Introduction to teal.data

NEST CoreDev

Introduction

The teal.data package specifies the data format used in teal applications.

A teal_data is meant to be used for reproducibility purposes. The class inherits from qenv and we encourage to get familiar with teal.code first. teal_data has following characteristics:

• It inherits from the environment and methods such as $, get(), ls(), as.list() work out of the box.
• teal_data is a locked environment, and data modification is only possible through the teal.code::eval_code() and within.qenv() functions.
• It stores metadata about the code used to create the data (see reproducibility).
• It supports slicing by [.
• It is immutable which means that each code evaluation does not modify the original teal_data environment directly.
• It maintains information about relationships between datasets (see Join Keys).

Quick Start

To create an object of class teal_data, use the teal_data function. teal_data has a number of methods to interact with the object.

library(teal.data)

# create teal_data object
my_data <- teal_data()

# run code within teal_data to create data objects
my_data <- within(
  my_data,
  {
    data1 <- data.frame(id = 1:10, x = 11:20)
    data2 <- data.frame(id = 1:10, x = 21:30)
    data3 <- data.frame(id = 1:10, x = 31:40)
  }
)

# get objects stored in teal_data
my_data[["data1"]]
my_data[["data2"]]

# limit objects stored in teal_data
my_data[c("data1", "data3")]

# get reproducible code
get_code(my_data)

# get code just for specific object
get_code(my_data, names = "data2")

# get datanames
names(my_data)

# print
print(my_data)

Reproducibility

The primary function of teal_data is to provide reproducibility of data. We recommend to initialize empty teal_data, which marks object as verified, and create datasets by evaluating code in the object, using within or eval_code. Read more in teal_data Reproducibility.

my_data <- teal_data()
my_data <- within(my_data, data <- data.frame(x = 11:20))
my_data <- within(my_data, data$id <- seq_len(nrow(data)))
my_data # is verified

## ✅︎ code verified
## <environment: 0x55e0913c3f38> 🔒 
## Parent: <environment: package:teal.data> 
## Bindings:
## - data: [data.frame]

Relational data models

The teal_data class supports relational data. Relationships between datasets can be described by joining keys and stored in a teal_data object. These relationships can be read or set with the join_keys function. See more in join_keys.

my_data <- teal_data()
my_data <- within(my_data, {
  data <- data.frame(id = 1:10, x = 11:20)
  child <- data.frame(id = 1:20, data_id = c(1:10, 1:10), y = 21:30)
})

join_keys(my_data) <- join_keys(
  join_key("data", "data", key = "id"),
  join_key("child", "child", key = "id"),
  join_key("child", "data", key = c("data_id" = "id"))
)

join_keys(my_data)

## A join_keys object containing foreign keys between 2 datasets:
## child: [id]
##   <-- data: [id]
## data: [id]
##   --> child: [data_id]

# join_keys for limited object
join_keys(my_data["child"])

## A join_keys object containing foreign keys between 1 datasets:
## child: [id]

Hidden objects

An object is hidden in teal_data if its name starts with a dot (.). This can be used to pass auxiliary objects in the teal_data instance, without being visible in the teal summary and filter panel.

my_data <- teal_data()
my_data <- within(my_data, {
  data <- data.frame(id = 1:10, x = 11:20)
  .data2 <- data.frame(id = 1:20, data_id = c(1:10, 1:10), y = 21:30)
})

ls(my_data)

## [1] "data"

names(my_data)

## [1] "data"