From 138eab201673bdeed163dcbe2a1db7476c3f5fb6 Mon Sep 17 00:00:00 2001
From: Alexander O'Neill <aloneill@gmail.com>
Date: Wed, 3 May 2023 16:42:03 -0300
Subject: [PATCH] Issue #941: Only add <br/> tags to plain text extracted text
 fields. (#942)

* Issue #941: Only add <br/> tags to plain text extracted text fields.

* Fix PHPCS errors.

* Don't add <br /> tags to edited OCR text field if it looks like hOCR.

* Respond to PHPCS errors.
---
 .../islandora_text_extraction.module                        | 4 ++++
 .../src/Controller/MediaSourceController.php                | 6 +++++-
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/modules/islandora_text_extraction/islandora_text_extraction.module b/modules/islandora_text_extraction/islandora_text_extraction.module
index 5d6f6437..ca330dd4 100644
--- a/modules/islandora_text_extraction/islandora_text_extraction.module
+++ b/modules/islandora_text_extraction/islandora_text_extraction.module
@@ -40,6 +40,10 @@ function islandora_text_extraction_media_presave(MediaInterface $media) {
       $file = File::load($file_id);
       if ($file) {
         $data = file_get_contents($file->getFileUri());
+        // Check if it's already markup like hOCR.
+        if (substr($data, 0, 5) == '<?xml') {
+          return;
+        }
         $data = nl2br($data);
         $media->set('field_edited_text', $data);
         $media->field_edited_text->format = 'basic_html';
diff --git a/modules/islandora_text_extraction/src/Controller/MediaSourceController.php b/modules/islandora_text_extraction/src/Controller/MediaSourceController.php
index 14c36ebd..f15e42d5 100644
--- a/modules/islandora_text_extraction/src/Controller/MediaSourceController.php
+++ b/modules/islandora_text_extraction/src/Controller/MediaSourceController.php
@@ -108,7 +108,11 @@ class MediaSourceController extends ControllerBase {
         $this->getLogger('islandora')->warning("Field $destination_field is not defined in  Media Type {$media->bundle()}");
       }
       if ($media->hasField($destination_text_field)) {
-        $media->{$destination_text_field}->setValue(nl2br($contents));
+        // @todo The request actually has a malformed parameter string, ?text_format=plain_text?connection_close=true.
+        if (substr($request->query->get('text_format'), 0, 10) == 'plain_text') {
+          $contents = nl2br($contents);
+        }
+        $media->{$destination_text_field}->setValue($contents);
       }
       else {
         $this->getLogger('islandora')->warning("Field $destination_text_field is not defined in Media Type {$media->bundle()}");