[CASSANDRA-4804] Wrong assumption for KeyRange about range.end_token in get_range_slices().

• Type: Bug
• Status: Resolved
• Priority: Low
• Resolution: Fixed
• Fix Version/s: 1.1.7, 1.2.0 beta 2
• Component/s: Legacy/CQL
• Severity: Low
• Assignee: Nikolay
• Reporter: Nikolay

Authors

Nikolay

Reviewers

David Brosius

• Created: 14/Oct/12 15:45
• Updated: 16/Apr/19 09:32
• Resolved: 17/Oct/12 04:01

Description

In get_range_slices() there is parameter KeyRange range.

There you can pass start_key - end_key, start_token - end_token, or start_key - end_token.

This is described in the documentation.

in thrift/ThriftValidation.java there is validation function validateKeyRange() (line:489) that validates correctly the KeyRange, including the case start_key - end_token.

However in thrift/CassandraServer.java in function get_range_slices() on line: 686 wrong assumption is made:

if (range.start_key == null)

{ ... // populate tokens }

else

{ bounds = new Bounds<RowPosition>(RowPosition.forKey(range.start_key, p), RowPosition.forKey(range.end_key, p)); }

This means if there is start key, no end token is checked.
The opposite - null is "inserted" as end_key.

Solution:
same file - thrift/CassandraServer.java on next function - get_paged_slice(), on line:741 same code is written correctly

if (range.start_key == null)

{ ... // populate tokens }

else

{ RowPosition end = range.end_key == null ? p.getTokenFactory().fromString(range.end_token).maxKeyBound(p) : RowPosition.forKey(range.end_key, p); bounds = new Bounds<RowPosition>(RowPosition.forKey(range.start_key, p), end); }