RELEASE_NOTES.txt

                              KijiSchema Release Notes

Version 1.6.0
  * SCHEMA-892. [Breaking Change] Remove hadoop1 and HBase 0.92 support.
  * SCHEMA-874. Bumps root pom version and our slf4j dependencies. Removes
      KijiSchema's implementation of slf4j's LoggerFactory.
  * SCHEMA-875. Revert binary incompatibility introduced in SCHEMA-826.
  * POM-40. Require JDK 7 to compile Kiji projects.
  * SCHEMA-851. Deprecate CellSpec table reader overrides; fix hashCode & equality
      of KijiDataRequest; cleanups to CellDecoderProvider.
  * SCHEMA-695. Rebase and merge all Cassandra code onto master.

Version 1.5.0
  * SCHEMA-828. Remove finalization in favor of DebugResourceTracker.
  * SCHEMA-826. Make KijiURI, KijiInstaller, and KijiFactory generic to the
      underlying database.
  * SCHEMA-823. Add Kiji Services tool.
  * SCHEMA-775. Refactor KijiColumnNameTranslator/HBaseColumnName/LayoutCapsule

Version 1.4.3
  * SCHEMA-813. Split ZooKeeperFactory out of HBaseFactory, remove LocalLock,
      reference track Curator ZK connections

Version 1.4.2
  * SCHEMA-802. Log warning when uninstalling instances and deleting tables
      with registered users
  * SCHEMA-776. Enable bridge support for CDH 4.6

Version 1.4.1
  * SCHEMA-741. Integration tests now clean up dummy instances when using an
      external HBase.
  * SCHEMA-678. If HADOOP_HOME, HBASE_HOME, or JAVA_HOME are not set the kiji
      script will attempt to resolve those environment variables by
      utilizing the corresponding binary.

Version 1.4.0
  * SCHEMA-675. Adds support for configuration of Kiji column name translation.
  * SCHEMA-687. DebugResourceTracker no longer accepts null error messages.
  * SCHEMA-672. DecodedCell and KijiCell now support getting the reader Schema
      with which the cell was decoded (if any). Decoded cell adds a constructor
      which accepts a writer schema, reader schema, and value.
  * SCHEMA-648. Adds support for CDH5 and HBase 96 and updates platform bridge
      API to be compatible. Due to the remove of client-exposed row locks,
      deleteMapFamily will no longer hold a lock while deleting qualifiers.

Version 1.3.7
  * SCHEMA-663. KijiTablePool now properly sets the retain count when spawning
      tables to reach the minimum size.
  * SCHEMA-656. Resources which should be closed before JVM shutdown may now
      be registered with DebugResourceTracker, which will log errors for all
      unclosed resources when the JVM shuts down.
  * SCHEMA-651. cleanup deleted instances from zookeeper in
      /kiji-schema/instances.
  * SCHEMA-491. Share/reuse Zookeeper connections instances when possible.
  * SCHEMA-650. The KijiTableReaderPool now retains the KijiTable associated with
      its KijiReaderFactory. Closing the pool will release the table.

Version 1.3.5
  * SCHEMA-627. The schema table tool now permits listing all schemas and
      output to StdOut.

Version 1.3.4
  * SCHEMA-604. Adds support for MapR's distribution of HBase.
  * SCHEMA-557. Removes HTablePool usage.
  * SCHEMA-605. Fixes compatibility with the latest 1.X Hadoop and 0.94.X
      HBase releases.

Version 1.3.3
  * SCHEMA-584. KijiTableAnnotator now supports storing string key-value
      annotations on tables and columns.

Version 1.3.2
  * SCHEMA-586. KijiTableReaderPool now supports managing a pool of open
      KijiTableReaders for thread safe concurrent reads.

Version 1.3.1
  * SCHEMA-544. Introduce new layout version 1.4 with PROTOBUF cell encoding.
  * SCHEMA-574. Fixed broken validation logic from Avro.
  * SCHEMA-581. Fixed potential race condition in client IDs used to identify
      users of Kiji instances and Kiji tables.

Version 1.3.0
  * SCHEMA-548. Introduce RAW_BYTES cell encoding that expose pure bytes cells.
  * SCHEMA-547. AtomicKijiPutter.checkAndCommit() can check for the
      non-existence of a cell.
  * SCHEMA-515. Exposes table and data request in HBaseKijiRowData.
  * SCHEMA-485. Adds profiling pointcuts for HBase reads/writes.
  * SCHEMA-514. DeleteTool correctly formats the delete confirmation message
      when deleting all entries from a cell.
  * SCHEMA-543. Added a table layout containing all relevant cell types that
      can be used in tests.

Version 1.2.0
  * SCHEMA-335, SCHEMA-489. Fix: FormattedEntityId and KijiTableLayout trigger
      NPE if RowKeyFormat2.salt is null. Omitted salt entries now default to a
      2-byte MD5-based hash prefix.
  * SCHEMA-474. KijiTable maintains a pool of HTable objects.
      Readers, writers, scanners and pagers borrow HTable objects from the pool.
  * SCHEMA-469. Fix: VersionInfo should recognize backward-compatibility
      requirements for instance versions.
  * SCHEMA-396. HBase KijiTable readers and writers can now receive table
      layout updates by subscribing to a LayoutConsumer list managed by the
      table. Registered readers and writers will recreate their layout dependent
      state in response to a table layout update.

Version 1.1.2
  * SCHEMA-358. Adds a withTable(TableLayoutDesc) method to
      InstanceBuilder.TableBuilder.
  * SCHEMA-475. Include support for profiling. Use the enable-profiling.sh
      script to switch to profiling-enabled KijiSchema (and KijiMR) jars.
  * SCHEMA-462. Include support for CDH 4.3.
  * SCHEMA-258. Fix: HBaseKijiTable.getRegions() closes the underlying HTable.
  * SCHEMA-415. KijiRandomRowFilter and KijiFirstKeyOnlyColumnFilter
  * SCHEMA-467. Fix: Upgrade checkin notification thread may hang.
  * SCHEMA-466. Fall back to $HADOOP_HOME/lib/native/ for native libraries.
      Correctly detect native libraries in CDH 4.2 and CDH 4.3 packaging.
  * SCHEMA-465. Fix: Can't create 256 regions on a table with 1 byte hash prefix
  * SCHEMA-459. Improve performance of ColumnId translation.
  * SCHEMA-456. HashedEntityId should throw an exception when getting
      row key components.

Version 1.1.1
  * SCHEMA-455. Add CellByteSizeAsValueFilter for computing table sizes.
  * SCHEMA-445. Fixes a bug in the DataRequest adapter that causes requests
      with mixed paging and non-paging columns to discard all values from all
      columns in a row, causing IOException while decoding the cells later on.
  * SCHEMA-139. Correctly close Kiji instances if an exception occurs during
      construction.
  * SCHEMA-332. Add basic maven archetype for KijiSchema-based user projects.
  * SCHEMA-439. Properly close HTable connections in HBaseKijiBufferedWriter.
  * SCHEMA-433. Kiji ls now defaults to listing Kiji instances if no argument
      is given.
  * SCHEMA-440. An IOException or KijiNotInstalledException occuring during the
      instantiation of an HBaseKiji instance no longer leads to retained
      reference count warnings once the JVM garbage collects.
  * SCHEMA-438. Fix bug detecting some HBase versions in compatibility shims.

Version 1.1.0
  * SCHEMA-434. Fixes a race condition in ZooKeeperClient which might cause
      a NPE when closing a Kiji instance.
  * SCHEMA-429. Kiji command line now asks the user to type in the instance or
      table name when uninstalling or deleting them.
  * SCHEMA-405. Adds KijiRowKeyComponents, a way to specify the components
      of a Kiji row key independent of a RowKeyFormat. May be converted into
      an EntityId, given a KijiTable or EntityIdFactory.
  * SCHEMA-388. KijiRowScanner handles HBase scanner timeouts and reopens
      the scanner as needed. Introduces two scanner options:
       - KijiScannerOptions.setReopenScannerOnTimeout() is true by default,
	 which means HBase scanners are automatically reopened on timeouts
	 instead of throwing a ScannerTimeoutException;
       - KijiScannerOptions.setRowCaching() configures the number of rows
	 fetched for each RPC to the region server. By defaults, this has
	 the of value the configuration property hbase.client.scanner.caching,
	 which is 1 unless overridden in hbase-site.xml.
  * SCHEMA-380. Version pagers may now be created on fully-qualified columns
      belonging to paging-enabled map-type families.
  * SCHEMA-355. KijiRowKeySplitter is parametrized by row key resolution.
  * SCHEMA-346. AtomicKijiPutter is no longer restricted to writing within a
      single locality group.  Calls to commit() no longer accept a locality group name.
  * SCHEMA-354. Change use of filters to comply with HBase 0.92 and 0.94 semantics.
    This provides experimental support for HBase 0.94, though this is not the
    default.  KijiSchema uses HBase filters to convert KijiDataRequests into the
    smallest possible HBase request. Where columns have a different maxVersions,
    columns with maxVersions not equal to 1 may communicate more values across
    the wire than are requested; the actual amount of data transferred may be as
    many as the largest maxVersions observed in the data request.  Previously
    (<= 0.92) all per-column maxVersions could be respected on the wire in
    addition to within KijiRowData.
  * SCHEMA-283. Add more table and and locality group properties to
      TableLayoutDesc. These properties control e.g. max HFile size and bloom
      filter usage. You must use layout version "layout-1.2.0" to expose
      these properties.
  * SCHEMA-339. JVM System property to configure KijiClientTest to use a
      real HBase instance.
  * SCHEMA-322. Restrict behavior of KijiPager on map-type families.
      Paging on a map-type family yields qualifiers only, no actual cell.
  * SCHEMA-323. Adds KijiTable.getReaderFactory() to provide custom readers.
  * SCHEMA-327. AvroCellEncoder uses the proper writer schema,
      instead of the reader schema.
  * SCHEMA-187. Allows decoding records with the writer schema when no
      specific record is available (ie. when there is no reader schema).
  * SCHEMA-324. bin/kiji supports loading Hadoop distro-specific dependencies
      by inferring from "bin/hadoop version" or $KIJI_HADOOP_DISTRO_VER.

Version 1.0.2
  * SCHEMA-317. Silence spurious warning about not using GenericOptionsParser.
  * SCHEMA-318. Fixes a bug when closing KijiBufferedWriters.
  * SCHEMA-306. Fixes a bug on KijiPager on fully-qualified columns.
      There is a known issue with the versions returned from a map family.
  * SCHEMA-314. Row filters constructors are being replaced with
      factory methods such Filters.and(...) and Filters.or(...).
  * SCHEMA-303. Allows proper serialization of row filters.
      Contributed by Jeff Kolesky (jeff.kolesky@gmail.com).
  * SCHEMA-304. Adds additional column filters.
  * SCHEMA-297. Adds KijiCell.equals() and .hashCode().

Version 1.0.1
  * SCHEMA-254. Adds cellList methods to KijiRowData.
  * SCHEMA-296. Adds hashCode and equals methods for Filters.

Version 1.0.0
  * SCHEMA-227. Fix get/scan behavior on tables with no families.
  * SCHEMA-284. Fix usage doc string for kiji scan --start-row/limit-row flags.
  * SCHEMA-37. Adds stability annotations to all classes.
  * SCHEMA-280. Removes deprecated close() method from KijiTable.  Use
      table.release().
  * SCHEMA-8. Adds Schema Table Tool for reading and writing to a KijiSchemaTable.
  * SCHEMA-115. KijiManagedHBaseTableName no longer manages names for SchemaV5
      tables.
  * SCHEMA-287. Suppress slf4j classpath warnings without modifying classpath.
  * SCHEMA-189. Adds KijiBufferedWriter.
  * SCHEMA-106. Move CellSpec to org.kiji.schema.layout and mark as Framework.

Version 1.0.0-rc5
  * SCHEMA-240. Composite key format support in all CLI tools.
  * SCHEMA-134. Hides KijiDataRequestValidator's constructor and
      changes validate()'s signature. Framework developers should
      use the factory method validatorForLayout() to get a validator
      which can validate KijiDataRequests against a table layout.
  * SCHEMA-261. Bugfix: Restrict KijiRowData.getValues(family) and
      getMostRecentValues() to only work on map-type column families.
  * SCHEMA-238. Adds a warning for new tables created with a layout
      that uses RowKeyFormat. New tables should be made with RowKeyFormat2.
  * SCHEMA-239. KijiTable.close() is deprecated, use KijiTable.release().
  * SCHEMA-262. KijiPager and KijiRowScanner throw unchecked KijiIOException.
  * SCHEMA-173. Improves CLI parsing of EntityIds.  Adds abstract
      toShellString() to EntityId which returns a string which can be
      copied into CLI.
  * SCHEMA-207. KijiURIBuilder supports relative paths to
      DEFAULT_HBASE_URI.
  * SCHEMA-247. The ls tool can no longer perform get and scan.
  * SCHEMA-241. Change audience annotations of KijiDeleter,
      KijiIncrementer, and KijiPutter to @ApiAudience.Public.
  * SCHEMA-139. KijiTableKeyValueDatabase methods return the true
      instance type, not the interface type.
  * SCHEMA-248. Kiji.modifyTableLayout() no longer takes a table name.
      The Kiji.modifyTableLayout() methods that take a String name are
      now deprecated.
  * SCHEMA-223. Adds backup methods to KijiSystemTable and SystemTable
      backup to MetadataTool.
  * SCHEMA-244. Kiji.createTable() now takes a TableLayoutDesc object.
      The Kiji.createTable() methods that take a KijiTableLayout object
      are deprecated.
  * SCHEMA-233. Extract kiji get functionality out of kiji ls.
  * SCHEMA-242. Enforce constraint of no non-hashed components
      in a FORMATTED row key with SuppressKeyMaterialization=true.
  * SCHEMA-141. Improved KijiTablePool API.
  * SCHEMA-234. Extract kiji scan functionality out of kiji ls.
  * SCHEMA-219. Adds KijiWriterFactory with access in KijiTable and
      AtomicKijiPutter with access in KijiUtilityWriterFactory.
  * SCHEMA-103. Remove ColumnNameTranslator from public API.
  * SCHEMA-140. Update API for metatable backups. This changes the
      backup format. Please re-backup your tables using the new API.
  * SCHEMA-136. Make KijiRowKeySplitter a non-static class. Get an instance
      with the KijiRowKeySplitter.get() method.
  * SCHEMA-145. Change KijiColumnFilter.Context to abstract class.
  * SCHEMA-199. VALID_INSTANCE_NAME should not match the empty string.
  * SCHEMA-9. Add a CLI tool for KijiSchema System Table.
      supports putting and getting properties from the system table.
  * SCHEMA-198. KijiColumnName validates family names using
      KijiNameValidator. KijiInvalidNameException is unchecked.
  * SCHEMA-135. Clean up API for table layout backups. This changes the
      backup format. Please re-backup your tables using the new API.

Version 1.0.0-rc4
  * BENTO-15. When run within a BentoBox, the 'kiji' script will notify you
      of a new BentoBox version if one is available.
  * SCHEMA-218. Fix instance-listing capability in 'kiji ls'.
  * SCHEMA-67. Remove unused classes (CounterManager, NumberParser).
  * SCHEMA-197. Layout versions now use the "layout" namespace.
      Changes the expected layout format version to "layout-1.0" or
      "layout-1.1". This avoids a collision with instances, etc. All layouts
      in resources have had their version numbers changed to "layout-1.0" or
      "layout-1.1". Existing layouts with a 'kiji-1.0' version string are
      supported. "kiji version" will now show the minimum and maximum layout
      versions that this kiji version can recognize.
  * SCHEMA-211. Use "default" instance in default KijiURIBuilder.
  * SCHEMA-208. Make KijiURIException unchecked.
  * SCHEMA-201. Fix NPE in 'kiji install'.
  * SCHEMA-196. Instance format should not use "kiji" protocol namespace.
      Changes the instance format installed via 'kiji install' to
      'system-1.0', rather than 'kiji-1.0'. This avoids a namespace collision
      with layouts, etc. Existing deployments with a 'kiji-1.0' instance
      format are still supported.
  * SCHEMA-151. Add org.kiji.schema.util.ProtocolVersion to manage
      format/protocol version compatibility.
  * SCHEMA-150. Update CLI tools to support composite entity IDs.
      Entity IDs can now be specified as:
        --entity-id=hbase=hex:deadc0de as hexadecimal HBase keys
        --entity-id=hbase=[utf8:]'hbase\x0akey\x00' as UTF-8 encoded HBase keys
        --entity-id=[kiji=]'kiji-key' for raw, hashed or hash-prefixed Kiji keys
        --entity-id="['component1', 2, 'comp3']" for Kiji formatted row keys.
  * SCHEMA-151. Stricter checking of layout "version" fields for compatibility
      with supported layout features. VersionInfo.get(Client/Cluster)DataVersion()
      now return a ProtocolVersion, not a string.
  * SCHEMA-190. Removed KijiDataBuffer.
  * SCHEMA-27. Add timestamp-related constants to KConstants:
      KConstants.BEGINNING_OF_TIME = 0
      KConstants.END_OF_TIME = Long.MAX_VALUE
  * SCHEMA-3. Add support for composite entity ids. You can now create
      compound keys containing a mix of hashed, string, integer, and null
      components.
  * SCHEMA-175. kiji script can now use tools provided by kiji-mapreduce. To make
      kiji-mapreduce tools available for use through the kiji script, make sure
      the environment variable KIJI_MR_HOME is set to the path to a kiji-mapreduce
      distribution.
  * SCHEMA-153. Adds a getRegions() method to KijiTable. Returns logical
      region divisions in a kiji table.
  * SCHEMA-119. Remove TableLayoutSerializer class.
  * SCHEMA-147. Hide OperatorRowFilter. You should use AndRowFilter or
      OrRowFilter directly.
  * SCHEMA-118. KijiTableLayout constructors are now hidden. Create a new layout
      with KijiTableLayout.newLayout(TableLayoutDesc) or create one updated from
      an old one with KijiTableLayout.createUpdatedLayout(TableLayoutDesc,
      KijiTableLayout).
  * SCHEMA-10. CLI tools use KijiURIs:
      - "kiji ls" flags --instances and --table are remove and inferred
        from URI in --kiji=kiji://...
      - flag --table is no longer present; it is superseeded by the table
        component from Kiji URIs.
  * SCHEMA-123. Removed spin-wait from deleteTable in HBaseKiji.
  * SCHEMA-169. Add KijiPager class that allows users to easily paginate and
      iterate over values from a column while reading from a KijiTable.
  * SCHEMA-99. KijiConfiguration removed. Use a KijiURI paired with a
      Hadoop Configuration object instead:

      KijiURI uri = KijiURI.newBuilder().withInstanceName("default")
          .withTableName("myTable").build();
  * SCHEMA-155. Merged methods in KijiAdmin into Kiji.
  * SCHEMA-133. Introduce KijiDataRequestBuilder class. KijiDataRequest
      objects are now immutable. Construct one with KijiDataRequest.create(),
      or call KijiDataRequest.builder() to get a KijiDataRequestBuilder.
  * SCHEMA-102. Removed overloaded variety of toHBaseFilter that accepted a
      ColumnNameTranslator.
  * SCHEMA-131. Move HBase classes to org.kiji.schema.hbase.
  * SCHEMA-82. KijiCell is now the uniquely located cell in a table with
      family, qualifier, and timestamp.  KijiCounters no longer exist and
      are represented as KijiCell<Long> with no writer schema instead.
  * SCHEMA-143. KijiURI constructors are now hidden.  KijiURIs should be built
      with KijiURIBuilders.
  * SCHEMA-109. KijiInstaller is no longer static.  Get an instance of it with
      KijiInstaller.get().

Version 1.0.0-rc3
  * SCHEMA-83. Added containsCell method to KijiRowData.
  * SCHEMA-154. KijiAdmin is now an interface. Use Kiji.getAdmin() to
      get an instance.
  * SCHEMA-105. Hide KijiColumnPager from public API.
  * SCHEMA-127. Command-line tools --confirm is replaced with --interactive,
      which suppresses yes/no prompts if false.
  * SCHEMA-142. KijiTableReader.getScanner() now takes a KijiScannerOptions
      argument as configuration.
  * SCHEMA-50. Kiji.close() is replaced by Kiji.retain() and Kiji.release().
      Kiji instances are created with a reference count of 1. Use the retain()
      method to increase this reference count, if another consumer holds a
      reference to this object.

      You should call release(), or ReferenceCountableUtils.releaseQuietly()
      before discarding the reference. Instances are closed when the reference
      count hits zero.
  * SCHEMA-97. org.kiji.schema.Kiji is now an interface, not a class.
      Use Kiji.Factory.open() to get a Kiji instance.
  * SCHEMA-157. Add KijiTable.getURI().
  * SCHEMA-159. Fixes a bug where KijiRowData.getValues() would return values
      in the wrong order.
  * SCHEMA-128. Add minPoolSize option to KijiTablePool. Kiji will aggressively
      open the min number of connections the first time a table connection is
      requested and will maintain at least that size of a pool indefinitely.
  * SCHEMA-17. DeleteTool replaces and improves DeleteTableTool.
      `kiji delete-table` has been removed. See `kiji delete --help` for
      instructions on using this tool, which can delete rows / cells in addition
      to tables.
  * SCHEMA-129. KijiTableReader converted to an interface. Use
      KijiTable.openTableReader() to get an instance of KijiTableReader for a
      given table.
  * SCHEMA-114. Fix NullPointerException in `kiji install`.
  * SCHEMA-20. Improve validation of cell schema changes.
      - Counters must specify schema storage 'final'.
      - Final columns require inline schema specifications and cannot change.
      - Schema storage cannot be modified.
  * SCHEMA-113. Bugfix: `kiji metadata` command reports an incorrect usage message.
  * SCHEMA-1. Writes through KijiTableWriter are validated against common reader schema.
  * SCHEMA-111. Adds RELEASE_NOTES.txt
  * SCHEMA-90. KijiURIException subclasses IOException.
  * SCHEMA-13. Removed HBaseTableRecordReader from public API.

Version 1.0.0-rc2
  * SCHEMA-93. Deprecate the org.kiji.schema.mapreduce package. This marks
      classes in this package as @Deprecated in preparation for the eventual
      release of KijiMapReduce.
  * SCHEMA-52. Initial effort at multiple Hadoop version support. While
      MapReduce code requires CDH4, non-MR-based operations provided by
      KijiSchema can be performed with both CDH4 or Apache Hadoop 1.0
      and HBase 0.92.1.
  * SCHEMA-60. Multiple platform profiles for KijiSchema. To facilitate
      multiple Hadoop version support in Kiji, your project must now
      explicitly declare a reference to the underlying Hadoop distribution.
      Specifically, put this in your pom.xml file to continue working with
      CDH4:
      <dependency>
        <groupId>org.kiji.platforms</groupId>
        <artifactId>kiji-cdh4-platform</artifactId>
        <version>1.0.0-rc2</version>
        <scope>provided</scope>
      </dependency>
      See https://github.com/kijiproject/kiji-platforms/blob/master/README.md
      for a full list of available platforms.
  * SCHEMA-7. Add metadata backup and restore tools ('kiji metadata')
  * SCHEMA-28. Removed KijiDataWrapper and its subclasses. This change
      refactors large parts of KijiSchema and has several incompatible
      changes associated with it:
      ** KijiURIs are now used more pervasively to reference Kiji instance
         and table names. KijiConfiguration is deprecated; you should open
         Kiji instances through KijiURIs.
      ** KijiCellEncoder and KijiCellDecoder were significantly revamped.
      ** KijiTable is now an interface.
      ** Interaction between Kiji, KijiTable, and KijiTableReader/Writer
         reworked. See the javadoc for the Kiji.open() and
         KijiTable.openTableReader() / openTableWriter() methods.
      ** KijiTableWriter was significantly simplified. You no longer need to
         specify the schema when writing to a cell; the schema is induced from
         the type when you write. The forms of the put() method where schemas
         are specified have been removed.
      ** KijiTableReader and KijiRowData have been significantly simplified.
         You no longer need to explicitly specify the type when you read a cell
         (e.g., via getStringValue()). Methods like getIntValue(),
         getStringValues(), etc. have been removed. Instead, a universal
         getMostRecentValue() method can now use type inference in combination
         with the default reader schema specified in the table's layout to
         deserialize the data with the correct type signature.
  * SCHEMA-40. Add audience annotations to classes. Provisional annotations
      have been added to indicate whether classes are intended for use by all
      clients, other framework modules, or only the current module.
      See github.com/kijiproject/annotations for more documentation.
      Related to this, several classes have been made 'final', to prevent end
      users from extending them.
  * SCHEMA-51. Use Kiji artifacts, not odiago/wibi artifacts.
  * SCHEMA-43. bugfix: bin-script cannot be directly symlinked.
  * SCHEMA-36. kiji-schema jar is now always added to the front of the
      classpath.
  * SCHEMA-34. Handle general exceptions more gracefully in Kiji tools.