Ports the Datomic Client to the IGraph protocols.
Part of the ont-app library, dedicated to Ontology-driven development.
Follow the getting started guide for Datomic. To get a datomic service up and running.
Leiningen:
[com.datomic/client-pro "0.9.41"]
[ont-app/datomic-client "0.1.0-SNAPSHOT"]
[ont-app/igraph "0.1.4-SNAPSHOT"]
[ont-app/igraph-vocabulary "0.1.0-SNAPSHOT"]
Require as:
(ns my.ns
(:require
[datomic.client.api :as d]
[environ.core :refer [env]]
[ont-app.datomic-client.core :as dg]
[ont-app.igraph.core :as igraph :refer :all]
[ont-app.igraph-vocabulary.core :as igv :refer [mint-kwi]]
)
See the Datomic docs on connecting to a database. This will give you instructions for installing and running Datomic.
For purposes of this discussion, let's assume that you have a datomic service set up on $DATOMIC_HOST : $DATOMIC_PORT, using a database named $DATOMIC_DB_NAME with $DATOMIC_ACCESS_KEY AND $DATOMIC_SECRET.
Somewhere at the top level of your application you will acquire a datomic connection per the Datomic instructions:
(def cfg {:server-type :peer-server
:access-key (env :datomic-access-key)
;; ... "myaccesskey"
:secret (env :datomic-secret)
;; ... "mysecret"
:endpoint (str (env :datomic-host) ":" (env :datomic-port))
;; ... "localhost:8998"
:validate-hostnames false})
(def client (d/client cfg))
(def conn (d/connect client {:db-name (env :datomic-db-name)}))
Graph instances are created with make-graph
, with two arities. This
function returns an instance of
ont_app.datomic_client.core.DatomicClient
, with members :conn
and
:db
:
With just the connection (db will be set as of the most recent transaction):
> (def g (make-graph conn))
{:conn
{:db-name "hello", :database-id "5e4608e8-8e48-49e5-a43f-9b73168ca4b8", :t 1034, :next-t 1040, :type :datomic.client/conn},
:db
{:t 1034, :next-t 1040, :db-name "hello", :database-id "5e4608e8-8e48-49e5-a43f-9b73168ca4b8", :as-of 1034, :type :datomic.client/db}}
>
Or we can provide a db
argument as of any previous transaction:
> (def db-on-valentines-day (d/as-of (d/db conn) #inst "2020-02-14"))
> (def g-on-valentines-day (make-graph conn db-on-valentines-day))
{:conn
{:db-name "hello", :database-id "5e4608e8-8e48-49e5-a43f-9b73168ca4b8", :t 1034, :next-t 1040, :type :datomic.client/conn},
:db
{:t 1034, :next-t 1040, :db-name "hello", :database-id "5e4608e8-8e48-49e5-a43f-9b73168ca4b8", :as-of #inst "2020-02-14T00:00:00.000-00:00", :type :datomic.client/db}}
>
A new database comes with a set of standard schema declarations.
The following supporting attributes are automatically added to those schema declarations:
> dg/igraph-schema
[#:db{:ident :igraph/kwi,
:valueType :db.type/keyword,
:unique :db.unique/identity,
:cardinality :db.cardinality/one,
:doc "Uniquely names a graph element"}
#:db{:ident :igraph/edn?,
:valueType :db.type/boolean,
:cardinality :db.cardinality/one,
:doc
"Domain is string-valued property. True if value should be encoded and
read as an edn representation of some object."}]
>
We can access the Datomic native representation directly:
> (:conn g)
{:db-name "hello", :database-id "5e4608e8-8e48-49e5-a43f-9b73168ca4b8", :t 1034, :next-t 1040, :type :datomic.client/conn}
>
> (:db g)
{:t 1034, :next-t 1040, :db-name "hello", :database-id "5e4608e8-8e48-49e5-a43f-9b73168ca4b8", :as-of 1034, :type :datomic.client/db}
>
Each graph implements the IGraph protocol.
We can access the contents of the graph with the standard IGraph accessor functions.
We can get a lazy sequence of all subjects:
> (subjects g)
(:db/code
:db.sys/reId
:ig-ctest/isa
:db.entity/preds
:db/tupleAttrs
...
)
>
Zero args gives us normal form:
> (g)
{:db.type/instant
{:db/doc
#{"Value type for instants in time. Stored internally as a number of milliseconds since midnight, January 1, 1970 UTC. Representation type will vary depending on the language you are using."},
:db/ident #{:db.type/instant},
:fressian/tag #{:inst}},
:db/excise
#:db{:cardinality #{:db.cardinality/one},
:ident #{:db/excise},
:valueType #{:db.type/ref}},
:db.type/tuple
{:fressian/tag #{:list}, :db/ident #{:db.type/tuple}},
...
}
>
The atom dg/normal-form-timeout
sets the timeout when querying for
normal form, and defaults to 1000 (ms). If the timeout is exceeded it
will throw ex-info of type ::igraph/Intractable
.
One arg gives us a description of a subject s:
> (g :db/code) ;; returns description of :db/code in normal form
#:db{:cardinality #{:db.cardinality/one},
:fulltext #{true},
:doc
#{"String-valued attribute of a data function that contains the function's source code."},
:ident #{:db/code},
:valueType #{:db.type/string}}
>
Two args gives us the set of objects for s and p:
> (g :db/code :db/doc) ;; returns set of objects
#{"String-valued attribute of a data function that contains the function's source code."}
>
(unique (g :db/code :db/doc)) ;; because we know :db/doc is cardinality-1
"String-valued attribute of a data function that contains the function's source code."
>
Three args is truthy:
(g :db/code :db/cardinality :db.cardinality/one) ;; truthy
true
>
As with all IGraph implementations, a
traversal
function may be provided as the p
argument for 2- and 3- arg invocations.
While both IGraph and Datomic's native representation are graph-oriented, the IGraph representation is based on an RDF-inspired subject-predicate-object (SPO) model, whereas the Datomic representation is based on what is sometimes called an entity-attribute-value (EAV) model.
Entities (E) in this case are DB-specific integer identifiers, whereas Subjects (S) are assumed to be keyword identifiers (KWIs), used as (and mappable to) URIs. Subjects 'pivot' off of Entity numbers via attributes which are either :db/ident or :db/unique :db.unique/identity. :db/ident provides faster access, but should be used relatively sparingly for large models, as they must be held always in memory.
The entity-id
function returns the e
for any unique ID in a given DB:
> (dg/entity-id (:db g) :db/cardinality)
41
>
You may find the discussion below on the method for minting KWIs useful.
Attributes (A) and Properties (P) are approximately equivalent. They must be declared in the schema as :db/ident.
Datomic Values (V) and IGraph Objects (O) are also approximately equivalent. Values can be either refs or literal values, interpreted per the Attribute declarations in the Datomic schema. Objects can be specified as KWIs (interpreted as Datomic :db.type/refs), as literal values supported by Datomic, or if the object is not supported natively by Datomic, they may be encoded/decoded as (non-queriable except by regex) EDN strings.
The
ont-app.igraph-vocabulary.core
module provides a multi-method called mint-kwi
. This can be useful
for creating unique keyword identifiers to serve as subjects or
objects in your graph. The default behavior is thus:
> (mint-kwi :myNs/Head :myNs/p1 "foo" :myNs/p2 "bar")
:myNs/Head_p1_foo_p2_bar
>
... where p1
and p2
should be sufficient to uniquely
distinguish your instance of :myNs/Head in whatever universe you
expect to be playing in.
The mint-kwi
multi-method is dispatched on the value of the first
'head' argument. Here is an example of a defmethod that mints a KWI
for movies based on the title and the year of release:
> (defmethod mint-kwi :movie/Movie
[head-kwi & args]
;; Generates unique KWI for <title> made in <year>
(let [{title :movie/title
year :movie/year
} args
_ns (namespace head-kwi)
_name (name head-kwi)
stringify (fn [x]
(cond (string? x) (str/replace x #" " "_")
(keyword? x) (name x)
:default (str x)))
kwi (keyword _ns (str _name "_" (str/join "_"
[(stringify title)
(or year "NoDate")])))
]
kwi))
#multifn[mint-kwi 0x1caf7807]
>
> (mint-kwi :movie/Movie
:movie/title "The Goonies"
:movie/year 1985)
:movie/Movie_The_Goonies_1985
>
The IGraph query
method can be used for datalog queries with either
a vector for simple queries or a map for datomic's
arity-one
format:
> (query g
'[:find ?e ?v
:where [?e :db/ident ?v]])
[[22 :db.type/long]
[38 :db.unique/identity]
[19 :db.alter/attribute]
...
]
>
> (query g
{:query '[:find ?e ?v
:in $ ?a
:where [?e ?a ?v]]
:args [(:db g) :db/ident]
:limit 3
})
[[22 :db.type/long]
[38 :db.unique/identity]
[19 :db.alter/attribute]]
Or if you want to use multiple-arity queries or Datomic's pull syntax, you can access the db with (:db g) and do that directly.
The domain-element?
function returns true for refs which are not
part of the standard schema, and thus presumably part of your domain
model:
> (filter dg/domain-element? (subjects g))
(
:movie/date
:movie/title
:movie/certifications
:movie/genre
:movie/Movie_The_Goonies_1985
...
)
>
Datomic describes its mutability model as "accumulate
only". This
means that you can access any earlier state of the DB using the
as-of
function. Assertions can be added and removed using the
datomic API functions add
and retract
. Assertions which have been
retracted at some point in time are still available to earlier states
of the DB.
This library implements an IGraph mutability model dedicated to this:
> (mutability g)
:ont-app.igraph.core/accumulate-only
>
> (satisfies? igraph/IGraphAccumulateOnly g)
true
>
As it happens, add
is already dedicated to the
igraph/IGraphImmutable
protocol, so the
igraph/IGraphAccumulateOnly
protocol uses retract
to remove assertions, and its antonym claim
to add assertions.
claim
implements the
igraph/add-to-graph
multimethod, and arguments can be in any of its associated triples-format
s.
Any given instance of the graph is immutable...
> (def g' (claim g [:john :fullName "John Smith"]))
#'user/g'
>
> (g :john)
nil
>
> (g' :john)
{:fullName #{"John Smith"}}
>
Since in this example :fullName is not already in the schema, its schema is automatically inferred based on the type of its object, with default cardinality of 'many':
> (g' :fullName)
#:db{:doc #{"Declared automatically while importing"},
:cardinality #{:db.cardinality/many},
:ident #{:fullName},
:valueType #{:db.type/string}}
>
Keywords are presumed to be KWIs, and interpreted as :db.type/ref. Properties that range over keywords need to be declared explicitly in the schema.
If an object is provided whose data type is not supported natively by Datomic, it is stored as an EDN string, and read back in when retrieved:
> (def g' (claim g' [:john :hasVector [1 2 3]]))
#user/g'
>
> (unique (g' :john :hasVector))
[1 2 3]
>
> (g' :hasVector)
{:db/doc #{"Declared automatically while importing"},
:db/ident #{:hasVector},
:db/cardinality #{:db.cardinality/many},
:igraph/edn? #{true},
:db/valueType #{:db.type/string}}
>
However, since these values are stored as strings, our ability to query against such objects is limited. The example above was provided as an illustration, but in practice it would often make more sense to declare it as :db.type/tuple with a :db/tupleType as :db.type/long.
retract
The retract
method implements the
remove-from-graph
multimethod, dispatched on igraph/triples-removal-format.
> (def g' (retract g' [:john]))
#user/g
>
> (g' :john)
nil
>
And we can also call retract with more specific s p or s p o arguments.
The ont-app.datomic-client.core-test
module requires that you have a
datomic service set up on $DATOMIC_HOST : $DATOMIC_PORT, using a
database named $DATOMIC_DB_NAME with $DATOMIC_ACCESS_KEY AND
$DATOMIC_SECRET, for which see the datomic
documentation.
$ lein test
... will test to ensure that the pertinent examples in IGraph's README work, as well as functions specific to datomic-client.
Copyright © 2020 Eric D. Scott
This program and the accompanying materials are made available under the terms of the Eclipse Public License 2.0 which is available at http://www.eclipse.org/legal/epl-2.0.
This Source Code may also be made available under the following Secondary Licenses when the conditions for such availability set forth in the Eclipse Public License, v. 2.0 are satisfied: GNU General Public License as published by the Free Software Foundation, either version 2 of the License, or (at your option) any later version, with the GNU Classpath Exception which is available at https://www.gnu.org/software/classpath/license.html.
Can you improve this documentation?Edit on GitHub
cljdoc is a website building & hosting documentation for Clojure/Script libraries
× close