Skip to content

Latest commit

 

History

History
654 lines (522 loc) · 23.1 KB

README.md

File metadata and controls

654 lines (522 loc) · 23.1 KB

sparklyr: R interface for Apache Spark

Github workflow status AppVeyor build status CRAN_Status_Badge codecov Join the chat at https://gitter.im/sparklyr/sparklyr

  • Install and connect to Spark using YARN, Mesos, Livy or Kubernetes.
  • Use dplyr to filter and aggregate Spark datasets and streams then bring them into R for analysis and visualization.
  • Use MLlib, H2O, XGBoost and GraphFrames to train models at scale in Spark.
  • Create interoperable machine learning pipelines and productionize them with MLeap.
  • Create extensions that call the full Spark API or run distributed R code to support new functionality.

Installation

You can install the sparklyr package from CRAN as follows:

install.packages("sparklyr")

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

library(sparklyr)
spark_install()

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

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 function:

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 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):

install.packages(c("nycflights13", "Lahman"))
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)
## [1] "batting" "flights" "iris"

To start with here’s a simple filtering example:

# filter by departure delay and print the first few records
flights_tbl %>% filter(dep_delay == 2)
## # Source: spark<?> [?? x 19]
##     year month   day dep_time sched_dep_time dep_delay arr_time sched_arr_time
##    <int> <int> <int>    <int>          <int>     <dbl>    <int>          <int>
##  1  2013     1     1      517            515         2      830            819
##  2  2013     1     1      542            540         2      923            850
##  3  2013     1     1      702            700         2     1058           1014
##  4  2013     1     1      715            713         2      911            850
##  5  2013     1     1      752            750         2     1025           1029
##  6  2013     1     1      917            915         2     1206           1211
##  7  2013     1     1      932            930         2     1219           1225
##  8  2013     1     1     1028           1026         2     1350           1339
##  9  2013     1     1     1042           1040         2     1325           1326
## 10  2013     1     1     1231           1229         2     1523           1529
## # … with more rows, and 11 more variables: arr_delay <dbl>, carrier <chr>,
## #   flight <int>, tailnum <chr>, origin <chr>, dest <chr>, air_time <dbl>,
## #   distance <dbl>, hour <dbl>, minute <dbl>, time_hour <dttm>

Introduction to dplyr provides additional dplyr examples you can try. For example, consider the last example from the tutorial which plots data on flight delays:

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)
## `geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'

Window Functions

dplyr window functions are also supported, for example:

batting_tbl %>%
  select(playerID, yearID, teamID, G, AB:H) %>%
  arrange(playerID, yearID, teamID) %>%
  group_by(playerID) %>%
  filter(min_rank(desc(H)) <= 2 & H > 0)
## # Source:     spark<?> [?? x 7]
## # Groups:     playerID
## # Ordered by: playerID, yearID, teamID
##    playerID  yearID teamID     G    AB     R     H
##    <chr>      <int> <chr>  <int> <int> <int> <int>
##  1 aaronha01   1959 ML1      154   629   116   223
##  2 aaronha01   1963 ML1      161   631   121   201
##  3 aaronto01   1962 ML1      141   334    54    77
##  4 aaronto01   1968 ATL       98   283    21    69
##  5 abadfe01    2012 HOU       37     7     0     1
##  6 abbated01   1905 BSN      153   610    70   170
##  7 abbated01   1904 BSN      154   579    76   148
##  8 abbeych01   1894 WAS      129   523    95   164
##  9 abbeych01   1895 WAS      133   516   102   142
## 10 abbotda01   1890 TL2        3     7     0     1
## # … with more rows

For additional documentation on using dplyr with Spark see the dplyr 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 interface for Spark, so you can use dbGetQuery to execute SQL and return the result as an R data frame:

library(DBI)
iris_preview <- dbGetQuery(sc, "SELECT * FROM iris LIMIT 10")
iris_preview
##    Sepal_Length Sepal_Width Petal_Length Petal_Width Species
## 1           5.1         3.5          1.4         0.2  setosa
## 2           4.9         3.0          1.4         0.2  setosa
## 3           4.7         3.2          1.3         0.2  setosa
## 4           4.6         3.1          1.5         0.2  setosa
## 5           5.0         3.6          1.4         0.2  setosa
## 6           5.4         3.9          1.7         0.4  setosa
## 7           4.6         3.4          1.4         0.3  setosa
## 8           5.0         3.4          1.5         0.2  setosa
## 9           4.4         2.9          1.4         0.2  setosa
## 10          4.9         3.1          1.5         0.1  setosa

Machine Learning

You can orchestrate machine learning algorithms in a Spark cluster via the machine learning 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 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.

# 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
## Formula: mpg ~ wt + cyl
##
## Coefficients:
## (Intercept)          wt         cyl
##   33.499452   -2.818463   -0.923187

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.

summary(fit)
## Deviance Residuals:
##    Min     1Q Median     3Q    Max
## -1.752 -1.134 -0.499  1.296  2.282
##
## Coefficients:
## (Intercept)          wt         cyl
##   33.499452   -2.818463   -0.923187
##
## R-Squared: 0.8274
## Root Mean Squared Error: 1.422

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 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.

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)
## [1] "batting"      "flights"      "iris"         "iris_csv"     "iris_json"
## [6] "iris_parquet" "mtcars"

Distributed R

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

spark_apply(iris_tbl, function(data) {
  data[1:4] + rgamma(1,2)
})
## # Source: spark<?> [?? x 4]
##    Sepal_Length Sepal_Width Petal_Length Petal_Width
##           <dbl>       <dbl>        <dbl>       <dbl>
##  1         8.80        7.20         5.10        3.90
##  2         8.60        6.70         5.10        3.90
##  3         8.40        6.90         5.00        3.90
##  4         8.30        6.80         5.20        3.90
##  5         8.70        7.30         5.10        3.90
##  6         9.10        7.60         5.40        4.10
##  7         8.30        7.10         5.10        4.00
##  8         8.70        7.10         5.20        3.90
##  9         8.10        6.60         5.10        3.90
## 10         8.60        6.80         5.20        3.80
## # … with more rows

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

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"
)
## # Source: spark<?> [?? x 6]
##   Species    term         estimate std.error statistic  p.value
##   <chr>      <chr>           <dbl>     <dbl>     <dbl>    <dbl>
## 1 setosa     (Intercept)   -0.0482    0.122     -0.396 6.94e- 1
## 2 setosa     Petal_Length   0.201     0.0826     2.44  1.86e- 2
## 3 versicolor (Intercept)   -0.0843    0.161     -0.525 6.02e- 1
## 4 versicolor Petal_Length   0.331     0.0375     8.83  1.27e-11
## 5 virginica  (Intercept)    1.14      0.379      2.99  4.34e- 3
## 6 virginica  Petal_Length   0.160     0.0680     2.36  2.25e- 2

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:

# 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)
## [1] 336777

To learn more about creating extensions see the Extensions section of the sparklyr website.

Table Utilities

You can cache a table into memory with:

tbl_cache(sc, "batting")

and unload from memory using:

tbl_uncache(sc, "batting")

Connection Utilities

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

spark_web(sc)

You can show the log using the spark_log function:

spark_log(sc, n = 10)
## 20/04/15 11:49:57 INFO DAGScheduler: Submitting 1 missing tasks from ResultStage 70 (/tmp/Rtmp3NF2Jb/filece955cc88fe.csv MapPartitionsRDD[336] at textFile at NativeMethodAccessorImpl.java:0) (first 15 tasks are for partitions Vector(0))
## 20/04/15 11:49:57 INFO TaskSchedulerImpl: Adding task set 70.0 with 1 tasks
## 20/04/15 11:49:57 INFO TaskSetManager: Starting task 0.0 in stage 70.0 (TID 78, localhost, executor driver, partition 0, PROCESS_LOCAL, 7900 bytes)
## 20/04/15 11:49:57 INFO Executor: Running task 0.0 in stage 70.0 (TID 78)
## 20/04/15 11:49:57 INFO HadoopRDD: Input split: file:/tmp/Rtmp3NF2Jb/filece955cc88fe.csv:0+33313106
## 20/04/15 11:49:57 INFO Executor: Finished task 0.0 in stage 70.0 (TID 78). 832 bytes result sent to driver
## 20/04/15 11:49:57 INFO TaskSetManager: Finished task 0.0 in stage 70.0 (TID 78) in 196 ms on localhost (executor driver) (1/1)
## 20/04/15 11:49:57 INFO TaskSchedulerImpl: Removed TaskSet 70.0, whose tasks have all completed, from pool
## 20/04/15 11:49:57 INFO DAGScheduler: ResultStage 70 (count at NativeMethodAccessorImpl.java:0) finished in 0.200 s
## 20/04/15 11:49:57 INFO DAGScheduler: Job 47 finished: count at NativeMethodAccessorImpl.java:0, took 0.204883 s

Finally, we disconnect from Spark:

  spark_disconnect(sc)

RStudio IDE

The latest RStudio Preview Release 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:

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:

You can also connect to Spark through Livy through a new connection dialog:

The RStudio IDE features for sparklyr are available now as part of the RStudio Preview Release.

Using H2O

rsparkling is a CRAN package from H2O that extends sparklyr to provide an interface into Sparkling Water. For instance, the following example installs, configures and runs h2o.glm:

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)
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
spark_disconnect(sc)

Connecting through Livy

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:

livy_install()
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:

sc <- spark_connect(master = "http://localhost:8998", method = "livy", version = "2.4.0")
copy_to(sc, iris, overwrite = TRUE)
## # Source: spark<iris> [?? x 5]
##    Sepal_Length Sepal_Width Petal_Length Petal_Width Species
##           <dbl>       <dbl>        <dbl>       <dbl> <chr>
##  1          5.1         3.5          1.4         0.2 setosa
##  2          4.9         3            1.4         0.2 setosa
##  3          4.7         3.2          1.3         0.2 setosa
##  4          4.6         3.1          1.5         0.2 setosa
##  5          5           3.6          1.4         0.2 setosa
##  6          5.4         3.9          1.7         0.4 setosa
##  7          4.6         3.4          1.4         0.3 setosa
##  8          5           3.4          1.5         0.2 setosa
##  9          4.4         2.9          1.4         0.2 setosa
## 10          4.9         3.1          1.5         0.1 setosa
## # … with more rows
spark_disconnect(sc)

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

livy_service_stop()

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

config <- livy_config(username="<username>", password="<password>")
sc <- spark_connect(master = "<address>", method = "livy", config = config)
spark_disconnect(sc)

To connect to remote livy clusters with Kerberos authentication, you should first get a valid Kerberos ticket on the local client, then:

config <- livy_config(negotiate=TRUE, custom_headers=list("X-Requested-By"="<user>"))
sc <- spark_connect(master = "<address>", method = "livy", config = config)
spark_disconnect(sc)

The remote livy cluster will require outgoing connection to GitHub in order to download sparklyr jars. You can avoid this by uploading the required jar to HDFS and connect as:

config <- livy_config(...)
config$sparklyr.livy.jar = "hdfs://path/to/sparklyr-X.X-Y.Y.jar"
sc <- spark_connect(master = "<address>", method = "livy", config = config)
spark_disconnect(sc)

Connecting through Databricks Connect

Databricks Connect allows you to connect sparklyr to a remote Databricks Cluster. You can install Databricks Connect Python pakcage 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 to setup the client:

  1. Make sure pyspark is not installed
  2. Install the latest version of Databricks Connect python package.
  3. Run databricks-connect configure and provide the configuration information
    • Databricks account URL of the form ttps://<account>.cloud.databricks.com.
    • User token
    • Cluster ID
    • Port (default port number is 15001)

Now simply create a spark connection as follows

spark_home <- system("databricks-connect get-spark-home")
sc <- spark_connect(method = "databricks",
                    spark_home = spark_home)
copy_to(sc, iris, overwrite = TRUE)

spark_disconnect(sc)