自从阿里云推出函数计算（FC）以来，一直期待着能用Java语言实现FC，今年早些时候发现已经放出了Java FC的实现，这样我们的DICOM预处理就可以做成函数了。这周终于空闲下来，决定来踩坑。
按照阿里云Java教程很容易写了一个DemoFC，不过在用fcli发布函数的时候该教程的命令是不太对的，mkf 命令应该使用参数 -f YourPackage.jar 来指定要上传的jar包。完整的命令如下：
mkf demo-java -t java8 -h com.lifeccp.fc.dcmpp.DemoFC::handleRequest -f fc-dcmpp-1.0.0.jar
接下来就开始将dcm4che引入项目了。
首先遇到的第一个坑是maven不能下载dcm4che。
Java and Let's Encrypt Certificates
DCM4CHE自己有一个maven repository，启用了HTTPS，并且使用的是Let's Encrypt的证书（全世界码农的选择都一样）。JDK8U101之前的版本没有装入Let's Encrypt的根证书，导致maven下载依赖包的时候出错：
Failed to collect dependencies at org.dcm4che:dcm4che-core:jar:5.14.0: Failed to read artifact descriptor for org.dcm4che:dcm4che-core:jar:5.14.0: Could not transfer artifact org.dcm4che:dcm4che-core:pom:5.14.0 from/to www.dcm4che.org (https://www.dcm4che.org/maven2/): sun.security.validator.ValidatorException: PKIX path building failed: sun.security.provider.certpath.SunCertPathBuilderException: unable to find valid certification path to requested target
最简单的解决办法：升级JDK到8U101以后版本。
参考这篇blog： https://www.compose.com/articles/compose-notes-java-and-lets-encrypt-certificates/
OpenCV
Java底下一直没有很好的JPEGLS/JPEG2K支持，之前dcm4che用的是一个很久不更新的jai_imageio(版本1.2-pre-dr-b04，对，一直停留在pre-b04版本号上)，而且底层还依赖clipwrap的jni，这个native依赖常常让预处理器崩溃。dcm4che 5.0以后引入了OpenCV这个强大的图像库来提供图像IO，我们在生产环境上升级dcm4che 5.0以后，处理JPG编码的DICOM没有再奔溃了。不过OpenCV还是native依赖的，不知道阿里云的FC环境是否能够支持呢？
先试试再说，加上opencv依赖：
<dependency>
<groupId>org.dcm4che</groupId>
<artifactId>dcm4che-imageio-opencv</artifactId>
<version>5.14.0</version>
</dependency>
构建过程中并没有自动包native包拉进来，还得将需要的native包依赖上才行。虽然构建后发现已经将libopencv_java-3.4.0-dcmR3-linux-x86-64.so 拷贝到lib了，但还是找不到opencv_java的本地依赖包。上面的方法拷贝的.so文件名并不是OpenCV core去加载的文件名，还得把名称改成libopencv_java.so (在mac则是libopencv_java.jnilib）才行。使用maven-dependenciey-plugin:copy 来完成。
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-dependency-plugin</artifactId>
<version>3.1.1</version>
<executions>
<execution>
<id>copy-dependencies</id>
<phase>prepare-package</phase>
<goals>
<goal>copy-dependencies</goal>
</goals>
<configuration>
<outputDirectory>${project.build.directory}/classes/lib</outputDirectory>
<includeScope>runtime</includeScope>
</configuration>
</execution>
<!-- copy libopencv_java.so for linux-x86-64 for distribution and testing,
run package goal once before test on linux, otherwise vm can't find native library -->
<execution>
<id>copy</id>
<phase>prepare-package</phase>
<goals>
<goal>copy</goal>
</goals>
<configuration>
<artifactItems>
<artifactItem>
<groupId>org.weasis.thirdparty.org.opencv</groupId>
<artifactId>libopencv_java</artifactId>
<type>so</type>
<version>${weasis.opencv.native.version}</version>
<classifier>linux-x86-64</classifier> <outputDirectory>${project.build.directory}/classes/lib</outputDirectory>
<destFileName>libopencv_java.so</destFileName>
</artifactItem>
</artifactItems>
</configuration>
</execution>
<!-- copy libopencv_java.jnilib for macosx-x86-64 for testing under my Mac -->
<execution>
<id>copy-mac</id>
<phase>process-test-classes</phase>
<goals>
<goal>copy</goal>
</goals>
<configuration>
<artifactItems>
<artifactItem>
<groupId>org.weasis.thirdparty.org.opencv</groupId>
<artifactId>libopencv_java</artifactId>
<type>jnilib</type>
<version>${weasis.opencv.native.version}</version>
<classifier>macosx-x86-64</classifier>
<outputDirectory>${project.build.directory}/test-classes/lib</outputDirectory>
<destFileName>libopencv_java.jnilib</destFileName>
</artifactItem>
</artifactItems>
</configuration>
</execution>
</executions>
</plugin>
同时需要配置maven-surefire-plugin ，加上<argLine>-Djava.library.path=./target/test-classes/lib:./target/classes/lib</argLine>，不然test的时候还是找不到本地依赖，mvn -Djava.library.path=./target/test-classes/lib 指定都不管用，因为test启动的是另外的vm。
<!-- specify java.library.path for testing -->
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-surefire-plugin</artifactId>
<version>2.22.0</version>
<configuration>
<argLine>-Djava.library.path=./target/test-classes/lib:./target/classes/lib</argLine>
</configuration>
</plugin>
接下来就是激动人心的上FC真实环境测试了。按照上面pom的配置，libopencv_java.so 拷贝到lib目录并且打包到jar以后，vm能够成功加载native包。FC上的Java runtime：
user.dir:/code
java.library.path:/code/:/code//lib:/usr/java/packages/lib/amd64:/usr/lib/x86_64-linux-gnu/jni:/lib/x86_64-linux-gnu:/usr/lib/x86_64-linux-gnu:/usr/lib/jni:/lib:/usr/lib
>fcli shell
Welcome to the function compute world. Have fun!
>>> cd deidentifier
>>> ls
dcminfo
demo-java
>>> invk dcminfo -s samples/CT1_J2KR
loading native library:opencv_java...
native library loaded.
samples/CT1_J2KR 512x512px
color model:ColorModel: #pixelBits = 8 numComponents = 1 color space = java.awt.color.ICC_ColorSpace@12ab536b transparency = 1 has alpha = false isAlphaPre = false
Job done!
