README.Rmd

---
title: "sparklyr: R interface for Apache Spark"
output:
  github_document:
    fig_width: 9
    fig_height: 5
---

![Github workflow status](https://github.com/sparklyr/sparklyr/workflows/CI/badge.svg) [![AppVeyor build status](https://ci.appveyor.com/api/projects/status/lxaj4kgu3liovmhp?svg=true)](https://ci.appveyor.com/project/yl790/sparklyr-ybp7x) [![CRAN_Status_Badge](https://www.r-pkg.org/badges/version/sparklyr)](https://cran.r-project.org/package=sparklyr) <a href="https://www.r-pkg.org/pkg/sparklyr"><img src="https://cranlogs.r-pkg.org/badges/sparklyr?color=brightgreen" style=""></a> [![codecov](https://codecov.io/gh/sparklyr/sparklyr/branch/master/graph/badge.svg)](https://codecov.io/gh/sparklyr/sparklyr) [![Join the chat at https://gitter.im/sparklyr/sparklyr](https://badges.gitter.im/sparklyr/sparklyr.svg)](https://gitter.im/sparklyr/sparklyr?utm_source=badge&utm_medium=badge&utm_campaign=pr-badge&utm_content=badge)

<img src="tools/readme/sparklyr-illustration.png" width="320" align="right" style="margin-left: 20px; margin-right: 20px"/>

```{r setup, include=FALSE}
knitr::opts_chunk$set(eval = TRUE)
knitr::opts_chunk$set(warning = FALSE)
knitr::opts_chunk$set(fig.path = "tools/readme/", dev = "png")
```

- Install and connect to [Spark](http://spark.apache.org/) using YARN, Mesos, Livy or Kubernetes.
- Use [dplyr](https://spark.rstudio.com/dplyr/) to filter and aggregate Spark datasets and [streams](https://spark.rstudio.com/guides/streaming/) then bring them into R for analysis and visualization.
- Use [MLlib](http://spark.apache.org/docs/latest/mllib-guide.html), [H2O](https://spark.rstudio.com/guides/h2o/), [XGBoost](https://github.com/rstudio/sparkxgb) and [GraphFrames](https://github.com/rstudio/graphframes) to train models at scale in Spark.
- Create interoperable machine learning [pipelines](https://spark.rstudio.com/guides/pipelines/) and productionize them with [MLeap](https://spark.rstudio.com/guides/mleap/).
- Create [extensions](https://spark.rstudio.com/extensions.html) that call the full Spark API or run [distributed R](https://spark.rstudio.com/guides/distributed-r/) code to support new functionality.

## Installation

You can install the **sparklyr** package from [CRAN](https://CRAN.r-project.org) as follows:

```{r, eval=FALSE}
install.packages("sparklyr")
```

You should also install a local version of Spark for development purposes:

```{r, eval=FALSE}
library(sparklyr)
spark_install()
```

To upgrade to the latest version of sparklyr, run the following command and restart your r session:

```{r, eval=FALSE}
install.packages("devtools")
devtools::install_github("sparklyr/sparklyr")
```

## Connecting to Spark

You can connect to both local instances of Spark as well as remote Spark clusters. Here we'll connect to a local instance of Spark via the [spark_connect](https://spark.rstudio.com/reference/spark-connections/) function:

```{r sparklyr-connect, message=FALSE}
library(sparklyr)
sc <- spark_connect(master = "local")
```

The returned Spark connection (`sc`) provides a remote dplyr data source to the Spark cluster.

For more information on connecting to remote Spark clusters see the [Deployment](https://spark.rstudio.com/deployment.html) section of the sparklyr website.

## Using dplyr

We can now use all of the available dplyr verbs against the tables within the cluster.

We'll start by copying some datasets from R into the Spark cluster (note that you may need to install the nycflights13 and Lahman packages in order to execute this code):

```{r, eval=FALSE}
install.packages(c("nycflights13", "Lahman"))
```

```{r dplyr-copy, message=FALSE}
library(dplyr)
iris_tbl <- copy_to(sc, iris, overwrite = TRUE)
flights_tbl <- copy_to(sc, nycflights13::flights, "flights", overwrite = TRUE)
batting_tbl <- copy_to(sc, Lahman::Batting, "batting", overwrite = TRUE)
src_tbls(sc)
```

To start with here's a simple filtering example:

```{r dplyr-filter}
# filter by departure delay and print the first few records
flights_tbl %>% filter(dep_delay == 2)
```

[Introduction to dplyr](https://CRAN.R-project.org/package=dplyr) provides additional dplyr examples you can try. For example, consider the last example from the tutorial which plots data on flight delays:

```{r dplyr-ggplot2}
delay <- flights_tbl %>%
  group_by(tailnum) %>%
  summarise(count = n(), dist = mean(distance), delay = mean(arr_delay)) %>%
  filter(count > 20, dist < 2000, !is.na(delay)) %>%
  collect

# plot delays
library(ggplot2)
ggplot(delay, aes(dist, delay)) +
  geom_point(aes(size = count), alpha = 1/2) +
  geom_smooth() +
  scale_size_area(max_size = 2)
```


### Window Functions

dplyr [window functions](https://CRAN.R-project.org/package=dplyr) are also supported, for example:

```{r dplyr-window}
batting_tbl %>%
  select(playerID, yearID, teamID, G, AB:H) %>%
  arrange(playerID, yearID, teamID) %>%
  group_by(playerID) %>%
  filter(min_rank(desc(H)) <= 2 & H > 0)
```

For additional documentation on using dplyr with Spark see the [dplyr](https://spark.rstudio.com/dplyr.html) section of the sparklyr website.

## Using SQL

It's also possible to execute SQL queries directly against tables within a Spark cluster. The `spark_connection` object implements a [DBI](https://github.com/rstats-db/DBI) interface for Spark, so you can use `dbGetQuery` to execute SQL and return the result as an R data frame:

```{r sql-dbi}
library(DBI)
iris_preview <- dbGetQuery(sc, "SELECT * FROM iris LIMIT 10")
iris_preview
```

## Machine Learning

You can orchestrate machine learning algorithms in a Spark cluster via the [machine learning](http://spark.apache.org/docs/latest/mllib-guide.html) functions within **sparklyr**. These functions connect to a set of high-level APIs built on top of DataFrames that help you create and tune machine learning workflows.

Here's an example where we use [ml_linear_regression](https://spark.rstudio.com/reference/ml_linear_regression/) to fit a linear regression model. We'll use the built-in `mtcars` dataset, and see if we can predict a car's fuel consumption (`mpg`) based on its weight (`wt`), and the number of cylinders the engine contains (`cyl`). We'll assume in each case that the relationship between `mpg` and each of our features is linear.

```{r}
# copy mtcars into spark
mtcars_tbl <- copy_to(sc, mtcars, overwrite = TRUE)

# transform our data set, and then partition into 'training', 'test'
partitions <- mtcars_tbl %>%
  filter(hp >= 100) %>%
  mutate(cyl8 = cyl == 8) %>%
  sdf_partition(training = 0.5, test = 0.5, seed = 1099)

# fit a linear model to the training dataset
fit <- partitions$training %>%
  ml_linear_regression(response = "mpg", features = c("wt", "cyl"))
fit
```

For linear regression models produced by Spark, we can use `summary()` to learn a bit more about the quality of our fit, and the statistical significance of each of our predictors.

```{r}
summary(fit)
```

Spark machine learning supports a wide array of algorithms and feature transformations and as illustrated above it's easy to chain these functions together with dplyr pipelines. To learn more see the [machine learning](https://spark.rstudio.com/mlib/) section.

## Reading and Writing Data

You can read and write data in CSV, JSON, and Parquet formats. Data can be stored in HDFS, S3, or on the local filesystem of cluster nodes.

```{r}
temp_csv <- tempfile(fileext = ".csv")
temp_parquet <- tempfile(fileext = ".parquet")
temp_json <- tempfile(fileext = ".json")

spark_write_csv(iris_tbl, temp_csv)
iris_csv_tbl <- spark_read_csv(sc, "iris_csv", temp_csv)

spark_write_parquet(iris_tbl, temp_parquet)
iris_parquet_tbl <- spark_read_parquet(sc, "iris_parquet", temp_parquet)

spark_write_json(iris_tbl, temp_json)
iris_json_tbl <- spark_read_json(sc, "iris_json", temp_json)

src_tbls(sc)
```


## Distributed R

You can execute arbitrary r code across your cluster using `spark_apply`. For example, we can apply `rgamma` over `iris` as follows:

```{r}
spark_apply(iris_tbl, function(data) {
  data[1:4] + rgamma(1,2)
})
```

You can also group by columns to perform an operation over each group of rows and make use of any package within the closure:

```{r}
spark_apply(
  iris_tbl,
  function(e) broom::tidy(lm(Petal_Width ~ Petal_Length, e)),
  columns = c("term", "estimate", "std.error", "statistic", "p.value"),
  group_by = "Species"
)
```

## Extensions

The facilities used internally by sparklyr for its dplyr and machine learning interfaces are available to extension packages. Since Spark is a general purpose cluster computing system there are many potential applications for extensions (e.g. interfaces to custom machine learning pipelines, interfaces to 3rd party Spark packages, etc.).

Here's a simple example that wraps a Spark text file line counting function with an R function:

```{r}
# write a CSV
tempfile <- tempfile(fileext = ".csv")
write.csv(nycflights13::flights, tempfile, row.names = FALSE, na = "")

# define an R interface to Spark line counting
count_lines <- function(sc, path) {
  spark_context(sc) %>%
    invoke("textFile", path, 1L) %>%
      invoke("count")
}

# call spark to count the lines of the CSV
count_lines(sc, tempfile)
```


To learn more about creating extensions see the [Extensions](https://spark.rstudio.com/extensions.html) section of the sparklyr website.


## Table Utilities

You can cache a table into memory with:

```{r, eval=FALSE}
tbl_cache(sc, "batting")
```

and unload from memory using:

```{r, eval=FALSE}
tbl_uncache(sc, "batting")
```


## Connection Utilities

You can view the Spark web console using the `spark_web` function:

```{r, eval=FALSE}
spark_web(sc)
```

You can show the log using the `spark_log` function:

```{r}
spark_log(sc, n = 10)
```

Finally, we disconnect from Spark:

```{r}
  spark_disconnect(sc)
```

## RStudio IDE

The latest RStudio [Preview Release](https://www.rstudio.com/products/rstudio/download/preview/) of the RStudio IDE includes integrated support for Spark and the sparklyr package, including tools for:

- Creating and managing Spark connections
- Browsing the tables and columns of Spark DataFrames
- Previewing the first 1,000 rows of Spark DataFrames

Once you've installed the sparklyr package, you should find a new **Spark** pane within the IDE. This pane includes a **New Connection** dialog which can be used to make connections to local or remote Spark instances:

<img src="tools/readme/spark-connect.png" class="screenshot" width=389 />

Once you've connected to Spark you'll be able to browse the tables contained within the Spark cluster and preview Spark DataFrames using the standard RStudio data viewer:

<img src="tools/readme/spark-dataview.png" class="screenshot" width=639 />

You can also connect to Spark through [Livy](http://livy.io) through a new connection dialog:

<img src="tools/readme/spark-connect-livy.png" class="screenshot" width=389 />

<div style="margin-bottom: 15px;"></div>

The RStudio IDE features for sparklyr are available now as part of the [RStudio Preview Release](https://www.rstudio.com/products/rstudio/download/preview/).

## Using H2O

[rsparkling](https://cran.r-project.org/package=rsparkling) is a CRAN package from [H2O](https://h2o.ai/) that extends [sparklyr](https://spark.rstudio.com/) to provide an interface into [Sparkling Water](https://github.com/h2oai/sparkling-water). For instance, the following example installs, configures and runs [h2o.glm](http://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/glm.html):

```{r results='hide', message=FALSE, eval=FALSE}
library(rsparkling)
library(sparklyr)
library(dplyr)
library(h2o)

sc <- spark_connect(master = "local", version = "2.3.2")
mtcars_tbl <- copy_to(sc, mtcars, "mtcars", overwrite = TRUE)

mtcars_h2o <- as_h2o_frame(sc, mtcars_tbl, strict_version_check = FALSE)

mtcars_glm <- h2o.glm(x = c("wt", "cyl"),
                      y = "mpg",
                      training_frame = mtcars_h2o,
                      lambda_search = TRUE)
```

```{r eval=FALSE}
mtcars_glm
```
```
## Model Details:
## ==============
##
## H2ORegressionModel: glm
## Model ID:  GLM_model_R_1527265202599_1
## GLM Model: summary
##     family     link                              regularization
## 1 gaussian identity Elastic Net (alpha = 0.5, lambda = 0.1013 )
##                                                                lambda_search
## 1 nlambda = 100, lambda.max = 10.132, lambda.min = 0.1013, lambda.1se = -1.0
##   number_of_predictors_total number_of_active_predictors
## 1                          2                           2
##   number_of_iterations                                training_frame
## 1                  100 frame_rdd_31_ad5c4e88ec97eb8ccedae9475ad34e02
##
## Coefficients: glm coefficients
##       names coefficients standardized_coefficients
## 1 Intercept    38.941654                 20.090625
## 2       cyl    -1.468783                 -2.623132
## 3        wt    -3.034558                 -2.969186
##
## H2ORegressionMetrics: glm
## ** Reported on training data. **
##
## MSE:  6.017684
## RMSE:  2.453097
## MAE:  1.940985
## RMSLE:  0.1114801
## Mean Residual Deviance :  6.017684
## R^2 :  0.8289895
## Null Deviance :1126.047
## Null D.o.F. :31
## Residual Deviance :192.5659
## Residual D.o.F. :29
## AIC :156.2425
```

```{r eval=FALSE}
spark_disconnect(sc)
```

## Connecting through Livy

[Livy](https://github.com/cloudera/livy) enables remote connections to Apache Spark clusters. However, please notice that connecting to Spark clusters through Livy is much slower than any other connection method.

Before connecting to Livy, you will need the connection information to an existing service running Livy. Otherwise, to test `livy` in your local environment, you can install it and run it locally as follows:

```{r eval=FALSE}
livy_install()
```

```{r livy-start}
livy_service_start()
```

To connect, use the Livy service address as `master` and `method = "livy"` in `spark_connect`. Once connection completes, use `sparklyr` as usual, for instance:

```{r livy-connect}
sc <- spark_connect(master = "http://localhost:8998", method = "livy", version = "2.4.0")
copy_to(sc, iris, overwrite = TRUE)

spark_disconnect(sc)
```

Once you are done using `livy` locally, you should stop this service with:

```{r livy-stop}
livy_service_stop()
```

To connect to remote `livy` clusters that support basic authentication connect as:

```{r eval=FALSE}
config <- livy_config(username="<username>", password="<password>")
sc <- spark_connect(master = "<address>", method = "livy", config = config)
spark_disconnect(sc)
```

## Connecting through Databricks Connect
[Databricks Connect](https://docs.databricks.com/dev-tools/databricks-connect.html#databricks-connect) allows you to connect
sparklyr to a remote Databricks Cluster. You can install [Databricks Connect python package](https://pypi.org/project/databricks-connect/) and use it to submit Spark jobs written in sparklyr APIs and have them execute remotely on a Databricks cluster instead of in the local Spark session.

To use sparklyr with Databricks Connect first launch a Cluster on Databricks. Then follow [these instructions](https://docs.databricks.com/dev-tools/databricks-connect.html#client-setup) to setup the client:

1. Make sure pyspark is not installed
2. Install the Databricks Connect python package. The latest supported version is 6.4.1.
3. Run `databricks-connect configure` and provide the configuration information
    * Databricks account URL of the form `https://<account>.cloud.databricks.com`.
    * [User token](https://docs.databricks.com/dev-tools/api/latest/authentication.html#token-management)
    * Cluster ID
    * Port (default port number is `15001`)

To configure `sparklyr` with Databricks Connect, set the following environment variables:
```{bash, eval=FALSE}
export SPARK_VERSION=2.4.4
```

Now simply create a spark connection as follows
```{r, eval=FALSE}
spark_home <- system("databricks-connect get-spark-home")
sc <- spark_connect(method = "databricks",
                    spark_home = spark_home)
copy_to(sc, iris, overwrite = TRUE)
```

<img src="tools/readme/databricks-connect.png" class="screenshot" width=750 />

```{r eval=FALSE}
spark_disconnect(sc)
```