FAQ

← Previous

What license is KijiSchema under?

KijiSchema is under the Apache License, Version 2.0

How do I add Kiji and its dependencies to my application's classpath?

Like hadoop, the kiji binary supports a jar operation. If your program is packaged as a single jar and you want to run it with Kiji on the classpath, you can run:

$ kiji jar myapp.jar com.myorg.MyApp [args...]

If your application constructs a more complex classpath, then you can also add Kiji to your application's classpath explicitly. Like hadoop and hbase, the kiji binary has a classpath command which will print a string that you can use directly in your $CLASSPATH environment variable or on the java command line as the -cp option. For example:

$ java -cp myapp.jar:`kiji classpath` com.myorg.MyApp

I'm confident that I'm writing data to my Kiji table correctly. Why don't I see the new cells when I try to read them back?

Check the retention policy for the cells in the locality group you are writing to. The locality groups of your Kiji table are configured in your table's layout. See Managing Data for details.

In particular, check to make sure the timestamps of the new cells you are writing are in milliseconds, and that they fall within the bounds of your locality group's ttl_seconds setting.

How do I specify the schema of a column in my Kiji table?

Kiji uses Avro for data type definitions and serialization within cells of Kiji tables. Avro can define simple types like "int" and "string" as well as complex record types, composed of scalars or other records, arrays, maps, etc. Every column (or map-type family) in Kiji has a type (or «schema» in Avro parlance) associated with it.

For primitives types like, eg. "int", use the following in your column layout descriptor:

column_schema: {type: "INLINE", value = '"int"'}

For complex record types, you may instead specify a generated record class:

column_schema: {type: "CLASS", value = "org.package.GeneratedRecordClass"}

You can write Avro schemas .avsc or .avdl files. These can be compiled into Java classes that represent the record type. This makes working with these objects easier in your mapper and reducers. For complete information on Avro, visit http://avro.apache.org.

See Managing Data for more details on Kiji cell schemas.

How does data in Kiji map to Java types?

Kiji uses Avro for data type management and serialization. Each cell of typed data has an associated Avro schema. When reading or writing data from Kiji, the Java type must be compatible with the Avro schema. For example, if a Kiji column has the Avro schema "string", data read from cells of the column will implement the Java CharSequence interface. For record schemas that have been compiled into Java classes using the Avro compiler, the Java type corresponding to the schema is the compiled class.

The full list of mappings between Avro schemas and Java types are documented in the Avro Java API documentation. See the package documentation for org.apache.avro.specific and org.apache.avro.generic.

Are there restrictions on the characters I use in table names, family names, and column names?

Yes. All table names, column family names, and column qualifier names may only contain numbers, letters and underscores ([A-Za-z_][A-Za-z0-9_]*). However, the qualifiers of a map-type column family may contain any printable characters.

← Previous

FAQ