Permalink
Commits on Mar 8, 2010
  1. Add link to new v0.2 branch

    felixge committed Mar 8, 2010
  2. Current status update

    felixge committed Mar 8, 2010
Commits on Jan 27, 2010
  1. Finished load test

    Also found a bug in remove(), fixed now.
    felixge committed Jan 27, 2010
  2. ensure _docs[] doesn't contain deleted docs when load()ing

    Signed-off-by: Aaron Heckmann <aaron.heckmann@gmacfs.com>
    Aaron Heckmann committed with felixge Jan 27, 2010
  3. Test for load()

    felixge committed Jan 27, 2010
  4. Test case for e415f8

    felixge committed Jan 27, 2010
Commits on Jan 22, 2010
  1. Feature: Memory-only support

    It is now possible to create a memory-only dirty database by omitting
    the `filename` parameter of the constructor. This is useful for testing.
    
    See: http://github.com/felixge/node-dirty/issues#issue/1
    felixge committed Jan 22, 2010
Commits on Jan 15, 2010
  1. Removing wrong paste

    felixge committed Jan 15, 2010
Commits on Jan 14, 2010
  1. Doc typos

    felixge committed Jan 14, 2010
Commits on Jan 13, 2010
  1. More explanation what Dirty is

    felixge committed Jan 13, 2010
  2. Added note about append only

    felixge committed Jan 13, 2010
  3. Key value store, not id value

    Damn you search & replace! : )
    felixge committed Jan 13, 2010
  4. Updated latest benchmark results

    felixge committed Jan 13, 2010
  5. Renamed 'key' to 'id'

    Seems to be the accepted standard in the NoSQL community. Oh well ...
    felixge committed Jan 13, 2010
  6. Bug fixes for remove

    felixge committed Jan 13, 2010
  7. Update docs

    Added new benchmark results & docs for Dirty.remove()
    felixge committed Jan 13, 2010
  8. Support for remove + bug fixes

    Adding remove required some rather serious underlaying changes.
    There might be a few bugs now, but things just got a lot faster
    and awesome!
    felixge committed Jan 13, 2010
Commits on Jan 12, 2010
  1. Reduced default flush interval to 10

    Seems to have no impact on performance, but is very convenient!
    felixge committed Jan 12, 2010
  2. Unlink file after flush test

    felixge committed Jan 12, 2010
  3. Added docs & license

    felixge committed Jan 12, 2010
Commits on Jan 11, 2010
  1. Fixing small mistake in benchmarks

    new Date() was executed on every loop by mistake, but it doesn't seem
    to have a measurable impact on the benchmark.
    felixge committed Jan 11, 2010
Commits on Jan 9, 2010
  1. Fix implied global

    felixge committed Jan 9, 2010
Commits on Jan 8, 2010
  1. Made filter benchmark more honest

    Now there is actual filtering happening, and we're still getting
    ~10mio. docs / sec processed. Need to start testing with only filtering
    through all records once, but that will take a lot of records for
    testing. After that there might still be some v8 optimizations which
    are screwing with the performance, but we're certainly getting closer
    to the truth here.
    
    975100 docs added in 10003ms
    flushed to disc, starting filtering ...
    
    Filtered 98485100 docs in 10066 ms 	(9783936 per sec)
    felixge committed Jan 8, 2010
  2. Rename Dirty.docs to Dirty._docs

    Users should not access the document array directly.
    felixge committed Jan 8, 2010
  3. Less flawed benchmark for filtering.

    The previous benchmark was not entirely fair as it was not accessing
    any properties of the filtered documents.
    
    This version accesses the ._key property of each element, but there is
    no actual filtering happening. When doing real filtering the performance
    should become aligned to the set() performance where the main bottleneck
    is pushing elements onto the array.
    
    Anyway, this being said, here are the new, less-flawed numbers:
    
    105800 docs added in 1009ms
    flushed to disc, starting filtering ...
    
    Filtered 25180400 docs in 1004 ms 	(12540040 per sec)
    felixge committed Jan 8, 2010
  4. Small fix

    felixge committed Jan 8, 2010
  5. Only emit flush event if all docs are on disk

    There was a bug causing the flush event to potentially emit while some
    documents were still in memory and not yet scheduled for flushing.
    felixge committed Jan 8, 2010
  6. I like singular

    felixge committed Jan 8, 2010
  7. Use .dirty extension for dirty files

    Also gitignore *.dirty
    felixge committed Jan 8, 2010
  8. Output that test is done

    felixge committed Jan 8, 2010