Merge changes from github.

PiperOrigin-RevId: 202585094

6 files changed, 107 insertions, 2 deletions

diff --git a/tensorflow/java/maven/.gitignore b/tensorflow/java/maven/.gitignore
index ff080515d5..657e2a60bc 100644
--- a/tensorflow/java/maven/.gitignore
+++ b/tensorflow/java/maven/.gitignore
@@ -11,4 +11,10 @@ tensorflow/src
 tensorflow/target
 proto/src
 proto/target
+hadoop/src
+hadoop/target
+spark-connector/src
+spark-connector/target
+spark-connector/dependency-reduced-pom.xml
+spark-connector/spark-warehouse
 pom.xml.versionsBackup
diff --git a/tensorflow/java/maven/README.md b/tensorflow/java/maven/README.md
index c7e8f03806..3e030dcd09 100644
--- a/tensorflow/java/maven/README.md
+++ b/tensorflow/java/maven/README.md
@@ -53,6 +53,12 @@ There are seven artifacts and thus `pom.xml`s involved in this release:
 7.  [`parentpom`](https://maven.apache.org/pom/index.html): Common settings
     shared by all of the above.
 
+8. `hadoop`: The TensorFlow TFRecord InputFormat/OutputFormat for Apache Hadoop.
+    The source code for this package is available in the [TensorFlow Ecosystem](https://github.com/tensorflow/ecosystem/tree/master/hadoop)
+
+9. `spark-connector`: A Scala library for loading and storing TensorFlow TFRecord
+    using Apache Spark DataFrames. The source code for this package is available
+    in the [TensorFlow Ecosystem](https://github.com/tensorflow/ecosystem/tree/master/spark/spark-tensorflow-connector)
 
 ## Updating the release
 
diff --git a/tensorflow/java/maven/hadoop/pom.xml b/tensorflow/java/maven/hadoop/pom.xml
new file mode 100644
index 0000000000..0642be06fa
--- /dev/null
+++ b/tensorflow/java/maven/hadoop/pom.xml
@@ -0,0 +1,24 @@
+<project
+        xmlns="http://maven.apache.org/POM/4.0.0"
+        xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
+        xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
+    <!-- Placeholder pom which is replaced by TensorFlow ecosystem Hadoop pom during build -->
+    <modelVersion>4.0.0</modelVersion>
+    <description>TensorFlow TFRecord InputFormat/OutputFormat for Apache Hadoop</description>
+    <artifactId>hadoop</artifactId>
+    <packaging>jar</packaging>
+
+    <scm>
+        <url>https://github.com/tensorflow/ecosystem.git</url>
+        <connection>git@github.com:tensorflow/ecosystem.git</connection>
+        <developerConnection>scm:git:https://github.com/tensorflow/ecosystem.git</developerConnection>
+    </scm>
+
+    <url>https://github.com/tensorflow/ecosystem/</url>
+    <parent>
+        <groupId>org.tensorflow</groupId>
+        <artifactId>parentpom</artifactId>
+        <version>1.9.0-rc0</version>
+        <relativePath>../</relativePath>
+    </parent>
+</project>
\ No newline at end of file
diff --git a/tensorflow/java/maven/pom.xml b/tensorflow/java/maven/pom.xml
index 3890f3fcaa..b4746794ea 100644
--- a/tensorflow/java/maven/pom.xml
+++ b/tensorflow/java/maven/pom.xml
@@ -32,6 +32,8 @@
     <module>libtensorflow_jni_gpu</module>
     <module>tensorflow</module>
     <module>proto</module>
+    <module>hadoop</module>
+    <module>spark-connector</module>
   </modules>
 
   <!-- Two profiles are used:
diff --git a/tensorflow/java/maven/run_inside_container.sh b/tensorflow/java/maven/run_inside_container.sh
index bf19c09b1d..2e771064e4 100644
--- a/tensorflow/java/maven/run_inside_container.sh
+++ b/tensorflow/java/maven/run_inside_container.sh
@@ -19,6 +19,7 @@
 
 
 RELEASE_URL_PREFIX="https://storage.googleapis.com/tensorflow/libtensorflow"
+TF_ECOSYSTEM_URL="https://github.com/tensorflow/ecosystem.git"
 
 # By default we deploy to both ossrh and bintray. These two
 # environment variables can be set to skip either repository.
@@ -44,7 +45,9 @@ clean() {
   # (though if run inside a clean docker container, there won't be any dirty
   # artifacts lying around)
   mvn -q clean
-  rm -rf libtensorflow_jni/src libtensorflow_jni/target libtensorflow_jni_gpu/src libtensorflow_jni_gpu/target libtensorflow/src libtensorflow/target tensorflow-android/target
+  rm -rf libtensorflow_jni/src libtensorflow_jni/target libtensorflow_jni_gpu/src libtensorflow_jni_gpu/target \
+    libtensorflow/src libtensorflow/target tensorflow-android/target proto/src proto/target \
+    hadoop/src hadoop/target spark-connector/src spark-connector/target
 }
 
 update_version_in_pom() {
@@ -183,6 +186,43 @@ generate_java_protos() {
   rm -rf "${DIR}/proto/tmp"
 }
 
+
+# Download the TensorFlow ecosystem source from git.
+# The pom files from this repo do not inherit from the parent pom so the maven version
+# is updated for each module.
+download_tf_ecosystem() {
+  ECOSYSTEM_DIR="/tmp/tensorflow-ecosystem"
+  HADOOP_DIR="${DIR}/hadoop"
+  SPARK_DIR="${DIR}/spark-connector"
+
+  # Clean any previous attempts
+  rm -rf "${ECOSYSTEM_DIR}"
+
+  # Clone the TensorFlow ecosystem project
+  mkdir -p  "${ECOSYSTEM_DIR}"
+  cd "${ECOSYSTEM_DIR}"
+  git clone "${TF_ECOSYSTEM_URL}"
+  cd ecosystem
+  git checkout r${TF_VERSION}
+
+  # Copy the TensorFlow Hadoop source
+  cp -r "${ECOSYSTEM_DIR}/ecosystem/hadoop/src" "${HADOOP_DIR}"
+  cp "${ECOSYSTEM_DIR}/ecosystem/hadoop/pom.xml" "${HADOOP_DIR}"
+  cd "${HADOOP_DIR}"
+  update_version_in_pom
+
+  # Copy the TensorFlow Spark connector source
+  cp -r "${ECOSYSTEM_DIR}/ecosystem/spark/spark-tensorflow-connector/src" "${SPARK_DIR}"
+  cp "${ECOSYSTEM_DIR}/ecosystem/spark/spark-tensorflow-connector/pom.xml" "${SPARK_DIR}"
+  cd "${SPARK_DIR}"
+  update_version_in_pom
+
+  # Cleanup
+  rm -rf "${ECOSYSTEM_DIR}"
+
+  cd "${DIR}"
+}
+
 # Deploy artifacts using a specific profile.
 # Arguments:
 #   profile - name of selected profile.
@@ -240,7 +280,8 @@ cd "${DIR}"
 # Comment lines out appropriately if debugging/tinkering with the release
 # process.
 # gnupg2 is required for signing
-apt-get -qq update && apt-get -qqq install -y gnupg2
+apt-get -qq update && apt-get -qqq install -y gnupg2 git
+
 clean
 update_version_in_pom
 download_libtensorflow
@@ -248,6 +289,8 @@ download_libtensorflow_jni
 download_libtensorflow_jni_gpu
 update_tensorflow_android
 generate_java_protos
+download_tf_ecosystem
+
 # Build the release artifacts
 mvn verify
 # Push artifacts to repository
diff --git a/tensorflow/java/maven/spark-connector/pom.xml b/tensorflow/java/maven/spark-connector/pom.xml
new file mode 100644
index 0000000000..19c752d08b
--- /dev/null
+++ b/tensorflow/java/maven/spark-connector/pom.xml
@@ -0,0 +1,24 @@
+<project
+        xmlns="http://maven.apache.org/POM/4.0.0"
+        xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
+        xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
+    <!-- Placeholder pom which is replaced by TensorFlow ecosystem Spark pom during build -->
+    <modelVersion>4.0.0</modelVersion>
+    <description>TensorFlow TFRecord connector for Apache Spark DataFrames</description>
+    <artifactId>spark-connector</artifactId>
+    <packaging>jar</packaging>
+
+    <scm>
+        <url>https://github.com/tensorflow/ecosystem.git</url>
+        <connection>git@github.com:tensorflow/ecosystem.git</connection>
+        <developerConnection>scm:git:https://github.com/tensorflow/ecosystem.git</developerConnection>
+    </scm>
+
+    <url>https://github.com/tensorflow/ecosystem/</url>
+    <parent>
+        <groupId>org.tensorflow</groupId>
+        <artifactId>parentpom</artifactId>
+        <version>1.9.0-rc0</version>
+        <relativePath>../</relativePath>
+    </parent>
+</project>
\ No newline at end of file