Tripal 4 Core Dev Task

Issue #1716

Tripal Version: 4

Description

This PR fixes issues when you run two publish jobs in the same tripal job. Specifically, when you publish two content types in the same Tripal job, the values from the first publish are leaked into the second publish.

I believe this is happening due to some unwritten Drupal rule in how the TripalPublish service is written. I think it's related to the fact that this service uses an init() method to setup everything rather then typical dependency injection with a constructor... but I originally got so caught up in writing a test to demonstrate the problem that I couldn't get my brain to move on to finding the fix.

Luckily, after posting this as a draft PR and asking for help, @dsenalik realized that we could just reset the leaked values in the init() 😅 which is a really clean solution. It may not be ideal and we may run into other issues with this service in the future... but based on my understanding of what is happening, I do think it is a good, reliable fix :-) And I don't know what the ideal fix is so I say we go with this! 🤪

What exactly does this PR do?

1. Updates the TripalPublish::init() and other variables to ensure they are reset on subsequent runs of this service.
2. Creates two simple tests that actually run the publish: these were failing before fix no. 1 and are now working. They are not ideal and do not have as many assertions as I would like but they will help us catch situations where the publish job straight up fails so they are worth keeping.
3. Creates a new testing helper method to help setup content types with fields based on the YAML. This replaces the current process where you have to first setup the content type and then attach the fields in the test. Plus it ensures the content type you are testing has all the default fields.

Testing?

Setup Testing environment

1. Create a fresh site/docker
2. Use Drush php:cli or devel php code to add records to Chado. The following is how I do this for Organism + Contacts. Alternatively you could load a GFF3 file.

$connection = \Drupal::service('tripal_chado.database');
for ($i=1; $i <= 10; $i++) {
  $species = 'databasica' . uniqid();
  $infra_name = 'postgresquelus' . $i; 
  $connection->insert('1:organism')
    ->fields([
      'genus' => 'Tripalus', 
      'species' => $species, 
      'infraspecific_name' => $infra_name, 
      'type_id' => 500
    ])->execute();
}
for ($i=1; $i <= 5; $i++) {
  $name = 'Tripal' . uniqid() . ' collaborator #' . $i;
  $description = md5(mt_rand()).' '.md5(mt_rand());
  $connection->insert('1:contact')
    ->fields([
      'name' => $name, 
      'description' => $description, 
      'type_id' => 500
    ])->execute();
}
for ($i=1; $i <= 15; $i++) {
  $name = 'Project TestingPublish' . uniqid() . ' interation #' . $i;
  $description = md5(mt_rand()).' '.md5(mt_rand());
  $connection->insert('1:project')
    ->fields([
      'name' => $name, 
      'description' => $description, 
    ])->execute();
}

Single publish job per call to job launcher works

1. Run tripal jobs to see that the queue is empty.
2. Go to Admin > Tripal > Content > Publish and choose a random content type with unpublished data. If you setup as I did above then you can use "contact".
3. Immediately run the tripal job to complete the publishing and then confirm.
4. Repeat above steps exactly and confirm all general content types publish perfectly.

When multiple publish jobs are in the queue, the second one fails.

From a fresh install of Tripal or a fresh docker container

1. Run tripal jobs to see that the queue is empty.
2. Go to Admin > Tripal > Content > Publish and choose a random content type with unpublished data. If you setup as I did above, then you can use "organism".
3. WITHOUT RUNNING THE JOB LAUNCHER, repeat step 2 for another content type. If you setup as I did above then you can use "project".
4. Now run the tripal jobs and confirm you see a big error and the second content type, regardless of what you chose, failed because it tried to publish data from fields for the first content type.