add "required" to non-optional fields. Thrift sucks.

Fixes to make BinaryMemtable useful. Highlights are configurable threads for [binary]memtable flushing and flushAndShutdown JMX/nodeprobe directive.

call doAntiCompaction in the tests the same way we do in the actual code -- by submitting it to the MinorCompactionManager so it serializes with other compaction tasks. patch by Sandeep Tata; reviewed by jbellis for CASSANDRA-395

fix typo reported by Arin.

bump thrift version to 808609 to fix regression from THRIFT-562. patch by jbellis for CASSANDRA-387

remove multiget_count until/unless there is a demonstrated need for it. patch by jbellis for CASSANDRA-70

add apache license boilerplate

Update AsyncResult to know when it started waiting so you can get() multiple Results in serial but have them wait in parallel. Use this to implement multiget methods. Remove unused MultiAsyncResult code and StorageProxy methods using it.

update to thrift trunk and rename jar after the svn revision (806014). inline our map typedef to work around regression introduced in THRIFT-144. add slf4j dependencies (required since THRIFT-558). regenerate thrift structs with new version.

fix getUsedDiskSpace to look at data directories recursively. Patch by Sammy Yu; reviewed by jbellis for CASSANDRA-394

better fix for COPP. patch by jbellis

fix unit test. patch by jbellis for CASSANDRA-242

refactors COPP to use BytesToken. patch by Stu Hood; reviewed by jbellis for CASSANDRA-242

1. Switch bootstrapNodes in TokenMetadata to Map<Token,EndPoint> so RackUnawareStrategy can use it.

add more (debug) logging to log recovery process. patch by jbellis

combine batch_insert and batch_insert_super the way we did get_slice and get_slice_super

[fixed version] avoid opening multiple writers for the current file; the buffered nature of the global logWriter_ could cause problems

need to include column container's deletion status when determining whether to include a column in the live count.

Revert "CASSANDRA-370 avoid opening multiple writers for the current file; the buffered nature of the global logWriter_ could cause problems"

read(bytes) doesn't automatically throw EOFException if it reads less than asked for, so we need to check for that.

CASSANDRA-370 avoid opening multiple writers for the current file; the buffered nature of the global logWriter_ could cause problems

updated log file location

allow blocking write to create hints if not enough of the "correct" nodes are live, but enough are to fulfil the requested consistency level. patch by Sandeep Tata; reviewed by jbellis and Michael Greene for CASSANDRA-383

change LongType to read longs in big endian order to be consistent with network order and the UUID types.

CASSANDRA-373 re{format,factor} conf/storage-conf.xml

finish removing Groovy. patch by Michael Greene; reviewed by jbellis for CASSANDRA-380

remove groovy jar, half-baked Callout code. patch by Michael Greene; reviewed by jbellis for CASSANDRA-380

mention snapshot support in changelog

set miscellaneous svn properties (mime-type, eol-style LF on .sh). patch by Sebb; reviewed by Michael Greene for CASSANDRA-378

svn ps svn:eol-style native *. patch by Sebb; reviewed by Michael Greene for CASSANDRA-378

sanity check start, finish slice args

remove authorship comments. patch by Michael Greene; reviewed by jbellis for CASSANDRA-379

rename getInitialToken to getToken; it doesn't actually have anything to do w/ initial or default tokens. log token on startup.

rename underscores away

Add "bootstrap mode" to node startup. This causes the node to tell the

use the subcomparator when toStringing slice commands on subcolumns. this exposes a couple bugs: fix getString in non-string types to accept byte[0], and fix a test to send a long to a LongType subcolumn. patch by jbellis and Evan Weaver for CASSANDRA-377

fix typo setting syncperiod. patch by jbellis

include index and bloom filter in snapshots. patch by Sammy Yu; reviewed by jbellis for CASSANDRA-353

human-readable error for bad CommitLogSyncDelay.

tests + fix for bsearch returning not-found-to-the-right

inline excessively abstracted reTest setup

compact slicelarge row to ensure we are in fact testing with more than the minimum index count

final cleanup of SSTableSliceIterator, now with less unnecessary copying of Columns around

avoid copying variables to ColumnGroupReader that it can get from the parent class. rename underscored variables.

move comparator out of IndexInfo

always write at least one index (with first and last column of the range) for the columns in row. this vastly simplifies column reading code and makes indexing bugs much more obvious (since there is only one read path each for names / slices now).

don't serialize unused column count into column index. remove DataInput/Output round-tripping from ColumnGroupReader

add test for multi-block reversal.

have bin/cassandra startup script respect JAVA_HOME

fix range query buglet; add debug logging

Improve the speed of RandomPartitioner comparator. patch by Sammy Yu; reviewed by jbellis for CASSANDRA-346

fix more findbugs complaints. patch by Michael Greene; reviewed by jbellis for CASSANDRA-338

fix SC.serializedSize. patch by jbellis; reviewed by Chris Goffinet for CASSANDRA-341

fix NPE on iterator close. patch by jbellis; reviewed by Chris Goffinet for CASSANDRA-341

fix .properties line endings. patch by jbellis; reviewed by Jeff Hodges for CASSANDRA-111

The change to LexicalUUIDType and TimeUUIDType didn't get propagated to the default storage-conf. patch by Michael Greene; reviewed by jbellis for CASSANDRA-240

Added LZOP fix for files with 1 split

Bootstrap code + tests. Patch by Sandeep Tata; reviewed by jbellis for CASSANDRA-213

brings findbugs count from 266 -> 250 bugs by fixing some scopes, removing some dead code, and not implementing Serializable unnecessarily. patch by Michael Greene; reviewed for CASSANDRA-338 by jbellis

forgot SSTable part of last.

renames and cleanup for bootstrap-related code.

remove AbstractWriter. all it was providing was append methods, which are inlined into SSTableWriter now; the rest was all wrappers. also, rename some variables in SST/SSTW.

r/m SequenceFile. ColumnGroupReader moved to SSTableSliceIterator mostly unchanged for now. (finish cleaning this up in #332)

replace SF.next with code in NamesIterator that doesn't rewrite data through an extra layer of DataOutput/Input

SF shouldn't duplicate position checking that was already done by SSTable. move utility methods to IndexHelper. fix off-by-4 in dataSizeReturned.

r/m code in SF.next handling being called multiple times, since it can only be called once.

r/m special case that no column names=all columns; add test for no-columns cases.

r/m SF.next(DataOutput) in favor of more-efficient IteratingRow.

move next, getColumnGroupReader out of SSTableReader.

make CommitLog use a raw BufferedRandomAccessFile instead of overloading AbstractWriter/IFileReader unnecessarily. the set of AW features used by both CommitLog and SSTable was nil.

Make range queries continue on the next node when necessary.

allow start of [] to mean "start with the largest value" when ascending=false. patch by Jun Rao; reviewed by jbellis for CASSANDRA-263

run source files through dos2unix to standardize on \n line endings. patch by Jeff Hodges; reviewed by Michael Greene for CASSANDRA-111

table -> keyspace in stress.py. patch by Michael Greene; reviewed by jbellis for CASSANDRA-271

table -> keyspace

Revert "Pluggable replicaplacement, endpointsnitch classes." Breaks test suite.

Pluggable replicaplacement, endpointsnitch classes. patch by Sammy Yu; reviewed by jbellis for CASSANDRA-323

Add missing changelog entries from trunk/0.4. patch by jbellis for CASSANDRA-321

add isTraceEnabled checks similar to isDebugEnabled. patch by jbellis for CASSANDRA-174

add more if-isDebugEnabled lines to reduce overhead when logging at INFO or higher. patch by Michael Greene; reviewed by jbellis for CASSANDRA-174

combine switchMemtable and enqueueFlush so that they always operate on the same object. move forceFlush guts from MT to CFS since it encapsulates better there

Use arrayblockingqueue in commitlog executor; this cleans up the code a bit (performance is unaffected since the writes and syncs are far more expensive than any queue ops)

add config options for commitlog syncing

custom CommitLogExecutorService that can fsync per multiple CL additions

move log ops to callables on a threadpoolexecutor instead of synchronizing. this prepares the way to merge multiple add() calls into a single sync.

naive fsync-after-each-log-entry

mv AbstractWriter to its own top-level class and remove redundant IFileWriter

handle incomplete CL entries on recover

update stress.py for current thrift api

use CSLS instead of NBHS for memtablesPendingFlush. See explanation here: https://sourceforge.net/tracker/?func=detail&aid=2828100&group_id=194172&atid=948362

add ConsistencyLevel Thrift enum to avoid client having to compute quorum size. patch by Sammy Yu; reviewed by jbellis for CASSANDRA-300

fix bug in comparing-to-empty-array; realize that we need compare-to-empty-array to specify start/finish of slices.

finish snapshot support. patch by Sammy Yu; reviewed by jbellis and Michael Greene for CASSANDRA-279

people are still having trouble with GCs in small heap sizes. flush more often with small columns. patch by jbellis

close temporary logWriters to avoid leaking FD. patch by Brandon Williams; reviewed by jbellis for CASSANDRA-313

r/m unwanted logger.debug, pointed out by Michael Greene. patch by jbellis

add perl namespace to thrift interface

better handling of file names with spaces on windows

add BytesType comparator (like AsciiType but doesn't require that the bytes mean anything)

rename DD.getType -> getComparator

add get_slice for supercolumn, tests. add back updated TimeSortTest. we need to pass gcBefore to the filters so Slice can count correctly.

r/m tests that won't pass until we do #303.

change String column names to byte[] to allow user-defined ordering.

r/m time-sorted columns.

introduce column() compatibility layer to make writing tests a little less verbose.

remove ganglia analytics package

move data files into per-table subdirectories. patch by Arin Sarkissian; reviewed by Michael Greene and jbellis