The Ontology Builder is a Docker Image that executes the build process, testing and publishing the git-based content of an Ontology "Family".
The Ontology Builder can process any Ontology Family as long as it is based into one git repository.
It has as input a git-clone of the repository of the Ontology Family (the source) and as output a target directory with tested and publishable content.
The Ontology Builder can be run "locally" on your Ontology Development machine provided that you have installed Docker.
Run the publisher with the following command:
./docker-run.sh --run
If you'd like to have a shell inside the Docker Container use this:
./docker-run.sh --shell
If you'd like the ontology-publisher's container to use your local drive for both input and output then add
the --dev
parameter:
./docker-run.sh --shell --dev
If you'd like the ontology-publisher's container to start with a clean slate in the output directory then add --clean
:
./docker-run.sh --shell --dev --clean
If you'd like to just build the image:
./docker-run.sh --buildimage
The --rebuildimage
option forces the image to be rebuilt with the ability
to change the default values for certain environment variables (see docker-run.sh --rebuild option), e.g .:
env ONTPUB_FAMILY="onto" ./docker-run.sh --rebuildimage
sets the default ontology family to onto
.
If you'd like to publish the image to Docker Hub:
./docker-run.sh --pushimage
When running ontology-publisher it is possible to set environment variables appropriate for the processed ontology as follows:
env <VARIABLE_NAME>=<VARIABLE_VALUE> [<VARIABLE_NAME1>=<VARIABLE_VALUE1>...] ./docker-run.sh [OPTION...]
for example:
env ONTPUB_FAMILY="onto" ./docker-run.sh --shell --dev
List of allowed <VARIABLE_NAME>
(and defaults for <VARIABLE_VALUE>
):
ONTPUB_FAMILY
:- ontology name (base for many other values), e.g.onto
(default:fibo
); this variable is also possible to set during build time (see--rebuildimage
option)URI_SPACE
:- a common "URI namespace" for all ontologies - see 4.2 Patterns for resource URIs, e.g.https://spec.edmcouncil.org/fibo/ontology/
; overrides ontology IRI=https://${ONTPUB_SPEC_HOST}/${ONTPUB_FAMILY}/ontology/
BRANCH_TAG
:- suffix appended to common "URI namespace" used to constructversionIRI
andrdf:resource
forowl:imports
, e.g.release/202101
overrides the default branch/tag, wherebranch
andtag
are values calculated from the parameters of the git repository containing the Ontology Family use/
to set a blank valueDATADICTIONARY_COLUMNS
:- list of parameters (separated by a vertical bar|
), which are the values of theextract-data-column
options (see OntoViewer Toolkit - Options), used when generating the "data dictionary" product (extract-data
goal - see OntoViewer Toolkit - Goals) e.g.synonym=http://example.com/synonym,https://www.omg.org/spec/Commons/AnnotationVocabulary/synonym|example=http://www.w3.org/2004/02/skos/core#example
ONTPUB_SPEC_HOST
:- the basis of the ontology IRI=https://${ONTPUB_SPEC_HOST}/${ONTPUB_FAMILY}/ontology/
, e.g.onto.example.org
(default:spec.edmcouncil.org
)DEV_SPEC
:- the name of the file (inside the directory named${ONTPUB FAMILY}
containing the "Development" ontology, e.g.AboutFIBODev.rdf
(default:About${ONTPUB_FAMILY}Dev.rdf
)PROD_SPEC
:- the name of the file (inside the directory named${ONTPUB FAMILY}
containing the "Production" ontology, e.g.AboutFIBOProd.rdf
(default:About${ONTPUB_FAMILY}Prod.rdf
)HYGIENE_TEST_PARAMETER_VALUE
:- filter pattern, e.g.example
(default:edmcouncil
); this variable is also possible to set during build time (see--rebuildimage
option)HYGIENE_WARN_INCONSISTENCY_SPEC_FILE_NAME
:- the name of the file (inside the directory named${ONTPUB FAMILY}
for which the "warning" level consistency check test will be performed (i.e. in the case of a lack of consistency, the ontology building process is not terminated), e.g.AboutFIBODev.rdf
(no default - in the absence of a value, the tests will not be run)HYGIENE_ERROR_INCONSISTENCY_SPEC_FILE_NAME
:- the name of the file (inside the directory named${ONTPUB FAMILY}
for which the "error" level consistency check test will be performed (i.e. in the case of a lack of consistency, the ontology building process is terminated with an error message), e.g.AboutFIBOProd.rdf
(no default - in the absence of a value, the tests will not be run)CONSISTENCY_CHECK_TIMEOUT
:- maximum duration of a single consistency check process - at the "warning" or "error" level. After this time (default: 1 hour), the process will be killed (kill -s KILL ...
), an error message will be displayed and thehygiene
stage will continueONTPUB_MERGED_INFIX
:- infix for "merged" files, e.g.-Merged
(no default - in the absence of a value, the merged files will not be created)RDFTOOLKIT_VERSION
:- RDF Toolkit version (default:latest
), e.g.v1.2
; don't setRDFTOOLKIT_VERSION
to uselatest
version
TODO: The text below this point needs to be updated.
The Jenkins server runs at https://jenkins.edmcouncil.org. It uses Github user authentication, so everyone needs to have a Github userid in order to access the Jenkins server. This userid needs to be part of the EDM Council organization on Github.
The server that runs at https://jenkins.edmcouncil.org is the so called "Jenkins Master Server". The idea is that most common jobs will run there, until we need more capacity. In that case we can delegate the work of running Jenkins jobs to "slaves". These are Jenkins-servers that do not have their own GUI, they're installed on any other machine automatically by the Jenkins Master, and simply run Jenkins jobs.
The facility that Jenkins provides to run jobs on a slave, can also be used to run specific jobs on special hardware. That hardware could be hosted elsewhere, for instance at the premises of a vendor in the ontology space, such as a triple store vendor. These vendors could then run the ontology test jobs on their own hardware, all configured and tuned as good as it gets. Whenever a change is pulled into the ontology repository, all sorts of jobs can get triggered on many different machines, validating that change.
-
The
bin/
directory contains scripts and executable tools (like jena) that can be used on your own computer or in a Jenkins job context.