ISSUE-3: OCR specific Processor and new features/processing option

config/schema/strawberry_runners.schema.yml

@@ -31,6 +31,44 @@ strawberryfield_runners.strawberry_runners_postprocessor.*:
 strawberryfield_runners.strawberry_runners_postprocessor.binary:
   type: config_object
   label: 'Strawberry Runners Post Processor Config Entity Binary specific config'
+  mapping:
+    source_type:
+      type: string
+      label: 'The type of Source Data this Processor works on'
+    ado_type:
+      type: string
+      label: 'DO type(s) to limit this Processor to'
+    jsonkey:
+      type: sequence
+      label: 'The JSON key(s) containing the desired Source File(s)'
+      sequence:
+        - type: string
+    mime_type:
+      type: string
+      label: 'Mimetypes(s) to limit this Processor to'
+    path:
+      type: string
+      label: 'The path for he binary to execute'
+    arguments:
+      type: string
+      label: 'Any additional argument your executable binary requires'
+    output_type:
+      type: string
+      label: 'The expected and desired output of this processor'
+    output_destination:
+      type: sequence
+      label: 'Where and how the output will be used'
+      sequence:
+        - type: string
+    timeout:
+      type: integer
+      label: 'Timeout in seconds for this process'
+    weight:
+      type: integer
+      label: 'Order or execution in the global chain'
+strawberryfield_runners.strawberry_runners_postprocessor.ocr:
+  type: config_object
+  label: 'Strawberry Runners Post Processor Config Entity OCR specific config'
   mapping:
     source_type:
       type: string
@@ -49,6 +87,15 @@ strawberryfield_runners.strawberry_runners_postprocessor.binary:
     arguments:
       type: string
       label: 'Any additional argument your executable binary requires'
+    tesseract_arguments:
+      type: string
+      label: 'Any additional argument your executable binary requires'
+    path:
+      type: string
+      label: 'The path for he binary to execute'
+    tesseract_path:
+      type: string
+      label: 'The path for he binary to execute'
     output_type:
       type: string
       label: 'The expected and desired output of this processor'

src/Annotation/StrawberryRunnersPostProcessor.php

@@ -55,6 +55,13 @@ class StrawberryRunnersPostProcessor extends Plugin {
    */
   public $input_property;
 
+  /**
+   * The Object property that contains the additional data needed by the Processor ::run method
+   *
+   * @var string $input_arguments;
+   *
+   */
+  public $input_arguments;
 
   /**
    * Processing stage: can be Entity PreSave or Index time search_api
@@ -64,4 +71,4 @@ class StrawberryRunnersPostProcessor extends Plugin {
    */
   public $when = StrawberryRunnersPostProcessor::PRESAVE;
 
-}
+}

src/EventSubscriber/StrawberryRunnersEventPreSavePostProcessingSubscriber.php → src/EventSubscriber/StrawberryRunnersEventSavePostProcessingSubscriber.php

@@ -5,7 +5,6 @@
 use Drupal\Core\Entity\EntityTypeManagerInterface;
 use Drupal\Core\Session\AccountInterface;
 use Drupal\strawberryfield\Event\StrawberryfieldCrudEvent;
-use Drupal\strawberryfield\EventSubscriber\StrawberryfieldEventPresaveSubscriber;
 use Drupal\Core\StringTranslation\StringTranslationTrait;
 use Drupal\Core\StringTranslation\TranslationInterface;
 use Drupal\Core\Messenger\MessengerInterface;
@@ -14,14 +13,14 @@
 use Drupal\Component\Utility\Unicode;
 use Drupal\file\FileInterface;
 use Drupal\Core\StreamWrapper\StreamWrapperManagerInterface;
-use Drupal\Core\StreamWrapper\StreamWrapperInterface;
 use Drupal\Core\File\FileSystemInterface;
 use Drupal\strawberry_runners\Plugin\StrawberryRunnersPostProcessorPluginManager;
+use Drupal\strawberryfield\EventSubscriber\StrawberryfieldEventSaveSubscriber;
 
 /**
  * Event subscriber for SBF bearing entity json process event.
  */
-class StrawberryRunnersEventPreSavePostProcessingSubscriber extends StrawberryfieldEventPresaveSubscriber {
+class StrawberryRunnersEventSavePostProcessingSubscriber extends StrawberryfieldEventSaveSubscriber {
 
 
   use StringTranslationTrait;
@@ -150,7 +149,7 @@ public function __construct(
    * @throws \Drupal\Component\Plugin\Exception\PluginException
    * @throws \Drupal\Component\Plugin\Exception\PluginNotFoundException
    */
-  public function onEntityPresave(StrawberryfieldCrudEvent $event) {
+  public function onEntitySave(StrawberryfieldCrudEvent $event) {
 
     /* @var $plugin_config_entities \Drupal\strawberry_runners\Entity\strawberryRunnerPostprocessorEntity[] */
     $plugin_config_entities = $this->entityTypeManager->getListBuilder('strawberry_runners_postprocessor')->load();
@@ -232,9 +231,9 @@ public function onEntityPresave(StrawberryfieldCrudEvent $event) {
           /** @var $itemfield \Drupal\strawberryfield\Plugin\Field\FieldType\StrawberryFieldItem */
           $flatvalues = (array) $itemfield->provideFlatten();
           // Run first on entity:files
-          $sbf_type = NULL;
+          $sbf_type = [];
           if (isset($flatvalues['type'])) {
-            $sbf_type = $flatvalues['type'];
+            $sbf_type = (array) $flatvalues['type'];
           }
           foreach ($askeymap as $jsonkey => $activePlugins) {
             if (isset($flatvalues[$jsonkey])) {
@@ -243,12 +242,16 @@ public function onEntityPresave(StrawberryfieldCrudEvent $event) {
 
                   foreach($activePlugins as $activePluginId => $config) {
                     $valid_mimes = [];
-                    if (empty($config['ado_type']) || in_array($config['ado_type'] , $sbf_type)) {
+                    //@TODO also split $config['ado_type'] so we can check
+                    $valid_ado_type = [];
+                    $valid_ado_type = explode(',', $config['ado_type']);
+                     if (empty($config['ado_type']) || count(array_intersect($valid_ado_type , $sbf_type)) > 0) {
                       $valid_mimes = explode(',', $config['mime_type']);
                       if (empty($valid_mimes) || (isset($asstructure["dr:mimetype"]) && in_array($asstructure["dr:mimetype"], $valid_mimes))) {
                         $data = new \stdClass();
                         $data->fid = $asstructure['dr:fid'];
                         $data->nid = $entity->id();
+                        $data->nuuid = $entity->uuid();
                         // We are passing also the full file metadata.
                         // This gives us an advantage so we can reuse
                         // Sequence IDs, PDF pages, etc and act on them
@@ -260,6 +263,16 @@ public function onEntityPresave(StrawberryfieldCrudEvent $event) {
                         // $activePluginId? That would allow us to skip reprocessing
                         // Easier?
                         $data->metadata = $asstructure;
+
+                        // @TODO how to force?
+                        // Can be a state key, valuekey, or a JSON passed property.
+                        // Issue with JSON passed property is that we can no longer
+                        // Here modify it (Entity is saved)
+                        // So we should really better have a non Metadata method for this
+                        // Or/ we can have a preSave Subscriber that reads the prop,
+                        // sets the state and then removes if before saving
+
+                        $data->force = FALSE;
                         $data->plugin_config_entity_id = $activePluginId;
                         // See https://github.com/esmero/strawberry_runners/issues/10
                         // Since the destination Queue can be a modal thing

src/Plugin/QueueWorker/IndexPostProcessorQueueWorker.php

@@ -159,12 +159,13 @@ public function processItem($data) {
 
     $processor_instance = $this->getProcessorPlugin($data->plugin_config_entity_id);
 
-    if (!isset($data->fid) || $data->fid == NULL || !isset($data->nid) || $data->nid == NULL) {
+    if (!isset($data->fid) || $data->fid == NULL || !isset($data->nid) || $data->nid == NULL || !is_array($data->metadata)) {
       return;
     }
     $file = $this->entityTypeManager->getStorage('file')->load($data->fid);
 
-    if ($file === NULL) {
+    if ($file === NULL || !isset($data->metadata['checksum'])) {
+      error_log('Sorry the file does not exist or has no checksum yet. We really need the checksum');
       return;
     }
     //@TODO should we wrap this around a try catch?
@@ -188,37 +189,54 @@ public function processItem($data) {
 
       // Skip file if element is found in key_value collection.
       $processed_data = $this->keyValue->get($keyvalue_collection)->get($key);
-
-      if (empty($processed_data)) {
+      error_log('Is this already in our temp keyValue?');
+      error_log(empty($processed_data));
+      //@TODO allow a force in case of corrupted key value? Partial output
+      // Extragenous weird data?
+      if (true || empty($processed_data) ||
+        $data->force == TRUE ||
+        (!isset($processed_data->checksum) ||
+          empty($processed_data->checksum) ||
+          $processed_data->checksum != $data->metadata['checksum'])) {
         // Extract file and save it in key_value collection.
         $io = new \stdClass();
         $input =  new \stdClass();
         $input->filepath = $filelocation;
-
+        $input->page_number = 1;
+        // The Node UUID
+        $input->nuuid = $data->nuuid;
+        // All the rest of the associated Metadata in an as:structure
+        $input->metadata = $data->metadata;
         $io->input = $input;
         $io->output = NULL;
+        //@TODO implement the TEST and BENCHMARK logic here
+        // RUN should return exit codes so we can know if something failed
+        // And totally discard indexing.
         $extracted_data = $processor_instance->run($io, StrawberryRunnersPostProcessorPluginInterface::PROCESS);
         error_log ('processing just run');
-        error_log($io->ouput);
         error_log('writing to keyvalue');
         error_log($key);
-        $this->keyValue->get($keyvalue_collection)->set($key, $io->output);
-      }
-
-      // Get which indexes have our StrawberryfieldFlavorDatasource enabled!
-      $indexes = StrawberryfieldFlavorDatasource::getValidIndexes();
-
-      $item_ids = [];
-      if (is_a($entity, TranslatableInterface::class)) {
-        $translations = $entity->getTranslationLanguages();
-        foreach ($translations as $translation_id => $translation) {
-          $item_ids[] = $entity->id() . ':'.'1' .':'.$translation_id.':'.$file->uuid().':'.$data->plugin_config_entity_id;
+        $toindex = new \stdClass();
+        $toindex->fulltext = $io->output;
+        $toindex->checksum = $data->metadata['checksum'];
+        error_log(var_export($toindex,true));
+        $this->keyValue->get($keyvalue_collection)->set($key, $toindex);
+
+        // Get which indexes have our StrawberryfieldFlavorDatasource enabled!
+        $indexes = StrawberryfieldFlavorDatasource::getValidIndexes();
+
+        $item_ids = [];
+        if (is_a($entity, TranslatableInterface::class)) {
+          $translations = $entity->getTranslationLanguages();
+          foreach ($translations as $translation_id => $translation) {
+            $item_ids[] = $entity->id() . ':'.'1' .':'.$translation_id.':'.$file->uuid().':'.$data->plugin_config_entity_id;
+          }
+        }
+        error_log(var_export($item_ids,true));
+        $datasource_id = 'strawberryfield_flavor_datasource';
+        foreach ($indexes as $index) {
+          $index->trackItemsInserted($datasource_id, $item_ids);
         }
-      }
-      error_log(var_export($item_ids,true));
-      $datasource_id = 'strawberryfield_flavor_datasource';
-      foreach ($indexes as $index) {
-        $index->trackItemsUpdated($datasource_id, $item_ids);
       }
     }
     catch (\Exception $exception) {
@@ -252,7 +270,7 @@ private function ensureFileAvailability(FileInterface $file) {
     // Check first if the file is already around in temp?
     // @TODO can be sure its the same one? Ideas?
     if (is_readable(
-     $this->fileSystem->realpath(
+      $this->fileSystem->realpath(
         'temporary://sbr_' . $cache_key . '_' . basename($uri)
       )
     )) {
@@ -306,4 +324,4 @@ public function getRealpath($uri) {
     }
   }
 
-}
+}