Convert translations

[asmecher]

A proposal for a rearrangement of translation data; see pkp/pkp-lib#857 (comment)

Just in case I lose it -- the source of the throwaway conversion tool is...

<?php

// Get an associative mapping of English term => URI.
$englishRoleDocument = simplexml_load_file('https://raw.githubusercontent.com/JATS4R/jats-schematrons/master/schematrons/1.0/credit-roles.xml');
$englishRoleToUri = [];
foreach ($englishRoleDocument->item as $item) {
  $englishTerm = (string) $item['term'];
  $uri = (string) $item['uri'];
  $englishTerm = match($englishTerm) {
    'Writing – original draft' => 'Writing – Original Draft Preparation',
    'Writing – review & editing' => 'Writing – Review & Editing',
    'Data curation' => 'Data Curation',
    'Funding acquisition' => 'Funding Acquisition',
    'Project administration' => 'Project Administration',
    default => $englishTerm,
  };

  $englishRoleToUri[$englishTerm] = $uri;
}

$uriToId = [];

$englishData = [];


foreach (glob('translations/credit_translation_*.json') as $filename) {
  if (!preg_match('~translations/credit_translation_([a-z_]+)\.json~', $filename, $matches)) throw new Exception('Unexpected filename!');
  $localeCode = $matches[1];
  // Load a translation
  $data = json_decode(file_get_contents($filename),true);
  $newData = ['metadata' => $data['metadata'], 'translations' => []];
  foreach ($englishRoleToUri as $englishRole => $uri) {
    // Store old-style data in new translation data
    $newData['translations'][$uri] = [
      'name' => $data['translations'][$englishRole]['translation']['name'],
      'description' => $data['translations'][$englishRole]['translation']['description'],
    ];
  
    // Handle the IDs, which should be in a common descriptor file
    $id = $data['translations'][$englishRole]['id'];
    if (!isset($uriToId[$uri])) $uriToId[$uri] = $id;
    elseif ($uriToId[$uri] !== $id) throw new Exception('ID mismatch in ' . $filename . ' for role ' . $uri . ': ' . $id . ' vs. ' . $uriToId[$uri]);
  }
  file_put_contents("translations/$localeCode.json", json_encode($newData, JSON_PRETTY_PRINT | JSON_UNESCAPED_UNICODE));
}

// Write the mapping of CRediT roles to IDs
file_put_contents('credit_roles.json', json_encode($uriToId, JSON_PRETTY_PRINT | JSON_UNESCAPED_UNICODE));

// Finally, output the English versions (such as we have them)
$newData = ['metadata' => [], 'translations' => []];
foreach ($englishRoleToUri as $englishRole => $uri) {
  // Store old-style data in new translation data
  $newData['translations'][$uri] = [
    'name' => $englishRole,
    'description' => '',
  ];
}
file_put_contents('translations/en.json', json_encode($newData, JSON_PRETTY_PRINT | JSON_UNESCAPED_UNICODE));```

[marton-balazs-kovacs]

I want to loop in @mfenner who extensively worked on article metadata. Martin, as I mentioned in our call today we had a json-schema for storing translations that you can find in the main branch of the repo. Here, @asmecher suggests a new format that would be more streamlined and easier to translate to JATSXML for validation with the JATS schematron. The goal is to incorporate the CRediT translations in OJS so people can upload their metadata using the local language.

What do you think of the suggested format? Since we are modifying the previous version of the translation json-schema it is a good opportunity to discuss the structure.