specifications-AA_GLOBAL

Global definitions, documents and scripts for openEHR

How to set up the Asciidoctor publishing environment for the specifications

Tool via Docker

Asciidoctor can be used as a tool via Docker with building provided Dockerfile.

Building image:

docker build -t openehr/asciidoctor .

Instantiating container, mounting the parent directory (assuming that is containing all openEHR specifications repos):

cd ..
docker run --rm -it -u $(id -u):$(id -g) -v "$(pwd):/documents/" openehr/asciidoctor bash

Running asciidoctor inside the container:

./specifications-AA_GLOBAL/bin/spec_publish.sh -f -r -v -t -q -l Release-1.0.1 QUERY
./specifications-AA_GLOBAL/bin/spec_publish.sh -f -r -v -t -q -l development RM

Running asciidoctor as a tool (considering that the docker CMD is defined spec_publish.sh -f -r -v -t -q -l):

docker run -u $(id -u):$(id -g) -v "$(pwd):/documents/" openehr/asciidoctor Release-1.0.1 QUERY
docker run -u $(id -u):$(id -g) -v "$(pwd):/documents/" openehr/asciidoctor development RM

Manual Tool installation

Note	a Linux-like environment is assumed; cygwin under Windows will work fine.

Firstly, you need to install asciidoctor. This is generally painless - see here. Install also few other required things:

• gems: asciidoctor-diagram, asciidoctor-diagram-plantuml, asccidoctor-bibtext, asciidoctor-tabs and pygments.rb

• python

• jq (and bc)

Secondly you may need MagicDraw, since UML diagrams and definitions are extracted from the UML binary files via MagicDraw. Please contact the openEHR SEC if this is the case.

Installing MagicDraw is relatively painless on Linux, despite not being PPA-managed. instructions).

If you installed MagicDraw, you will also need to install the openEHR UML extractor plugin.

Git repo setup

Now you need to clone the openEHR specifications repos, as follows:

1. Create a directory openEHR-specifications or similar to contain the specifications git repo clones

2. Download the bash script /bin/setup_openehr_git.sh into that directory and run it (needs to be Linux, cygwin, or other unix-like environment).

You are now ready to go. From the parent directory, in a bash shell, type:

$ ./do_spec_publish -r RM

This will publish the HTML documents for the RM component in specifications-RM/docs. You can do the same for any other component, e.g. BASE, QUERY etc.

To find out what else you can do, type:

$ ./do_spec_publish -h

Making changes.

Text

Some of the typos (or other text improvements) are actually in the UML text, which is inside the tool. Other improvements are to text that is in the .adoc source files. However, the UML class definitions are also extracted to .adoc files, so it looks like everything is just .adoc files that could easily be updated. But if we apply a typo fix to one of the extracted adoc files (in /docs/UML/classes) it will be overwritten the next time the MD is used, and a new extraction is run. So some of the typos have to be done inside MD, while others can be done directly on adoc files not in that directroy

Whenever you want to publish changes made over time at some later date, re-run ./setup_openehr_git.sh, which will now perform git pull commands on the relevant repositories, and then run ./do_spec_publish.

UML

Ask the SEC how to do this using MAGICDRAW

How to do a Release of a Component

Components, i.e. RM, AM, LANG, PROC etc are the units of release. Each component has its own Github repo. A release is always named either:

• Release-N.N.N, where N.N.N is a semver-style 3-part release id, e.g. 1.2.14, OR

• Release-N.N.NvN, where the trailing vN identifies a new version of the existing release Release-N.N.N (see below)

To do a release of a component, e.g. 'RM Release-1.0,4', the following steps are required.

Jira

When all the work specified by the CRs in the component project (e.g. SPECRM) in Jira is done, resolve and close all Jira CRs and PRs. You can now create the relevant saved searches for the PRs and the CRs for this release in Jira, which will have URLs of the form <jira-home>/projects/SPECPR/versions/NNNNN and <jira-home>/projects/SPECR/versions/NNNNN respectively.

Git work area

In the Git clone work area of the relevant component (e.g. specifications-RM):

• make sure you have done every last change to the specifications files;

• In the manifest.json file, ensure:

  • the specifications[@id=<spec>]/spec_status value is correct for each <spec> in the component (e.g. this release may be the one that promotes a specification from TRIAL to STABLE) (example);

  • there is a correct entry under releases for this release - in particular, check the Jira links are correct. It should receive a date (i.e. the current date), and (when necessary) a new cooking release should be prepended by leaving the date empty.

• create a new branch with the name of the release, e.g. Release-1.0.4 and check it out. The changes from now will be on that branch, which becomes that releas’s maintenance branch.

• now run the publisher script in release mode, e.g. ./do_spec_publish -r -l Release-1.0.4 (as many times as needed) and check:

  • the HTML output front page and preface page of each specification document;

• commit all the changes locally;

• add an annotated Git tag with the release id e.g. Release-1.0.4, and a short description, usually of the form "RM Release 1.0.4", but may have more text if needed;

• push the changes to Github.

• don’t forget you are on a branch, for later work, you probably want to go back to master.

The last push will cause a normal commit at GitHub, and the webhook will cause the specifications.openEHR.org server to do a pull, and then a copy out of a) the latest commit, and b) also of any tagged commit not already copied out, to two different directories under the /var/www/vhosts/openehr.org/specifications.openehr.org/releases/COMPONENT area, namely Release-N.N.N and latest. If you just did a release, these are the same, but otherwise, usually just the latest is updated. These copy-outs are what enables the specifications.openehr.org to serve both the latest working version and also all previous releases of any spec.

Fixing a Release

Sometimes all the above work is performed, and then someone discovers a significant typo or other error in the newly released specifications (including diagrams etc). Fixing things is easy. Do the following in the Git work area:

• check out the relevant release branch (see above); any changes you now make are on that branch.

• make the changes you need to the source files;

• republish using ./do_spec_publish with the original release id, e.g. ./do_spec_publish -r -l Release-1.0.4;

• commit the changes;

• add a new annotated tag with the id Release-N.N.NvN+1, e.g. Release-1.0.4v1; the final number must be 1 higher than the previous fix release, so it could easily be Release-1.0.4v7 if a number of errors were discovered over time;

• push the new commit(s); the webhook-driven server scripts will over-write the last copy-out of the release with this new one;

• don’t forget to merge the changes across into the component’s Git repo master branch as well, if they are not already taken care of by changes there.