Release packages moving forward

[enridaga]

Hi all, since we are preparing the v1.0.0 release (#538), I would like to discuss possible changes to the way we release sparql anything. At the moment, apart from releasing the zipped code archives, we include:

• sparql-anything-.jar -- all dependencies + command line tool
• sparql-anythin g--fuseki.jar -- fuseki server
• artefacts published on Maven.
• docker image

There have been discussions regarding the size of jars, which often include dependencies that are only needed for corner cases (see #484 #513).

I also want to mention that PySPARQL-Anything testing, upgrade, and release follow any SA Java release; we can discuss that as well, but consider it a separate workflow.

One way would be to have a light, mid, and full version by including only some formats. E.g.

• light: CLI + csv, json, xml, html (without headless browser), txt, file system, Markdown
• medium: CLI + csv, json, xml, html (without headless browser), OOxmls (word, excel, etc....)
• full: CLI + FUSEKI + all features

(please note that I am not sure how to make a single binary for both CLI and Fuseki, but that's a different issue).

Opinions? Would anyone want a different way to package and release SA?

[MathiasVDA]

Docker :)

[enridaga]

Indeed, I edited the post!

[justin2004]

docker +1

i mostly use jena/fuseki via:
https://github.com/stain/jena-docker

and in a similar fashion i mostly use sparql anything by building and running this image:
https://github.com/SPARQL-Anything/sparql.anything/blob/v1.0-DEV/BROWSER.md#headless-browser

[enridaga]

There are two Docker files, one for the automated release (by @MathiasVDA) and another that @justin2004 made for setting up a server with the headless browser. I wonder whether we could publish both automatically.

[enridaga]

See https://github.com/SPARQL-Anything/sparql.anything/blob/v1.0-DEV/.github/workflows/docker-image.yml

[MathiasVDA]

Publishing both shouldn't be a problem. First choice to make is if you want to have the same docker image name but different tags, or just different docker image names. I would advise the latter so that the tags are not to confusing.

I actually can't find the docker image on dockerhub that is published by the github action. And I can't find a link in any of the markdown files. What did you put in the settings as repository name?

[enridaga]

The image is published here: https://hub.docker.com/r/facadex/sparql-anything

I agree on using different names

[enridaga]

Does anyone want to test the release? https://github.com/SPARQL-Anything/sparql.anything/releases/tag/v1.0.0