When populating a Datomic database I found it tedious to manually deal with temp IDs to refer entities to each other. It is difficult to both write and read.
I created a simple function TO-TRANSACTION
which accepts
natural Clojure datastructure (nested maps, vectors)
and generates a Datomic transaction to populate DB with
these interlinked entities - temp IDs are assigned automatically,
and references to nested entities (maps) are replaced by their temp IDs.
Also, when working with DB schema, it was inconvenient to work with plain long list of attributes. I quickly loose track of how entities are interlinked, difficult to see how I can improve the schema, difficult to consult it when I write queries.
So TO-SCHEMA-TRANSACTION
function helps to generate a schema-defining
transaction from a template, which resembles real shape of data
how we see it through the Entitiy API.
Example how we can define schema for the well known Seattle sample (distributed with Datomic in the <datomic-root>/samples/seattle/seattle-schema.edn and seattle-data0.edn):
(to-schema-transaction
{:community/name (ext {:db/fulltext true}
:db.type/string)
:community/url :db.type/string
:community/neighborhood {:neighborhood/name :db.type/string
:neighborhood/district {:district/name :db.type/string
:district/region #{:region/n
:region/ne
:region/e
:region/se
:region/s
:region/sw
:region/w
:region/nw}}}
:community/category [ (ext {:db/fulltext true}
:db.type/string) ]
:community/orgtype #{:community.orgtype/community
:community.orgtype/commercial
:community.orgtype/nonprofit
:community.orgtype/personal}
:community/type [ #{:community.type/email-list
:community.type/twitter
:community.type/facebook-page
:community.type/blog
:community.type/website
:community.type/wiki
:community.type/myspace
:community.type/ning} ] })
and some data for it:
(mapcat to-transaction
[{:community/name "15th Ave Community",
:community/category ["15th avenue residents"]
:community/orgtype :community.orgtype/community
:community/type :community.type/email-list
:community/url "http://groups.yahoo.com/group/15thAve_Community/"
:community/neighborhood {:neighborhood/name "Capitol Hill",
:neighborhood/district {:district/region :region/e
:district/name "East"}}}
{:community/category ["neighborhood association"]
:community/orgtype :community.orgtype/community
:community/type :community.type/email-list
:community/name "Admiral Neighborhood Association"
:community/url "http://groups.yahoo.com/group/AdmiralNeighborhood/"
:community/neighborhood {:neighborhood/name "Admiral (West Seattle)"
:neighborhood/district {:district/region :region/sw
:district/name "Southwest"}}}
{:community/category ["members of the Alki Community Council and residents of the Alki Beach neighborhood"]
:community/orgtype :community.orgtype/community
:community/type :community.type/email-list
:community/name "Alki News"
:community/url "http://groups.yahoo.com/group/alkibeachcommunity/"
:community/neighborhood {:neighborhood/name "Alki"
:neighborhood/district {:district/name "Southwest"}}}
{:community/category ["news" "council meetings"]
:community/orgtype :community.orgtype/community
:community/type :community.type/blog
:community/name "Alki News/Alki Community Council"
:community/url "http://alkinews.wordpress.com/"
:community/neighborhood {:neighborhood/name "Alki"}}
{:district/name "Southwest"}
{:community/category ["community council"]
:community/orgtype :community.orgtype/community
:community/type :community.type/website
:community/name "All About Belltown"
:community/url "http://www.belltown.org/"
:community/neighborhood {:neighborhood/name "Belltown"
:neighborhood/district {:district/region :region/w
:district/name "Downtown"}}}]))]
The complete datomic_helpers_sample.clj script demonstrates running the above Seattle sample, and also schema and data for another Datimc sample - [MusicBrainz] (https://github.com/Datomic/mbrainz-sample).
The notation is meant to be intuitively understandable, and here are the precise rules:
(to-schema-transaction type)
We represent schema of Datomic enities by Cloujure maps. Map keys are attribute idents, the key values are attribute types.
The type specification may be either:
Normal datomic types: :db.type/string
, :db.type/float
, etc.
Clojure map - means an entity. It translates to :db.type/ref
type,
and the map is processed recursively to define all its attributes too.
If you specify that your entity has :db/ident
attribute,
no attribute definition is generated for it
(because Datomic already has definition for :db/ident
).
Thus :db/ident
in entity types just serves human readers
of your schema.
Vector means the attibute will have :db.cardinality/many
.
The attirubte type is specified by the nested vector element
(thus only single element vectors make sense).
For example, an attribute stroing multiple strings:
:community/category [ :db.type/string ]
Set means an enum. The attribute is given type :db.type/ref
,
and every element of the set is used as :db/ident
for a
new, separate entity.
An expression (ext <extra properties> <typespec>)
may be
used to annotate attribute type with additional schema properties.
For example:
:community/category [ (ext {:db/fulltext true}
:db.type/string) ]
If several entities share attribute with the same name, you may either repeat the attribute type, or just use any symbol in place of the attribute type, in which case the attribute appearence will be ignored:
:some/repeated-attribue 'defined-above
If the same entity type is referenced from several places,
you may either repeat the entity type definition,
or just use :db.type/ref
in the second appearence.
If the repeated attribute definitions are different, an exception is thrown.
(to-transaction data-map)
Extends data-map
with :db/id
attribute.
If a data-map
key refers to another map, the reference
value is replaced by :db/id
of the child map processed recursively.
If a key refers to a vector, the vector is processed in
similar fasion - all its map elements are replaced by :db/id
's
assigned to them in recursive processing.
All other values (numbers, strings, dates, etc) are left as is.
This processing turns every map encountered into a valid Datomic transactoction map.
Returns a sequence of all those transaction maps, which
may be passed to datomic.api/transact
to populate
datebase with the required set of inter-linked entities
with attributes.
I think the notation may be improved, but the current form is enough for me, and helps me significantly.
Can you improve this documentation?Edit on GitHub
cljdoc is a website building & hosting documentation for Clojure/Script libraries
× close