Permalink
Branch: develop
Commits on Sep 27, 2013
  1. Merge pull request #150 from basho/adt-api-updates

    rzezeski committed Sep 27, 2013
    Update riak_search to support some new core apis
Commits on Aug 1, 2013
  1. Use 1.4.1 deps

    rzezeski committed Aug 1, 2013
Commits on Jul 10, 2012
  1. Fix vnode query crash regression

    rzezeski committed Jul 9, 2012
    While running the Search integration tests I noticed that queries with
    more than one range query were failing with `badarg` during the call
    to `ets:next` in `mi_utils:ets_keys`.  This indicates that the ETS
    table is being destroyed while the keys are being iterated.  After
    more digging it turns out that each query is causing a vnode restart.
    This regression was caused by a recent commit,
    96a97a6, I made to _not_ ignore exit
    signals from processes linked to the search vnode.  The query handler
    links to the vnode and thus causes `handle_exit` to be invoked upon
    completion of the query.
    
    The fix is to not stop the vnode when a linked process exits normally.
Commits on Jun 19, 2012
Commits on Jun 12, 2012
  1. Make async fold truly async

    rzezeski committed Jun 12, 2012
    Need to use the iterator API for `merge_index` to prevent blocking the
    `mi_server` instance.
Commits on Jun 11, 2012
  1. Factor general filter code into Core

    rzezeski committed Jun 11, 2012
    Factor out the repair filter code shared by Search & KV into Core.
  2. Let core extract the chash

    rzezeski committed Jun 9, 2012
    The fact that the chash is needed is a core detail so keep it there.
  3. Let vnode crash on exit signal

    rzezeski committed Jun 7, 2012
    If a linked process, such as the merge index process or worker pool,
    crashes then restart the vnode since it's pid values will be obsolete.
  4. Add repair ability

    rzezeski committed Mar 27, 2012
    Add the ability to repair a Search partition.  If a partition's data
    is lost or corrupted this command may be used to rebuild it from
    adjacent partitions.  This is preferable over a list-keys + re-index
    since it is more targeted, should complete more quickly and should put
    less strain on the cluster.
  5. Remove trailing whitespace

    rzezeski committed Mar 28, 2012
Commits on Jun 8, 2012
  1. Honor the `?` wildcard

    rzezeski committed Jun 8, 2012
    For some reason the range-sized op was passing down a size of `all`
    for the character wildcard.  Instead, pass down the correct size.
  2. Remove solr supervisor

    rzezeski committed Jun 8, 2012
    The solr supervisor does not supervise anything and hasn't since
    version 0.14.0, maybe even before that.
Commits on Jun 7, 2012
  1. Log stats for 3 unique search APIs

    rzezeski committed Jun 6, 2012
    Riak Search currently has 3 unique search APIs.
    
    1. Search for matching docs - `search`
    
    2. Search as input to map/reduce - `search_fold`
    
    3. Search and then fetch documents - `search_doc`
    
    Since each does varying amounts of work it makes more sense to track
    stats separate than together.  E.g. `search_doc` has potential to
    perform much more work than `search` since it returns matching objects
    and not just the keys.
Commits on Jun 6, 2012
Commits on May 26, 2012
  1. Remove some remaining qilr stuff

    rzezeski committed May 26, 2012
    There were still bits of qilr clinging to the code base.  I left the
    qilr stuff in the one EQC test alone for now as perhaps that test
    could be resurrected in the future or used as inspiration.  In any
    case if that test goes the entire module should be deleted at once
    rather than just removing some of it.
  2. Remove trailing whitespace

    rzezeski committed May 26, 2012
Commits on May 24, 2012
  1. Fix `encode_results` and FL

    rzezeski committed May 24, 2012
    * The `encode_search_doc` expects a proplist, not a tuple.
    
    * Need to parse FL.
  2. Remove defensive programming

    rzezeski committed May 24, 2012
    1. There is no chance for `get_schema` to return an error tuple when
    passed an index.  Instead it will throw.
    
    2. It will only throw if something is horribly wrong, so let it crash
    the process and kill the conn rather than doing defensive programming.
Commits on May 23, 2012
  1. Move `replace_schema_defaults` into utils

    rzezeski committed May 23, 2012
    This is another function that is shared between PB & HTTP.
  2. Fix dialyzer errors

    rzezeski committed May 23, 2012
    * The `search_field` can be binary because of inline fields code.
    
    * Typo in `riak_search_test`
  3. Use common fun for PB query

    rzezeski committed May 23, 2012
    Use the common `run_query` for the PB query execution.
  4. Move common HTTP/PB `run_query`

    rzezeski committed May 23, 2012
    Centralize the common logic between HTTP & PB query execution.
  5. Remove trailing whitespace

    rzezeski committed May 23, 2012
Commits on May 22, 2012
Commits on May 21, 2012
  1. Clear schema cache after schema update

    rzezeski committed May 21, 2012
    After updating the schema make sure to clear the cache on all nodes to
    make sure the update propagates to all nodes immediately.  If one of
    the remote calls fails then that node will still use the old schema so
    it's not perfect.  In the future the schema should be gossiped in a
    manner similar to the ring in order to get convergence.
  2. Remove trailing whitespace

    rzezeski committed May 21, 2012
Commits on May 8, 2012
Commits on Apr 25, 2012
  1. Ignore ordering of positions

    rzezeski committed Apr 25, 2012