[HBASE-8074] Consolidate map-side features across mapreduce tools into a single place

• Type: Sub-task
• Status: Closed
• Priority: Major
• Resolution: Won't Fix
• Component/s: mapreduce, Usability
• Assignee: Unassigned
• Reporter: Nick Dimiduk
• Created: 12/Mar/13 01:06
• Updated: 16/Jun/22 16:38
• Resolved: 16/Jun/22 16:38

Description

The mapreduce tools support a similar but divergent set of features for mapping over KeyValue data:

• Export supports specifying a version-range window, application of a rowkey regex or prefix filter, and a "raw mode" that includes delete markers.
• Import can apply an arbitrary filter and can also apply a "transform", renaming column families in the emitted KeyValues.
• CopyTable allows specifying a version-range window, limiting to a fixed number of versions, a "raw mode", and column family transformation.
• WALPlayer supports reading a time-range.
• ImportTsv could incorporate a number of these features, especially the filter and transform capabilities, allowing a user to avoid implementing a custom mapper where the existing parser is sufficient, but for a slight massage of the data.

The proposal is to create a single implementation for these features with a single configuration interface. Ideally, such an implementation would be exposed via the common utility classes as well (ie, IdentityTableMapper).

Issue Links

• depends upon: HBASE-8094 TestTableInputFormatScan doesn't assert anything
• is related to: HBASE-13787 use a cli argument parsing library instead of ad-hoc string handling