onyx-datomic

Onyx plugin providing read and write facilities for batch processing a Datomic database.

Installation

In your project file:

[org.onyxplatform/onyx-datomic "0.9.15.0"]

In your peer boot-up namespace:

(:require [onyx.plugin.datomic])

Functions

read-datoms

Reads datoms out of a Datomic database via datomic.api/datoms.

Catalog entry:

{:onyx/name :read-datoms
 :onyx/plugin :onyx.plugin.datomic/read-datoms
 :onyx/type :input
 :onyx/medium :datomic
 :datomic/uri db-uri
 :datomic/t t
 :datomic/datoms-index :eavt
 :datomic/datoms-components []
 :datomic/datoms-per-segment 20
 :onyx/max-peers 1
 :onyx/batch-size batch-size
 :onyx/doc "Reads a sequence of datoms from the d/datoms API"}

:datomic/datoms-components may be used to filter by a datomic index. See the Clojure Cookbook for examples.

Lifecycle entry:

{:lifecycle/task :read-datoms
 :lifecycle/calls :onyx.plugin.datomic/read-datoms-calls}

Attributes

key	type	description
:datomic/uri	string	The URI of the datomic database to connect to
:datomic/t	integer	The t-value of the database to read from
:datomic/datoms-index	keyword	datomic index to use in datomic.api/datoms call
:datomic/datoms-components	keyword	components of the datomic index to use (see datomic.api/datoms documentation)
:datomic/datoms-per-segment	integer	The number of datoms to compress into a single segment
:datomic/read-buffer	integer	The number of segments to buffer after partitioning, default is 1000

read-index-range

Reads datoms from an indexed attribute via datomic.api/index-range.

Catalog entry:

{:onyx/name :read-index-datoms
 :onyx/plugin :onyx.plugin.datomic/read-index-range
 :onyx/type :input
 :onyx/medium :datomic
 :datomic/uri db-uri
 :datomic/t t
 :datomic/index-attribute :your-indexed-attribute
 :datomic/index-range-start <<INDEX_START_VALUE>>
 :datomic/index-range-end <<INDEX_END_VALUE>>
 :datomic/datoms-per-segment 20
 :onyx/max-peers 1
 :onyx/batch-size batch-size
 :onyx/doc "Reads a range of datoms from the d/index-range API"}

Lifecycle entry:

{:lifecycle/task :read-index-datoms
 :lifecycle/calls :onyx.plugin.datomic/read-index-range-calls}

Attributes

key	type	description
:datomic/uri	string	The URI of the datomic database to connect to
:datomic/t	integer	The t-value of the database to read from
:datomic/index-attribute	keyword	datomic indexed attribute
:datomic/index-range-start	any	inclusive start value for the index range
:datomic/index-range-end	any	exclusive end value for the index range
:datomic/datoms-per-segment	integer	The number of datoms to compress into a single segment
:datomic/read-buffer	integer	The number of segments to buffer after partitioning, default is 1000

read-log

Reads the transaction log via repeated chunked calls of d/tx-range. Continues to read transactions until :datomic/log-end-tx is reached, or forever if :datomic/log-end-tx is nil.

Catalog entry:

{:onyx/name :read-log
 :onyx/plugin :onyx.plugin.datomic/read-log
 :onyx/type :input
 :onyx/medium :datomic
 :datomic/uri db-uri
 :datomic/log-start-tx <<OPTIONAL_TX_START_INDEX>>
 :datomic/log-end-tx <<OPTIONAL_TX_END_INDEX>>
 :checkpoint/force-reset? true
 :onyx/max-peers 1
 :onyx/batch-size batch-size
 :onyx/doc "Reads a sequence of datoms from the d/log API"}

Lifecycle entry:

{:lifecycle/task :read-log
 :lifecycle/calls :onyx.plugin.datomic/read-log-calls}

Task will emit a sentinel :done when it reaches the tx log-end-tx (exclusive).

Segments will be read in the form {:t tx-id :data [[e a v t added] [e a v t added]]}.

Log read checkpointing is per job - i.e. if a virtual peer crashes, and a new one is allocated to the task, the new virtual peer will restart reading the log at the highest acked point. If a new job is started, this checkpoint information will not be used. In order to persist checkpoint information between jobs, add :checkpoint/key "somekey" to the task-map. This will persist checkpoint information for cluster (on a given :onyx/tenancy-id) under the key, ensuring that any new jobs restart at the checkpoint. This is useful if the cluster needs to be restarted, or a job is killed and a new one is created in its place.

Attributes

key	type	description
:datomic/uri	string	The URI of the datomic database to connect to
:datomic/log-start-tx	integer	optional starting tx (inclusive) for log read
:datomic/log-end-tx	integer	optional ending tx (exclusive) for log read. Sentinel will emitted when this tx is passed.
:checkpoint/force-reset?	boolean	whether or not checkpointing should be re-initialised from log-start-tx, or 0 in the case of nil
:checkpoint/key	any	optional global (for a given onyx/tenancy-id) key under which to store the checkpoint information. By default the task-id for the job will be used, in which case checkpointing will only be resumed when a virtual peer crashes, and not when a new job is started.
:datomic/read-buffer	integer	The number of segments to buffer after partitioning, default is 1000

commit-tx

Writes new entity maps to datomic. Will automatically assign tempid's for the partition if a value for :datomic/partition is supplied and datomic transaction data is in map form. tx-data returned by datomic.api/transact is injected into the pipeline event map under :datomic/written.

Catalog entry:

{:onyx/name :write-datoms
 :onyx/plugin :onyx.plugin.datomic/write-datoms
 :onyx/type :output
 :onyx/medium :datomic
 :datomic/uri db-uri
 :datomic/partition :my.database/optional-partition-name
 :onyx/batch-size batch-size
 :onyx/doc "Transacts segments to storage"}

Lifecycle entry:

{:lifecycle/task :write-datoms
 :lifecycle/calls :onyx.plugin.datomic/write-tx-calls}

Attributes

key	type	description
:datomic/uri	string	The URI of the datomic database to connect to
:datomic/partition	keyword	Optional keyword. When supplied, :db/id tempids are added using this partition.

commit-bulk-tx (asynchronous)

Writes transactions via the :tx segment key to a Datomic database. The value of :tx should be as if it were ready for (d/transact uri tx). This lets you perform retractions and arbitrary db functions. tx-data returned by datomic.api/transact is injected into the pipeline event map under :datomic/written. Takes advantage of the Datomic transactor's ability to pipeline transactions by asynchronously transacting :onyx/batch-size transactions at once. Transaction futures are then derefed one by one after. Parallelism can thus be controlled by modifying the batch size appropriately. This is the recommended way to transact in bulk.

Catalog entry:

{:onyx/name :write-bulk-datoms-async
 :onyx/plugin :onyx.plugin.datomic/write-bulk-datoms-async
 :onyx/type :output
 :onyx/medium :datomic
 :datomic/uri db-uri
 :datomic/partition :my.database/partition
 :onyx/batch-size batch-size
 :onyx/doc "Transacts segments to storage"}

An example value of :tx would look like the following:

(require '[datomic.api :as d])

{:tx [[:db/add (d/tempid :db.part/user) :db/doc "Hello world"]]}

Lifecycle entry:

{:lifecycle/task :write-bulk-datoms-async
 :lifecycle/calls :onyx.plugin.datomic/write-bulk-tx-async-calls}

Attributes

key	type	description
:datomic/uri	string	The URI of the datomic database to connect to
:datomic/partition	keyword	Optional keyword. When supplied, :db/id tempids are added using this partition.

commit-bulk-tx (synchronous)

Exactly the same as commit-bulk-tx (asynchronous), but transacts each tx completely (blocking on the returned future) before proceeding to the next. You should generally prefer the async version.

Catalog entry:

{:onyx/name :write-bulk-datoms
 :onyx/plugin :onyx.plugin.datomic/write-bulk-datoms
 :onyx/type :output
 :onyx/medium :datomic
 :datomic/uri db-uri
 :datomic/partition :my.database/partition
 :onyx/batch-size batch-size
 :onyx/doc "Transacts segments to storage"}

An example value of :tx would look like the following:

(require '[datomic.api :as d])

{:tx [[:db/add (d/tempid :db.part/user) :db/doc "Hello world"]]}

Lifecycle entry:

{:lifecycle/task :write-bulk-datoms
 :lifecycle/calls :onyx.plugin.datomic/write-bulk-tx-calls}

Attributes

key	type	description
:datomic/uri	string	The URI of the datomic database to connect to
:datomic/partition	keyword	Optional keyword. When supplied, :db/id tempids are added using this partition.

Datomic Params Injection via Lifecycles

The datomic params lifecycles inject datomic dbs or conns into :onyx/fn params.

inject-conn

Injects a datomic conn into the event map. Will also inject as an :onyx/fn param if :onyx/param? is true.

{:lifecycle/task :use-conn-task
 :lifecycle/calls :onyx.plugin.datomic/inject-conn-calls
 :datomic/uri db-uri
 :onyx/param? true
 :lifecycle/doc "Initialises datomic conn as a :onyx.core/param"}

inject-db

Injects a datomic db into the event map. Will also inject as an :onyx/fn param if :onyx/param? is true.

:datomic/basis-t is optional, and if supplied it calls datomic.api/as-of on the db using :datomic/basis-t.

{:lifecycle/task :use-db-task
 :lifecycle/calls :onyx.plugin.datomic/inject-db-calls
 :datomic/uri db-uri
 :datomic/basis-t optional-basis-t
 :onyx/param? true
 :lifecycle/doc "Initialises datomic db as a :onyx.core/param"}

Contributing

Pull requests into the master branch are welcomed.

License

Copyright © 2015 Michael Drogalis

Distributed under the Eclipse Public License, the same as Clojure.