sedona-spark-shaded package should exclude some repetitive dependencies.

[freamdx]

pom.xml should exclude any dependencies that exist in spark jars, eg: edu.ucar:cdm-core exclude guava/httpclient/protobuf-java ... ... <dependency> <groupId>edu.ucar</groupId> <artifactId>cdm-core</artifactId> <exclusions> <exclusion> <groupId>com.google.guava</groupId> <artifactId>guava</artifactId> </exclusion> <exclusion> <groupId>org.apache.httpcomponents</groupId> <artifactId>httpclient</artifactId> </exclusion> <exclusion> <groupId>com.google.protobuf</groupId> <artifactId>protobuf-java</artifactId> </exclusion> </exclusions> </dependency> ... ... <artifactSet> <excludes> <exclude>org.scala-lang:scala-library</exclude> <exclude>org.apache.commons:commons-*</exclude> <exclude>commons-pool:commons-pool</exclude> <exclude>commons-lang:commons-lang</exclude> <exclude>commons-io:commons-io</exclude> <exclude>commons-logging:commons-logging</exclude> </excludes> </artifactSet> ... ...

[jiayuasu]

@freamdx would you mind create a PR to fix this yourself?

[Kontinuation]

A few notes:

• cdm-core has to be bundled in the shaded package, we include it in the shaded pom.xml to make it compile rather than provided.
• guava is better shaded than excluded, since we don't know whether the guava jar shipped with spark is compatible with sedona or not, and shading guava is a common practice.
• other package exclusion seems to be OK. Apache commons has good backward compatibility and the versions shipped with spark is later than what we bundled.

[zwu-net]

I agree with @Kontinuation. Guava is pretty tricky, as I learned it when upgrading Spark.

I'm having trouble understanding why this Google's library has such poor compatibility issues. It seems unexpected from a leading tech company

I'm willing to step in if @jiayuasu approves, given @freamdx's lack of response. However, I some worried proceeding without rigorous testing to mitigate potential risks.

[jiayuasu]

@zwu-net please feel free to create a PR to fix this

[zwu-net]

@jiayuasu (I would like also to bring @james-willis here since I got know him due to DBScan discussion. If you are busy on other tasks, you don't need to participate) @Kontinuation @freamdx

Research Findings and Proposal

After researching the feasibility of modifying Sedona's pom.xml (https://github.com/apache/sedona/blob/master/spark-shaded/pom.xml), I concluded that manual management of library inclusions/exclusions across various Spark and Sedona versions would be overly complex and labor-intensive.

Proposed Solution

To address this challenge, I suggest creating a custom tool to manage pom.xml generation. This approach is inspired by my previous work on a custom transformer of SQLs from Oracle to Snowflake(https://www.linkedin.com/pulse/revolutionizing-data-analysis-custom-transformer-seamless-paul-wu-4euxc/?trackingId=1AyXqiJh1sei4yybC%2FZW9g%3D%3D).

Tool Requirements

The tool should:

1. Fetch Spark releases (ideally from a prefetched repository to minimize build time and network stability issues)
2. Generate pom.xml files
3. Be implemented in Java or Python (with careful dependency management if using Python)

Implementation Considerations

Given the scope of this task, it may take several months for part-time contributors like myself to implement and test. Before proceeding, I'd appreciate feedback on:

1. The viability of this approach
2. Potential alternative solutions

Please share your thoughts.

[freamdx]

1. only NetCDF using ucar package
2. s2-geometry package depends guava:jar:25.1-jre ( mvn dependency:tree ) as default
3. mvn test passed

[freamdx]

@freamdx would you mind create a PR to fix this yourself?

how to get a ticket ID ?

[jiayuasu]

@freamdx Please sign up for an account here: https://issues.apache.org/jira/projects/SEDONA