HACKING.md

Coding conventions

• For files underneath bin:

  • Use - as separator for filenames (like my-nice-cool-script).

  • Do not use extensions for the used language (so no trailing .sh, and no trailing .py)

  • At least for programs that “act on” refinery, filenames should start in refinery-.

• For files outside of bin:

  • Use _ as separator for filenames (like my_nice_file.txt).

  • Do use proper extensions there. (like trailing .xml, or trailing .properties)

• We do not like camelCase in variable names. Only use them in Java code, where they are somewhat standard, and in Python class names.

• Filenames for files in HiveQL should end in .hql.

• When referencing tables, reference them using database.table wherever possible.

• For Hadoop jobs, use the “adhoc” queue per default, and make sure that it can be overridden to submit the job to queues with less limitations.

• For Oozie jobs, set oozie.use.system.libpath, and oozie.action.external.stats.write to true in the corresponding properties files. (This setting will get added as default at some point, but until then, we rely on it being set in the properties files)

• When needing to reference hdfs, try using hdfs:///. Where this does not work, try using hdfs://analytics-hadoop/, then hdfs://namenode.analytics.eqiad.wmnet/ and finally hdfs://namenode.analytics.eqiad.wmnet:8020/.

• Use 4 spaces to indent files.

• There is no rule on vertical alignment.

• In HiveQL, we prefer != over <>.

• In XML, put the beginning and ending tag on separate lines for multi-lined element values. So use

  <tag>
    foo
    bar
  </tag>
  

  instead of

  <tag>foo
  bar</tag>
  

  .

• For Oozie jobs, have the name end in -wf for workflows, -coord for coordinators, and -bundle for bundles.

• If you add Oozie jobs, update the diagrams/oozie-overview.dia.