2. Export that one object: ddev drush config:get search_api.index.{index_name} --format=yaml > config/default/search_api.index.{index_name}.yml (a blanket config:export also sweeps unrelated drift; see the drupal-config-mgmt skill)
3. Review exported changes: git diff config/default/
4. Revert any unintended changes (e.g., ngram fields changed to plain text)

Field Type Preservation

Problem: When Solr server config is modified (e.g., pointing to a different backend), config export may downgrade custom field types to generic types:

# BEFORE (correct)
field_display_name:
  type: 'solr_text_custom:ngramstring'

# AFTER export (incorrect)
field_display_name:
  type: text

Solution: After exporting, restore custom field types via PHP:

ddev drush php:eval "
\$config = \Drupal::configFactory()->getEditable('search_api.index.{index_name}');
\$config->set('field_settings.field_display_name.type', 'solr_text_custom:ngramstring');
\$config->set('field_settings.label.type', 'solr_text_custom:ngramstring');
\$config->set('field_settings.name.type', 'solr_text_custom:ngramstring');
\$config->set('field_settings.title.type', 'solr_text_custom:ngramstring');
\$config->save();
"
ddev drush config:get search_api.index.{index_name} --format=yaml > config/default/search_api.index.{index_name}.yml

Fields Using ngram Tokenization:

• field_display_name - Display names (partial matching for autocomplete)
• label - Entity labels/titles
• name - User account names
• title - Node titles

Never change these to type: text - it breaks partial name matching (e.g., "mich" won't find "Michael").

Config Export Side Effects

When modifying Search API server config (e.g., for local development), Drupal may update index configs to remove server dependencies:

# Unintended change in search_api.index.{index_name}.yml
dependencies:
  config:
-    - search_api.server.{server_name}  # Removed
server: null  # Changed from {server_name}

Prevention:

1. Don't modify search_api.server.{server_name}.yml for local development
2. Use DDEV's built-in Solr service without changing server config
3. If server config changes are needed, use config splits for environment-specific settings
4. Always review git diff config/default/ before committing

Recovery:

# Revert index files if server was set to null
git checkout HEAD -- config/default/search_api.index.*.yml

Reindexing After Changes

When Reindexing is Required

Always reindex after:

• Changing field type (integer -> boolean)
• Changing boost factors
• Adding/removing processors
• Modifying processor configuration

Reindex Commands

# Full reindex workflow
ddev drush cr
ddev drush search-api:clear {index_name}
ddev drush search-api:index {index_name}

# Check index status
ddev drush search-api:status {index_name}

Partial Reindexing

# Index only remaining items (incremental)
ddev drush search-api:index {index_name}

# Clear and reindex specific items (not available in core)
# Use full reindex instead

Testing Boosts

Search API Drush Commands

# Test search via drush
ddev drush search-api:search {index_name} "search query"

# Check indexed values
ddev drush search-api:status {index_name}

Web Endpoint Testing

# Test via web search endpoint
curl -H "Accept: application/vnd.api+json" \
  "https://example.com/search-api/web-search?query=test"

# Parse results with Python
curl -s "https://example.com/search-api/web-search?query=test" | \
  python3 -c "import sys, json; data = json.load(sys.stdin); \
  [print(f'{i+1}. {item[\"attributes\"][\"title\"]}') \
  for i, item in enumerate(data['data'][:10])]"

Validation Queries

Check Bookmark Counts

SELECT u.uid, u.name, fcf.count AS bookmark_count
FROM users_field_data u
LEFT JOIN flag_counts fcf ON fcf.entity_type = 'user'
  AND fcf.entity_id = u.uid
  AND fcf.flag_id = 'bookmark'
ORDER BY fcf.count DESC
LIMIT 10;

Check Favorite Counts

SELECT n.nid, n.title, fcl.count AS favorite_count
FROM node_field_data n
INNER JOIN flag_counts fcl ON fcl.entity_type = 'node'
  AND fcl.entity_id = n.nid
  AND fcl.flag_id = 'favorite'
WHERE n.type = 'article'
ORDER BY fcl.count DESC
LIMIT 10;

Performance Considerations

Index-Time vs Query-Time Boosting

Index-Time Boosting (Recommended)

• Better performance (computed once during indexing)
• Scales with high query volume
• Requires reindex when boost logic changes
• Static boosts (can't adjust per-query)

Query-Time Boosting

• Dynamic (can adjust per-query)
• No reindex needed for changes
• Performance impact on every query
• Complex to implement with Search API

Recommended Pattern: Index-Time Boosting

• All boosts applied during preprocess_index stage
• Scales better for high-traffic search
• Acceptable reindex requirement

Boost Stacking

How Boosts Combine

1. Base relevance score (from Solr)
2. Featured content boost: score x 2.0
3. Bookmark count boost: score + (bookmarks x 0.01)
4. Favorite count boost: score + (favorites x 0.1)

Example Calculation

Node: "Getting Started with Drupal" (featured, 138 bookmarks)
- Base score: 1.0
- Featured: 1.0 x 2.0 = 2.0
- 138 bookmarks: 2.0 + (138 x 0.01) = 3.38
- Final boost: 3.38

Node: "Advanced Drupal Techniques"
- Base score: 1.0
- 12 favorites: 1.0 + (12 x 0.1) = 2.2
- Final boost: 2.2

Common Issues

Issue: Boost Not Applied

Symptoms

• Results not ranking as expected
• No difference after reindex

Checklist

1. Processor enabled in index config
2. Field type correct (boolean/integer)
3. Full reindex completed
4. No conflicting processors
5. Cache cleared after config changes

Debug Commands

# Check processor configuration
ddev drush config:get search_api.index.{index_name} processor_settings

# Check field configuration
ddev drush config:get search_api.index.{index_name} field_settings

# Verify index status
ddev drush search-api:status {index_name}

Issue: Boost Too Strong/Weak

Symptoms

• Irrelevant results ranking too high
• Important results not ranking high enough

Solution: Adjust Boost Factors

Too strong (dominating relevance):

• Reduce multiplicative boosts (2.0 -> 1.5)
• Reduce additive boost factors (0.1 -> 0.05)

Too weak (no visible effect):

• Increase boost factors gradually
• Test with extreme values to confirm working, then dial back

Testing Pattern

1. Start with conservative boost (1.5x or 0.01)
2. Test with known queries
3. Increase gradually until effect visible
4. Reduce to optimal level

Module Dependencies

Required Modules

• search_api - Core Search API module
• search_api_solr - Solr backend (if using Solr)

Optional Modules

• flag_search_api - Index flag counts (bookmark, favorite)
• search_api_boolean_field_boost - Reference for boost patterns

References

Search API Processor Development

• Search API Processors
• Drupal API: ProcessorPluginBase

Boosting Strategies

• Search API Boost Module
• Solr Boosting

Flag Search API

• Flag Search API Module
• Provides flag count indexing for engagement metrics