Commits on Apr 21, 2009
  1. RVH should use RC.getRow for correctness and maintainability (in part…

    …icular the path for get_columns_since was missing). patch by Jun Rao; reviewed by jbellis for #93
    Jonathan Ellis committed Apr 21, 2009
  2. raise exception when column is not passed to get_column. patch by Jun…

    … Rao; reviewed by jbellis for #92
    Jonathan Ellis committed Apr 21, 2009
  3. configurable ListenAddress setting. allows running multiple instances…

    … of Cassandra
    
    on a single machine or VM.  patch and review by Per Mellqvist, Eric Evans, and
    jbellis for #43
    Jonathan Ellis committed Apr 21, 2009
  4. Revert "configurable ListenAddress setting" since it prevents server …

    …from starting.
    
    This reverts commit 1304a9d9e1f9f85b991d99a1ca96e02505eb430f.
    Jonathan Ellis committed Apr 21, 2009
  5. configurable ListenAddress setting. allows running multiple instances…

    … of Cassandra on a single machine or VM. patch by Per Mellqvist; reviewed by Eric Evans for #43
    Jonathan Ellis committed Apr 21, 2009
  6. fix strange line endings and indentation so we can apply patches sanely.

    patch by Per Mellqvist; reviewed by Eric Evans for #43
    Jonathan Ellis committed Apr 21, 2009
  7. allow SuperColumn.getSubColumn to return null. patch by Sandeep Tata;…

    … reviewed by jbellis for #91
    Jonathan Ellis committed Apr 21, 2009
  8. test and fix get_column regression. patch by Jun Rao; reviewed by jbe…

    …llis for #90
    Jonathan Ellis committed Apr 21, 2009
Commits on Apr 20, 2009
  1. refactor read path: first we make readColumnFamily accept a ReadComma…

    …nd, and use that to clean out duplicate code in CassandraServer. Then we clean up the duplicate versions of the read methods in StorageService by making them ReadCommand-based, too. [not touching multiget code for now.]
    
    patch by jbellis; reviewed by Eric Evans for #88
    Jonathan Ellis committed Apr 20, 2009
  2. rename ReadResponseMessage -> ReadResponse

    patch by jbellis; reviewed by Eric Evans for #88
    Jonathan Ellis committed Apr 20, 2009
  3. clean up ReadCommand; it's basically a struct so use public final fields

    patch by jbellis; reviewed by Eric Evans for #88
    Jonathan Ellis committed Apr 20, 2009
  4. rename get_cf -> readColumnFamily; ReadMessage -> ReadCommand.

    [Message message = ReadMessage.readMessage(readMessage) is just plain confusing]
    patch by jbellis; reviewed by Eric Evans for #88
    Jonathan Ellis committed Apr 20, 2009
  5. fix row size "returned" via bufOut (currently ignored). patch by Jun …

    …Rao; reviewed by jbellis for #33
    Jonathan Ellis committed Apr 20, 2009
Commits on Apr 18, 2009
  1. fix SequenceFile to know about new int in CF serialization. patch by …

    …Jun Rao; reviewed by jbellis for #33
    Jonathan Ellis committed Apr 18, 2009
Commits on Apr 17, 2009
  1. refactor Filter heirarchy, making hash generation easily customizable…

    …. Use Murmur
    
    hash + combinatorics to generate hashes.
    Jonathan Ellis committed Apr 17, 2009
  2. rename BloomFilter.fill -> add

    Jonathan Ellis committed Apr 17, 2009
  3. replace JenkinsHash w/ MurmurHash. its hash distribution is just as g…

    …ood, and it's faster
    Jonathan Ellis committed Apr 17, 2009
  4. r/m unused code, including entire CountingBloomFilter

    Jonathan Ellis committed Apr 17, 2009
  5. change column_t value to binary from string. patch by johano; reviewe…

    …d by jbellis for #29
    Jonathan Ellis committed Apr 17, 2009
  6. test and fix for CF clone bug during remove. patch by Jun Rao; review…

    …ed by jbellis for #33
    Jonathan Ellis committed Apr 17, 2009
  7. make GC_GRACE_IN_SECONDS customizable in storage.conf.

    patch by jbellis; reviewed by Eric Evans for #33
    Jonathan Ellis committed Apr 17, 2009
  8. omit tombstones from column_t and supercolumn_t returned to client.

    patch by jbellis; reviewed by Eric Evans for #33
    Jonathan Ellis committed Apr 17, 2009
  9. preserve tombstones until a GC grace period has elapsed.

    patch by jbellis; reviewed by Eric Evans for #33
    Jonathan Ellis committed Apr 17, 2009
  10. waitForFlush -> forceBlockingFlush. ServerTest.cleanup now flushes an…

    …d cleans out
    
    all ColumnFamilyStores and commitlog, allowing remove tests to not step on each
    others' toes (all tests pass now).
    
    patch by jbellis; reviewed by Sandeep Tata for #85
    Jonathan Ellis committed Apr 17, 2009
  11. use isSuper instead of DatabaseDescriptor.getColumnType.

    patch by jbellis; reviewed by Sandeep Tata for #85
    Jonathan Ellis committed Apr 17, 2009
  12. refactor CFS constructor to not depend on DatabaseDescriptor.

    patch by jbellis; reviewed by Sandeep Tata for #85
    Jonathan Ellis committed Apr 17, 2009
  13. add testRemoveColumnFamily, testRemoveSubColumn. (these do not pass y…

    …et b/c of
    
    cleanup problems.)  patch by jbellis; reviewed by Sandeep Tata for #85
    Jonathan Ellis committed Apr 17, 2009
Commits on Apr 16, 2009
  1. make forceFlush block until the flush action is queued on MemtableMan…

    …ager. That
    
    way calling forceFlush; waitForFlush will be guaranteed that the action waitFF puts
    on MtM will run after the flush completes, i.e., the wait will actually do what
    it's supposed to.
    
    patch by jbellis; reviewed by Eric Evans for #59
    Jonathan Ellis committed Apr 16, 2009
  2. r/m libfb303 as a dependency. (this completes the process begun by Pr…

    …ashant earlier.) patch by Jun Rao; reviewed by jbellis for #86
    Jonathan Ellis committed Apr 16, 2009
  3. expose redone remove to thrift. patch by jbellis; reviewed by Jun Rao…

    … for #83
    Jonathan Ellis committed Apr 16, 2009
  4. generate supercolumn tombstone when a 2-tuple delete is passed to a s…

    …uper CF.
    
    patch by jbellis; reviewed by Sandeep Tata for #84
    Jonathan Ellis committed Apr 16, 2009
  5. make remove test code use the same api that the thrift interface does…

    …. patch by
    
    jbellis; reviewed by Sandeep Tata for #84
    Jonathan Ellis committed Apr 16, 2009
Commits on Apr 15, 2009
  1. Move from ExecutorService per ColumnFamily to ES per Memtable. This a…

    …llows us to
    
    wait for the ES to quiesce completely before flushing, preventing the possibility
    of ConcurrentModificationException when a get scheduled before the switch executes
    concurrently with flush.  It also provides a simpler mental model (only one thread
    touches memtable at a time, period) which is a valuable property.  Finally, it is
    slightly more performant since it avoids hashing the CF name for each operation.
    
    Patch by jbellis; reviewed by Todd Lipcon for #9
    Jonathan Ellis committed Apr 15, 2009
  2. Add MemtableSizeInMB and MemtableObjectCountInMillions to the default…

    … config. Uses a more conservative value for MemtableSizeInMB the default. Patch by Eric Evans; reviewed by jbellis for #51
    Jonathan Ellis committed Apr 15, 2009
  3. upgrade to thrift svn r763981. for Java, this should be identical to …

    …thrift 0.1. patch by jbellis; reviewed by Eric Evans for #72
    Jonathan Ellis committed Apr 15, 2009