<?php
// $Id: recommender.module,v 1.18 2010-06-03 17:52:22 danithaca Exp $

/**
 * @file
 * Providing generic recommender system algorithms.
 */

require_once('Recommender.php');

/**
 * classical collaborative filtering algorithm based on correlation coefficient.
 * could be used in the classical user-user or item-item algorithm
 * see the README file for more details
 *
 * @param $app_name the application name that uses this function.
 * @param $table_name the input table name, don't use {tablename} here because it'll get expanded later
 * @param $field_mouse the input table field for mouse
 * @param $field_cheese the input table field for cheese
 * @param $field_weight the input table field weight
 * @param $options an array of options
 *          'performance': whether to do calculation in memory or in database. 'auto' is to decide automatically.
 *          'missing': how to handle missing data -- 'none' do nothing; 'zero' fill in missing data with zero; 'adjusted' skip mice that don't share cheese in common.
 *          'sensitivity': if similarity is smaller enough to be less than a certain value (sensitivity), we just discard those
 * @return null {recommender_similarity} will be filled with similarity data
 */
function recommender_similarity_classical($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options = array()) {
  $recommender = new CorrelationRecommender($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options);
  $recommender->computeSimilarity();
}


/**
 * Classical weight-average algorithm to calculate prediction from the similarity matrix, based on average weight.
 * Limitation: we only do database operation for now. no in-memory operation available until future release.
 * Limitation: we only do average weight. regression-based weight maybe included in future release.
 *
 * @param $app_name the application name that uses this function.
 * @param $table_name the input table name
 * @param $field_mouse the input table field for mouse
 * @param $field_cheese the input table field for cheese
 * @param $field_weight the input table field weight
 * @param $options an array of options
 *          'missing': how to handle missing data -- 'none' do nothing; 'zero' fill in missing data with zero; 'adjusted' skip mice that don't share cheese in common.
 *          'duplicate': how to handle predictions that already exists in mouse-cheese evaluation. 'preserve' or 'eliminate'
 *          'sensitivity': if similarity is smaller enough to be less than a certain value, we treat it as zero
 * @return null {recommender_prediction} will be filled with prediction data
 */
function recommender_prediction_classical($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options = array()) {
  $recommender = new CorrelationRecommender($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options);
  $recommender->computePrediction();
}


function recommender_user2user($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options = array()) {
  $options['sim_pred'] = TRUE;
  $recommender = new User2UserRecommender($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options);
  $recommender->computeSimilarity();
  $recommender->computePrediction();
}


function recommender_item2item($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options = array()) {
  $options['sim_pred'] = TRUE;
  $recommender = new Item2ItemRecommender($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options);
  $recommender->computeSimilarity();
  $recommender->computePrediction();
}

/**
 * Co-ocurrences algorithm that compute similarity among mice based on how many cheese they share.
 *
 * @param $app_name the application name that uses this function.
 * @param $table_name the input table name
 * @param $field_mouse the input table field for mouse
 * @param $field_cheese the input table field for cheese
 * @param $options an array of options
 *          'incremental': whether to rebuild the whole similarity matrix, or incrementally update those got changed. 'rebuild' or 'update'
 * @return null {recommender_similarity} will be filled with similarity data
 */
function recommender_similarity_coocurrences($app_name, $table_name, $field_mouse, $field_cheese, $options = array()) {
  $recommender = new CooccurrenceRecommender($app_name, $table_name, $field_mouse, $field_cheese, NULL, $options);
  $recommender->computeSimilarity();
}


/**
 * This is the implementation of slope-one algorithm, which is supposed to be faster than other algrithms.
 * From the original research paper, the author argues slope-one support incremental update. But this is quite hard to implement.
 * The incremental update doesn't work for now.
 * Missing data is just treated as missing. For slope-one, we don't automatically expand the matrix to have zeros.
 * The responsibility of handling missing data is on the caller functions.
 *
 * @param $app_name the application name that uses this function.
 * @param $table_name the input table name
 * @param $field_mouse the input table field for mouse
 * @param $field_cheese the input table field for cheese
 * @param $field_weight the input table field weight
 * @param $options an array of options
 *          'extention': whether to use 'basic', 'weighted', or 'bipolar' extensions of the algorithm.
 *                       Please refer to the original research paper. Usually it could just be 'weighted'.
 *          'duplicate': how to handle predictions that already exists in mouse-cheese evaluation. 'preserve' or 'eliminate'
 *          'incremental': whether to 'rebuild' or to 'update'. CAUTION: 'update' doesn't work now.
 * @return null {recommender_prediction} will be filled with prediction data
 */
function recommender_prediction_slopeone($app_name, $table_name, $field_mouse, $field_cheese, $field_weight, $options = array()) {

}


/////////////////////// Helper functions /////////////////////


/**
 * Get the application id from the application name.
 * @param $app_name
 * @return the id of the application.
 */
function recommender_get_app_id($app_name) {
  return Recommender::convertAppId($app_name);
}


/**
 * Remove the application. Usually used in calling module's hook_uninstall()
 * @param $app_name the application name to be removed.
 * @return null
 */
function recommender_purge_app($app_name) {
  Recommender::purgeApp($app_name);
}

/**
 * Return a list of items that are top similar with $id
 * @param $app_name The $app_name for the depending modules
 * @param $id usually the $node_id of the target item.
 * @param $top_n how many similar items to return
 * @param $test_func optional function to test whether an item satisfy some conditions
 * @return an array of the most similar items to $id.
 */
function recommender_top_similarity($app_name, $id, $top_n, $test_func=NULL) {
  // create a null recommender to access the topSimilarity() function.
  $recommender = new Recommender($app_name, NULL, NULL, NULL, NULL);
  return $recommender->topSimilarity($id, $top_n, $test_func);
}

/**
 * Return a list of items that are top prediction for $id
 * @param $app_name The $app_name for the depending modules
 * @param $id usually the $node_id of the target item.
 * @param $top_n how many predictions to return
 * @param $test_func optional function to test whether an item satisfy some conditions
 * @return an array of the most similar items to $id.
 */
function recommender_top_prediction($app_name, $id, $top_n, $test_func=NULL) {
  // create a null recommender to access the topPrediction() function.
  $recommender = new Recommender($app_name, NULL, NULL, NULL, NULL);
  return $recommender->topPrediction($id, $top_n, $test_func);
}


//////////////////////////// DRUPAL RELATED FUNCTIONS //////////////////////////

// implementation of hook_perm()
function recommender_perm() {
  $perm = array(
    "administer recommender",
  );
  return $perm;
}

// implementation of hook_menu()
function recommender_menu() {
  $items = array();

  $items['admin/settings/recommender'] = array(
    'title' => 'Recommender API',
    'description' => 'Configuration and trigger recommender modules',
    'page callback' => 'drupal_get_form',
    'page arguments' => array('recommender_settings_form'),
    'access arguments' => array('administer recommender'),
  );

  $items['recommender/run'] = array(
    'title' => 'Running recommender',
    'page callback' => 'recommender_run',
    'access arguments' => array('administer recommender'),
    'type' => MENU_CALLBACK,
   );
  return $items;
}

function recommender_settings_form() {
  $form = array();

  $form['settings'] = array(
    '#type' => 'fieldset',
    '#collapsible' => FALSE,
    '#collapsed' => FALSE,
    '#title' => t('Settings'),
    '#description' => t('Change settings for Recommender API based modules.'),
  );

  $form['settings']['cron_freq'] = array(
    '#title' => t('Recommender running frequency in cron job.'),
    '#type' => 'select',
    '#default_value' => variable_get('recommender_cron_freq', 'never'),
    '#options' => array(
      'never' => 'Never',
      'immediately' => 'Immediately',
      'hourly' => t('Hourly'),
      'every6hr' => t('Every 6 hours'),
      'every12hr' => t('Every 12 hours'),
      'daily' => t('Daily'),
      'weekly' => t('Weekly')
    ),
    '#description' => t("Please specify the optional frequency to run recommender algorithms in cron. Note that this is a time consuming operation and might timeout or affect your other cron tasks. Not recommended for large site. Consider using the Drush script with system cron.")
  );

  $form['settings']['save'] = array(
    '#type' => 'submit',
    '#value' => t('Save'),
    '#name' => 'save'
  );

  $form['run'] = array(
    '#type' => 'fieldset',
    '#collapsible' => FALSE,
    '#collapsed' => FALSE,
    '#title' => t('Run recommender'),
    '#description' => t('Running recommender involves complex matrix computation and could probably take some time. Please be patient. You can also run recommender with Drush.'),
  );

  $options = drupal_map_assoc(module_implements('run_recommender'));
  $modules = module_list();

  if (empty($options)) {
    $form['run']['note'] = array(
      '#title' => 'Note',
      '#type' => 'item',
      '#description' => t('No recommender modules available.'),
    );
  }
  else {
    $form['run']['modules'] = array(
      '#title' => t('Choose modules'),
      '#default_value' => variable_get('recommender_modules', array()),
      '#type' => 'checkboxes',
      '#description' => t('Please select which modules to run the recommender'),
      '#options' => $options,
    );
  }

  $form['run']['run'] = array(
    '#type' => 'submit',
    '#value' => t('Run recommender now'),
    '#name' => 'run',
    '#disabled' => $options==NULL ? TRUE : FALSE,
  );

  return $form;
}


function recommender_settings_form_submit($form, &$form_state) {
  $cron_freq = $form_state['values']['cron_freq'];
  $modules = $form_state['values']['modules'];

  variable_set('recommender_cron_freq', $cron_freq);
  variable_set('recommender_modules', $modules);

  if ($form_state['clicked_button']['#name'] == 'save') {
    drupal_set_message(t("The configuration options have been saved."));
  }
  else {
    // trigger recommender_run()
    $modules = array_values(array_diff($modules, array(0)));
    if (empty($modules)) {
      drupal_set_message(t("No module selected to run recommender"));
    }
    else {
      recommender_run($modules);
    }
  }
}


function recommender_cron() {
  $last_cron = variable_get('recommender_last_cron', 0);
  $cron_freq = variable_get('recommender_cron_freq', 'never');

  $current = time();
  switch ($cron_freq) {
    case 'immediately':
      $run = TRUE;
      break;
    case 'hourly':
      $run = (($current-$last_cron) >= 3600) ? TRUE : FALSE;
      break;
    case 'every6hr':
      $run = (($current-$last_cron) >= 21600) ? TRUE : FALSE;
      break;
    case 'every12hr':
      $run = (($current-$last_cron) >= 43200) ? TRUE : FALSE;
      break;
    case 'daily':
      $run = (($current-$last_cron) >= 86400) ? TRUE : FALSE;
      break;
    case 'weekly':
      $run = (($current-$last_cron) >= 604800) ? TRUE : FALSE;
      break;
    case 'never':
    default:
      $run = FALSE;
  }

  $msg = $run ? "Will run." : "Not running.";
  watchdog('recommender', "Recommender cron at frequency ${cron_freq}. $msg");

  if ($run == TRUE) {
    recommender_run();
    variable_set('recommender_last_cron', $current);
  }
}


function recommender_run($selected = NULL) {
  watchdog('recommender', "Invoking run_recommender. Might start a time-consuming process.");

  // hook_run_recommender() doesn't take any args. args setting is the responsibility of caller modules.
  //module_invoke_all('run_recommender');
  $operations = array();
  foreach (module_implements('run_recommender') as $module) {
    if ($selected===NULL || in_array($module, $selected)) {
      $operations[] = array("{$module}_run_recommender", array());
    }
  }

  $batch = array(
    'operations' => $operations,
    //'finished' => 'recommender_run_finished', // post-operations, not needed.
    'title' => t("Running recommender"),
    'init_message' => t("The recommender engine is running. For medium to large site (users>100, nodes>1000), this could be very slow, and might cause PHP memory overload or execution timeout. Please be patient."),
    'progress_message' => t("The recommender engine is running. Please be patient. Remaining @remaining out of @total"),
    'error_message' => t("Running recommender encounter an internal error. If it is due to PHP out of memory, or timeout, please use Drush script instead. Otherwise, please file an issue to the http://drupal.org/project/recommender"),
  );

  batch_set($batch);
  // FIXME: the batch mode needs to be more polished. e.g., create a summary page.
  // we are settled for now.
  batch_process('/admin/settings/recommender');
}


/**
 * Implementation of hook_views_api().
 */
function recommender_views_api() {
  return array('api' => 2.0);
}


// $simpred: 0 - similarity table, 1 - prediction table
// $un1: 0 - first field is user, 1 - first field is node
// $un2: 0 - second field is user, 1 - second field is node
function recommender_debug_table($app_name, $simpred, $un1, $un2) {
  $app_id = recommender_get_app_id($app_name);

  // load data
  if ($simpred == 0) {
    $sql = "SELECT mouse1_id un1, mouse2_id un2, similarity score FROM {recommender_similarity} WHERE app_id=%s ORDER BY mouse1_id, mouse2_id";
  }
  else if ($simpred == 1) {
    $sql = "SELECT mouse_id un1, cheese_id un2, prediction score FROM {recommender_prediction} WHERE app_id=%s ORDER BY mouse_id, cheese_id";
  }
  $result = db_query($sql, $app_id);

  // load users
  $users = array();
  $u_result = db_query("SELECT uid, name FROM {users}");
  while ($u = db_fetch_array($u_result)) {
    $users[$u['uid']] = $u['name'];
  }

  // load nodes
  $nodes = array();
  $n_result = db_query("SELECT nid, title FROM {node}");
  while ($n = db_fetch_array($n_result)) {
    $nodes[$n['nid']] = $n['title'];
  }

  $un1_content = $un1 ? $nodes : $users;
  $un2_content = $un2 ? $nodes : $users;

  $un1_header = $un1 ? 'node' : 'user';
  $un2_header = $un2 ? 'node' : 'user';
  $header = array($un1_header, $un2_header, 'score');
  $rows = array();

  while($record = db_fetch_array($result)) {
    $r = array();
    $r[] = l($un1_content[$record['un1']], "{$un1_header}/{$record['un1']}");
    $r[] = l($un2_content[$record['un2']], "{$un2_header}/{$record['un2']}");
    $r[] = $record['score'];
    $rows[] = $r;
  }

  return theme('table', $header, $rows);
}