ORC-477: BloomFilter for ACID table does not get created

[deniskuzZ]

No description provided.

[t3rmin4t0r]

This code should ideally be in the Hive codebase and ORC should interpret the selectedColumns there.

[deniskuzZ]

ACID schema generation code is under ORC SchemaEvolution class

public static TypeDescription createEventSchema(TypeDescription typeDescr) {
TypeDescription result = TypeDescription.createStruct()
.addField("operation", TypeDescription.createInt())
.addField("originalTransaction", TypeDescription.createLong())
.addField("bucket", TypeDescription.createInt())
.addField("rowId", TypeDescription.createLong())
.addField("currentTransaction", TypeDescription.createLong())
.addField("row", typeDescr.clone());
return result;
}

[deniskuzZ]

I am thinking on simplifying this a bit by introducing extra param (isAcid) in findColumn method.

if (resultTypeDesc.getCategory() == TypeDescription.Category.STRUCT && isAcid) {
//unfold
resultTypeDesc = findColumn(columnName, resultTypeDesc.getFieldNames(), resultTypeDesc.getChildren(), false);
}

[t3rmin4t0r]

This should already work if Hive sends "row." in selected columns?

[deniskuzZ]

not completely, column decr are embedded under row struct, so we need to unfold it.

[deniskuzZ]

@t3rmin4t0r do you know how to configure the AppVeyor build?

The build phase is set to "MSBuild" mode (default), but no Visual Studio project or solution files were found in the root directory. If you are not building Visual Studio project switch build mode to "Script" and provide your custom build command.