Skip to contents

Delayed Data Objects

Dawid Kałędkowski

15.05.2022

delayed-data-objects.Rmd

Basic understanding

Delayed data objects are R objects that contain instructions on how to acquire data. In practice, you will pass these DDL objects with their instructions into a teal application so that you can launch a teal app first and then pull the data afterwards.

The main difference between a DDL object and a non-DDL object is that data is available immediately after creating a non-DDL object. In contrast, data in a DDL object is not available after construction, only after pulling it (executing instructions stored in the object).

Key concepts

  • A TealDatasetConnector is an object used to pull a single delayed data set into a teal app. Connectors to pull data from proprietary data storage are not included in this package.
library(teal.data)
## Loading required package: shiny
library(scda)
## 
# specialized function to create delayed data using scda (could also set keys when using cdisc_dataset_connector)
adlb <- scda_dataset_connector(
  "ADLB",
  "adlb",
  keys = get_cdisc_keys("ADLB")
)

# generalized function to create delayed data from code - see package help for other connectors
x <- code_dataset_connector(
  dataname = "ADSL",
  keys = get_cdisc_keys("ADSL"),
  code = "library(scda)\nADSL <- synthetic_cdisc_data(\"latest\")$adsl"
)
  • A TealDataConnector is an object used to pull a set of delayed data sets into a teal app which all share a common connection (see Delayed Data Loading for the definition of a connection object).
# using scda
adsl <- scda_cdisc_dataset_connector(dataname = "ADSL", scda_dataname = "adsl")
adae <- scda_cdisc_dataset_connector(dataname = "ADAE", scda_dataname = "adae")
adsl_adae <- relational_data_connector(
  connection = data_connection(),
  connectors = list(adsl, adae)
)
  • The cdisc_data function takes a set of TealDataConnector, TealDatasetConnector and / or cdisc_datasets (non-delayed datasets) to create the TealData object which is used to create teal applications.
# create a TealDatasetConnector for ADVS
advs <- scda_cdisc_dataset_connector(dataname = "ADVS", scda_dataname = "advs")
# use cdisc_data() to create a `DDL` object
delayed_data <- cdisc_data(adsl_adae, advs)

Constructors

Below is a list of all of the constructors available in teal.data to create TealDataset and delayed TealDatasetConnector objects:

Description Base Constructor Constructor Wrappers
TealDataset Dataframe with name (and optionally keys) dataset, dataset_file dataset, cdisc_dataset
TealDatasetConnector Delayed Dataset dataset_connector, dataset_connector_file (see note 1 below) rds_dataset_connector, script_dataset_connector, code_dataset_connector, csv_dataset_connector, fun_dataset_connector, python_dataset_connector, scda_dataset_connector
TealDataConnector Group of TealDatasetConnector
TealData Group of TealDatasetConnector, TealDataConnector, TealDataset teal_data, teal_data_file (see note 2 below) cdisc_data, cdisc_data_file

Notes:

  1. All xyz_dataset_connector functions have an equivalent xyz_cdisc_dataset_connector function (for example rds_cdisc_dataset_connector) which specifies additional dataset metadata.
  2. cdisc_data is the standard function used to create a data object to be used within teal apps for standard CDISC study data. The more general teal_data function can be used to allow arbitrary relational data to be used within teal apps.

Dataset dependencies

The datasets passed into teal_data and cdisc_data are pulled in the order they are inputted. So if datasets depend on other datasets being available they should be placed later in the argument list:

adsl <- scda_cdisc_dataset_connector(dataname = "ADSL", "adsl")
adsl_2 <- code_cdisc_dataset_connector("ADSL_2",
  code = "head(ADSL, 5)",
  keys = get_cdisc_keys("ADSL"), ADSL = adsl
)

# launch method will be able to load the data as adsl will be pulled first
cdisc_data(adsl, adsl_2)
## A CDISCTealData object containing 2 TealDataset/TealDatasetConnector object(s) as element(s):
## --> Element 1:
## A CDISCTealDatasetConnector object, named ADSL, containing a TealDataset object that has not been loaded/pulled
## --> Element 2:
## A CDISCTealDatasetConnector object, named ADSL_2, containing a TealDataset object that has not been loaded/pulled
# launch method will not be able to load the data as adae is pulled first but it depends on adsl
cdisc_data(adsl_2, adsl)
## A CDISCTealData object containing 2 TealDataset/TealDatasetConnector object(s) as element(s):
## --> Element 1:
## A CDISCTealDatasetConnector object, named ADSL_2, containing a TealDataset object that has not been loaded/pulled
## --> Element 2:
## A CDISCTealDatasetConnector object, named ADSL, containing a TealDataset object that has not been loaded/pulled

Suggested development workflow

The following workflow facilitates building teal apps with DDL by minimizing debugging overhead.

  1. Run a teal app configured without delayed data to verify that the app starts as expected.
  2. Replace the cdisc_dataset functions with the appropriate TealDatasetConnector objects.
  3. Add preprocessing code (see Delayed Data Loading for preprocessing documentation) and verify once again by running the newly created object’s $launch method.
  4. First, include just a single module to verify that the teal app starts and that it loads all the expected data.
  5. Then, add the rest of the modules one by one, replacing all dataset calls with strings and iteratively verify that the app functions as expected.