key setting in data.table

keys.Rmd

---
title: "data.table keys"
author: "Steph Locke"
date: "4 April 2016"
output: 
  md_document:
    variant: markdown_github
---

# Glossary
 - **KEY**
 - **COMPOSITE KEY**
  
# Key setting
You can set keys on data.tables to facilitate joins, improve querying speed, and to sort your data.  You can set a key as you create a data.table with `data.table()`, and you can also set keys with dedicated functions, chiefly `setkey()` and `set2key()`.

The iris dataset will be used throughout.
```{r}
library(data.table)
head(setDT(copy(iris)))
```

## `data.table()`
You can create keys as you create data.tables.

### `data.table()`
When you make a data.table object via `data.table()` there is an argument `key=`. `key=` allows you to set a key as you produce a data.table - this will perform sorting like `setkey()` would.
```{r}
irisDT<-data.table(iris, key="Sepal.Width")
head(irisDT)
```

### `setDT()`
Alternatively, the fast setting of a data.frame to data.table function `setDT()` also has a `key=` argument.
```{r}
irisDT<-setDT(copy(iris), key="Sepal.Width")
head(irisDT)
```

## `setkey()`
`setkey()` assigns a key and performs physical sorting on the table.
```{r}
irisDT<-setDT(copy(iris))
setkey(irisDT,Sepal.Length)
head(irisDT)
```

It's possible to make a composite key:
```{r}
irisDT<-setDT(copy(iris))
setkey(irisDT, Sepal.Length, Sepal.Width)
head(irisDT)
```

The `setkey()` function takes named arguments but sometimes you may want to dynamically pass in column names. For this you can use the "v" variant `setkeyv()`:

```{r}
irisDT<-setDT(copy(iris))
key<-c("Sepal.Width","Sepal.Length")
setkeyv(irisDT,key )
head(irisDT)
```

## `set2key()`
`set2key()` assigns a key and **does not** perform physical sorting on the table.
```{r}
irisDT<-setDT(copy(iris))
set2key(irisDT,Sepal.Length)
head(irisDT)
```

It's possible to make a composite key:
```{r}
irisDT<-setDT(copy(iris))
set2key(irisDT, Sepal.Length, Sepal.Width)
head(irisDT)
```

The `set2key()` function takes named arguments but sometimes you may want to dynamically pass in column names. For this you can use the "v" variant `set2keyv()`:

```{r}
irisDT<-setDT(copy(iris))
key<-c("Sepal.Width","Sepal.Length")
set2keyv(irisDT,key )
head(irisDT)
```

-----
[Rmd file](https://gist.github.com/stephlocke/dcb5f0dad688b69afb712f7c55824eea)