Merge pull request apache#98 from shivaram/windows-fixes-build

[SPARKR-113] Build scripts for Windows

Author: concretevitamin

BUILDING.md

@@ -0,0 +1,17 @@
+## Building SparkR on Windows
+
+To build SparkR on Windows, the following steps are required
+
+1. Install R (>= 3.1) and [Rtools](http://cran.r-project.org/bin/windows/Rtools/). Make sure to
+include Rtools and R in `PATH`.
+2. Install
+[JDK7](http://www.oracle.com/technetwork/java/javase/downloads/jdk7-downloads-1880260.html) and set
+`JAVA_HOME` in the system environment variables.
+3. Install `rJava` using `install.packages(rJava)`. If rJava fails to load due to missing jvm.dll,
+you will need to add the directory containing jvm.dll to `PATH`. See this [stackoverflow post](http://stackoverflow.com/a/7604469]
+for more details.
+4. Download and install [Maven](http://maven.apache.org/download.html). Also include the `bin`
+directory in Maven in `PATH`.
+5. Get SparkR source code either using [`git]`(http://git-scm.com/downloads) or by downloading a
+source zip from github.
+6. Open a command shell (`cmd`) in the SparkR directory and run `install-dev.bat`

install-dev.bat

@@ -0,0 +1,8 @@
+@echo off
+
+rem Install development version of SparkR
+rem
+
+MKDIR .\lib
+
+R.exe CMD INSTALL --library=".\lib" pkg\

pkg/R/sparkR.R

@@ -4,6 +4,7 @@ assemblyJarName <- "sparkr-assembly-0.1.jar"
 
 sparkR.onLoad <- function(libname, pkgname) {
   assemblyJarPath <- paste(libname, "/SparkR/", assemblyJarName, sep="")
+  assemblyJarPath <- gsub(" ", "\\ ", assemblyJarPath, fixed=T)
   packageStartupMessage("[SparkR] Initializing with classpath ", assemblyJarPath, "\n")
 
   sparkMem <- Sys.getenv("SPARK_MEM", "512m")
@@ -67,16 +68,19 @@ sparkR.init <- function(
   }
 
   .jaddClassPath(sparkJars)
-  jars=c(as.character(.sparkREnv$assemblyJarPath), as.character(sparkJars))
-  
+  jars <- c(as.character(.sparkREnv$assemblyJarPath), as.character(sparkJars))
+
+  nonEmptyJars <- Filter(function(x) { x != "" }, jars)
+  localJarPaths <- sapply(nonEmptyJars, function(j) { paste("file://", j, sep="") })
+
   assign(
     ".sparkRjsc",
     J("edu.berkeley.cs.amplab.sparkr.RRDD",
       "createSparkContext",
       master,
       appName,
       as.character(sparkHome),
-      .jarray(jars, "java/lang/String"),
+      .jarray(localJarPaths, "java/lang/String"),
       sparkEnvirMap,
       sparkExecutorEnvMap),
     envir=.sparkREnv

pkg/src/Makefile.win

@@ -0,0 +1,35 @@
+SCALA_VERSION := 2.10
+SPARKR_VERSION := 0.1
+JAR_NAME := sparkr-assembly-$(SPARKR_VERSION).jar
+SBT_TARGET_NAME := target/scala-$(SCALA_VERSION)/$(JAR_NAME)
+
+MAVEN_JAR_NAME := sparkr-$(SPARKR_VERSION)-assembly.jar
+MAVEN_TARGET_NAME := target/$(MAVEN_JAR_NAME)
+
+SCALA_SOURCE_DIR := src/main/scala/edu/berkeley/cs/amplab/sparkr
+RESOURCE_DIR := src/main/resources
+
+SCALA_FILES := $(wildcard $(SCALA_SOURCE_DIR)/*.scala)
+RESOURCE_FILES := $(wildcard $(RESOURCE_DIR)/*)
+
+SPARK_VERSION ?= 1.1.0
+SPARK_HADOOP_VERSION ?= 1.0.4
+SPARK_YARN_VERSION ?= 2.4.0
+
+TARGET_NAME := $(MAVEN_TARGET_NAME)
+
+all: $(TARGET_NAME)
+
+$(MAVEN_TARGET_NAME): $(SCALA_FILES) $(RESOURCE_FILES)
+	mvn.bat -Dhadoop.version=$(SPARK_HADOOP_VERSION) -Dspark.version=$(SPARK_VERSION) -Dyarn.version=$(SPARK_YARN_VERSION) -DskipTests clean package shade:shade
+	cp -f $(MAVEN_TARGET_NAME) ../inst/$(JAR_NAME)
+
+clean:
+	mvn.bat clean
+	rm -rf target
+	rm -rf project/target
+	rm -rf project/project
+	-rm sbt/sbt-launch-*.jar
+	rm -f ../inst/$(JAR_NAME)
+
+.PHONY: all clean