Carga la secuencia de comandos

Un algoritmo basado en una secuencia de comandos se define mediante una secuencia de comandos que usa una sintaxis básica de Python. Sube la secuencia de comandos a Display & Video 360 a través de un archivo de texto.

Si usas un algoritmo basado en reglas, ve a la página Subir reglas.

Escribe un guion

Los algoritmos de ofertas personalizadas basados en secuencias de comandos emplean secuencias de comandos proporcionadas por el usuario para evaluar el valor de una impresión. Consulta el Centro de ayuda de Display & Video 360 para ver secuencias de comandos de ejemplo y una lista de campos.

Puedes ejecutar tu secuencia de comandos en la IU de Display & Video 360 para probar su rendimiento. Proporcionará una distribución de las puntuaciones para las impresiones muestreadas.

Guarda tu secuencia de comandos de forma local como un archivo TXT.

Genera un objeto de referencia de secuencia de comandos

Un objeto de referencia se usa para asociar un archivo subido con un script. Genera un objeto de referencia en tu algoritmo con una solicitud uploadScript.

Sigue estos pasos para generar un objeto de referencia de la secuencia de comandos:

Java

// Provide the ID of the advertiser that owns the algorithm.
long advertiserId = advertiser-id;

// Provide the ID of the algorithm.
long customBiddingAlgorithmId = algorithm-id;

// Generate the custom bidding script reference object.
CustomBiddingScriptRef response =
    service
        .customBiddingAlgorithms()
        .uploadScript(customBiddingAlgorithmId)
        .setAdvertiserId(advertiserId)
        .execute();

// Display the generated resource path.
System.out.printf(
    "The generated script reference object provided the following resource path: %s.",
    response.getResourceName());

Python

# Provide the ID of the advertiser that owns the algorithm.
advertiser_id = advertiser-id

# Provide the ID of the algorithm.
algorithm_id = algorithm-id

# Generate a script reference object under the algorithm.
custom_bidding_script_ref = (
    service.customBiddingAlgorithms()
    .uploadScript(
        customBiddingAlgorithmId=algorithm_id, advertiserId=advertiser_id
    )
    .execute()
)

# Print the resource path provided in the generated script reference object.
print(
    "The generated script reference object provided the following resource"
    " path: "
    f'{custom_bidding_script_ref["resourceName"]}.'
)

PHP

// Provide the ID of the advertiser that owns the algorithm.
$advertiserId = advertiser-id;

// Provide the ID of the algorithm.
$algorithmId = algorithm-id;

$uploadScriptOptParams = array(
    'advertiserId' => $advertiserId
);

try {
  // Retrieve a usable custom bidding script reference.
    $scriptRefResponse =
        $this->service->customBiddingAlgorithms->uploadScript(
            $customBiddingAlgorithmId,
            $uploadScriptOptParams
        );
} catch (\Exception $e) {
    $this->renderError($e);
    return;
}

// Print the retrieved resource path.
printf(
    '<p>The generated script reference object provided the following '
        . 'resource path: %s</p>',
    $scriptRefResponse->getResourceName()
);

Cómo subir un archivo de secuencia de comandos

Usa una solicitud de medios upload para subir tu archivo de secuencia de comandos a la ruta de acceso del recurso en el objeto de referencia. Usa una carga simple con un parámetro de consulta uploadType=media.

Advertencia: Las secuencias de comandos tienen un tamaño máximo de archivo de 7 MB.

Sigue estos pasos para subir un archivo de secuencia de comandos:

Java

// Provide the local path to the script file to upload.
String scriptPath = script-path;

// Provide the resource path to upload the script file to.
String resourcePath = resource-path;

// Create media object.
GoogleBytestreamMedia media = new GoogleBytestreamMedia();
media.setResourceName(resourcePath);

// Create input stream for the script file.
InputStreamContent scriptFileStream =
    new InputStreamContent(null, new FileInputStream(scriptPath));

// Create media.upload request.
Media.Upload uploadRequest = service.media().upload(resourcePath, media, scriptFileStream);

// Retrieve uploader from the request and set it to us a simple upload
// request.
MediaHttpUploader uploader = uploadRequest.getMediaHttpUploader();
uploader.setDirectUploadEnabled(true);

// Execute the upload using an Upload URL with the destination resource
// name.
uploader.upload(
    new GenericUrl("https://displayvideo.googleapis.com/upload/media/" + resourcePath));

Python

# Import HTTP objects used for file upload.
from apiclient.http import HttpRequest
from apiclient.http import MediaFileUpload

# Provide the local path to the script file to upload.
script_path = script-path

# Provide the resource path to upload the script file to.
resource_path = resource-path

# Create a media upload object.
media = MediaFileUpload(script_path)

# Create upload request.
upload_request = service.media().upload(
    resourceName=resource_path, media_body=media
)

# Override response handler to expect null response.
upload_request.postproc = HttpRequest.null_postproc

# Upload script to given resource path.
upload_request.execute()

PHP

// Provide the local path to the script file to upload.
$scriptPath = script-path;

// Provide the resource path to upload the script file to.
$resourcePath = resource-path;

// Create the media body.
$mediaBody = new Google_Service_DisplayVideo_GoogleBytestreamMedia();
$mediaBody->setResourceName($resourceName);

// Build params array for the upload request.
$mediaUploadOptParams = array(
    'data' => file_get_contents($scriptPath),
    'uploadType' => 'media',
    'resourceName' => $resourceName
);

try {
    // Call the API, uploading the script file to Display & Video 360.
    $this->service->media->upload(
        $resourceName,
        $mediaBody,
        $mediaUploadOptParams
    );
} catch (\Exception $e) {
    $this->renderError($e);
}

Crea un objeto de secuencia de comandos

Usa una solicitud create para crear un objeto de secuencia de comandos en tu algoritmo.

Si tu archivo de secuencia de comandos tiene un error de sintaxis, fallará tu solicitud de create. La respuesta incluirá detalles sobre el error. Corrige la secuencia de comandos y reinicia el proceso generando un nuevo objeto de referencia.

Precaución: Un algoritmo usará el objeto de secuencia de comandos válido más reciente una vez que se haya entrenado correctamente con los datos disponibles. El entrenamiento puede tardar entre 1 y 3 días. No crees un objeto de secuencia de comandos a menos que tengas la certeza de que deseas reemplazar la secuencia de comandos existente.

Sigue estos pasos para crear un objeto de secuencia de comandos:

Java

// Provide the ID of the advertiser that owns the parent algorithm.
long advertiserId = advertiser-id;

// Provide the ID of the parent algorithm.
long algorithmId = algorithm-id;

// Provide the resource path the script file was uploaded to.
String resourcePath = resource-path;

// Create the custom bidding script structure.
CustomBiddingScript customBiddingScript =
    new CustomBiddingScript()
        .setScript(new CustomBiddingScriptRef().setResourceName(resourcePath));

// Create the custom bidding script.
CustomBiddingScript response =
    service
        .customBiddingAlgorithms()
        .scripts()
        .create(customBiddingAlgorithmId, customBiddingScript)
        .setAdvertiserId(advertiserId)
        .execute();

// Print ID of new custom bidding script.
System.out.printf("Script was created with ID %s.", response.getCustomBiddingScriptId());

Python

# Provide the ID of the advertiser that owns the parent algorithm.
advertiser_id = advertiser-id

# Provide the ID of the parent algorithm.
algorithm_id = algorithm-id

# Provide the resource path the script file was uploaded to.
resource_path = resource-path

# Build script object.
script_obj = {"script": {"resourceName": resource_path}}

# Build and execute request.
script_response = (
    service.customBiddingAlgorithms()
    .scripts()
    .create(
        customBiddingAlgorithmId=algorithm_id,
        advertiserId=advertiser_id,
        body=script_obj,
    )
    .execute()
)

# Print ID of new custom bidding algorithm.
print(
    f'Script was created with ID {script_response["customBiddingScriptId"]}.'
)

PHP

// Provide the ID of the advertiser that owns the parent algorithm.
$advertiserId = advertiser-id;

// Provide the ID of the parent algorithm.
$algorithmId = algorithm-id;

// Provide the resource path the script file was uploaded to.
$resourcePath = resource-path;

// Build Script object.
$customBiddingScript = new Google_Service_DisplayVideo_CustomBiddingScript();
$scriptRef = new Google_Service_DisplayVideo_CustomBiddingScriptRef();
$scriptRef->setResourceName($resourcePath);
$customBiddingScript->setScript($scriptRef);

$createScriptOptParams = array(
    'advertiserId' => $advertiserId
);

// Call the API, creating the custom bidding script using the script
// file and under the advertiser and custom bidding algorithm given.
try {
    $result = $this->service->customBiddingAlgorithms_scripts->create(
        $customBiddingAlgorithmId,
        $algorithmId,
        $createScriptOptParams
    );
} catch (\Exception $e) {
    $this->renderError($e);
    return;
}

// Print ID of new custom bidding script.
printf('<p>Script was created with ID %s.</p>', $result['customBiddingScriptId']);

Estado de la secuencia de comandos de la encuesta

Display & Video 360 procesa la secuencia de comandos después de la creación. Luego, Display & Video 360 acepta o rechaza la secuencia de comandos para calificar las impresiones.

Después de crear tu secuencia de comandos, sondea para asegurarte de que se acepte. Verifica el campo state de la secuencia de comandos resultante:

• Si es PENDING, sigue sondeo la secuencia de comandos.
• Si es ACCEPTED, la secuencia de comandos es válida. Continúa para comprobar si el algoritmo está listo.
• Si es REJECTED, la secuencia de comandos tiene un error. Recupera el campo errors para obtener una lista de errores específicos que se encontraron en la secuencia de comandos. Corrige los errores y reinicia este proceso generando un nuevo objeto de referencia de la secuencia de comandos para subir un archivo de secuencia de comandos corregido.

Nota: Usa la retirada exponencial para conservar la cuota cuando realices sondeos.

A continuación, se explica cómo sondear el estado de la secuencia de comandos con una retirada exponencial:

Java

// Provide the ID of the advertiser that owns the parent algorithm.
long advertiserId = advertiser-id;

// Provide the ID of the parent algorithm.
long algorithmId = algorithm-id;

// Provide the ID of the script.
long scriptId = script-id;

// Configure the Operations.get request.
Scripts.Get scriptRequest =
    service
        .customBiddingAlgorithms()
        .scripts()
        .get(algorithmId, scriptId)
        .setAdvertiserId(advertiserId);

// Configure exponential backoff for checking the status of our operation.
ExponentialBackOff backOff =
    new ExponentialBackOff.Builder()
        .setInitialIntervalMillis(5000) // setting initial interval to five seconds
        .setMaxIntervalMillis(300000) // setting max interval to five minutes
        .setMaxElapsedTimeMillis(18000000) // setting max elapsed time to five hours
        .build();

long backoffMillis = 0;

CustomBiddingScript script = null;

do {
  // Sleep before retrieving operation again.
  Thread.sleep(backoffMillis);

  // Retrieve script.
  script = scriptRequest.execute();

  // If operation is not done, calculate next sleep period.
  if (script.getState() == null || script.getState().equals("PENDING")) {
    backoffMillis = backOff.nextBackOffMillis();
    if (backoffMillis == ExponentialBackOff.STOP) {
      System.out.printf("Polling deadline exceeded.%n");
      break;
    }
    System.out.printf(
        "Script with ID %s is still pending, sleeping for %s milliseconds.%n",
        scriptId, backoffMillis);
  }
} while (script.getState() == null || script.getState().equals("PENDING"));

// Check whether operation finished with an error, or is completed and ready to download.
if (script.getState().equals("ACCEPTED")) {
  System.out.printf("Script with ID %s has been accepted.%n", scriptId);
} else {
  System.out.printf("Script with ID %s has been rejected.%n", scriptId);
  if (script.getErrors() != null && script.getErrors().size() > 0) {
    System.out.println("The following errors were identified:");
    for (ScriptError error : script.getErrors()) {
      System.out.printf(
          "Line %s, Column %s (%s): %s%n",
          error.getLine(), error.getColumn(), error.getErrorCode(), error.getErrorMessage());
    }
  } else {
    System.out.println("No specific errors were identified.");
  }
}

Python

# Provide the ID of the advertiser that owns the parent algorithm.
advertiser_id = advertiser-id

# Provide the ID of the parent algorithm.
algorithm_id = algorithm-id

# Provide the ID of the script.
script_id = script-id

# Set the following values that control retry behavior while the script is
# being processed.
# Minimum amount of time between polling requests. Defaults to 5 seconds.
min_retry_interval = 5
# Maximum amount of time between polling requests. Defaults to 5 minutes.
max_retry_interval = 5 * 60
# Maximum amount of time to spend polling. Defaults to 5 hours.
max_retry_elapsed_time = 5 * 60 * 60

# Generate the next sleep interval using exponential backoff logic.
def next_sleep_interval(previous_sleep_interval):
    """Calculates the next sleep interval based on the previous."""
    min_interval = previous_sleep_interval or min_retry_interval
    max_interval = previous_sleep_interval * 3 or min_retry_interval
    return min(max_retry_interval, random.randint(min_interval, max_interval))

# Configure the script GET request.
get_request = (
    service.customBiddingAlgorithms()
    .scripts()
    .get(
        customBiddingAlgorithmId=algorithm_id,
        customBiddingScriptId=script_id,
        advertiserId=advertiser_id,
    )
)

# Poll the script until the script is finished being processed or the max
# time has fully elapsed.
sleep = 0
start_time = time.time()
while True:
  # Get current status of the script.
  script = get_request.execute()

  # Check whether script was accepted or rejected and print any relevant
  # errors.
  if script["state"] == "ACCEPTED":
    print(f"Script with ID {script_id} has been accepted.")
    break
  elif script["state"] == "REJECTED":
    print(f"Script with ID {script_id} has been rejected.")
    if "errors" in script and len(script["errors"]) > 0:
      print("The following errors were identified:")
      for err in script["errors"]:
        print(
            f'Line {err["line"]}, Column'
            f' {err["column"]} ({err["errorCode"]}): {err["errorMessage"]}'
        )
    else:
      print("No specific errors were identified.")
    break
  elif time.time() - start_time > max_retry_elapsed_time:
    print("Polling deadline exceeded.")
    break

  sleep = next_sleep_interval(sleep)
  print(
      f"Script with ID {script_id} is still pending, sleeping for"
      f" {sleep} seconds."
  )
  time.sleep(sleep)

PHP

// Provide the ID of the advertiser that owns the parent algorithm.
$advertiserId = advertiser-id;

// Provide the ID of the parent algorithm.
$algorithmId = algorithm-id;

// Provide the ID of the script.
$scriptId = script-id;

// The following values control retry behavior while the task is
// processing.
// Minimum amount of time between polling requests. Defaults to 5 seconds.
$minRetryInterval = 5;
// Maximum amount of time between polling requests. Defaults to 5 minutes.
$maxRetryInterval = 300;
// Maximum amount of time to spend polling. Defaults to 5 hours.
$maxRetryElapsedTime = 18000;

$sleep = 0;
$startTime = time();

$getScriptOptParams = array(
    'advertiserId' => $advertiserId
);

do {
    try {
        // Call the API, retrieving the script.
        $script = $this->service->customBiddingAlgorithms_scripts->get(
            $algorithmId,
            $scriptId,
            $getScriptOptParams
        );
    } catch (\Exception $e) {
        $this->renderError($e);
        return;
    }

    if ($script->getState() == "ACCEPTED") {
        printf("<p>Script with ID %s has been accepted.</p>", $scriptId);
        break;
    } elseif ($script->getState() == "REJECTED") {
        printf("Script with ID %s has been rejected.", $scriptId);
        if ($script->getErrors() !== null) {
            print("<p>The following errors were identified:</p><ul>");
            $errors = $script->getErrors();
            foreach ($errors as $error) {
                printf(
                    "<li>Line %s, Column %s (%s): %s</li>",
                    $error->getLine(),
                    $error->getColumn(),
                    $error->getErrorCode(),
                    $error->getErrorMessage());
            }
            print("</ul>");
        } else {
            print("<p>No specific errors were identified.</p>");
        }
        break;
    } elseif (time() - $startTime > $maxRetryElapsedTime) {
        print '<p>Script polling deadline exceeded.</p>';
        break;
    }

    // Generate next sleep interval.
    $minInterval = max($minRetryInterval, $sleep);
    $maxInterval = max(
        $minRetryInterval,
        $sleep * 3
    );
    $sleep = min($maxRetryInterval, rand($minInterval, $maxInterval));

    printf(
        '<p>Script with ID %s is still pending, sleeping for %d seconds</p>',
        $scriptId,
        $sleep
    );
    sleep($sleep);
} while (true);