osma
commented
1 year ago Thanks for opening the issue @kinow ! More specifically, the problem is that ARQ collation doesn't work with the Fuseki Docker image, apparently due to lack of locale support (though the underlying issue might be different or more complicated than just lack of locales). Copying my analysis from the relevant Skosmos issue:
I tested this with the example query from the Jena ARQ Collation documentation. I put this in collation.rq:
PREFIX arq: <http://jena.apache.org/ARQ/function#>
SELECT ?label WHERE {
VALUES ?label { "tsahurin kieli"@fi "tšekin kieli"@fi "tulun kieli"@fi "töyhtöhyyppä"@fi }
}
ORDER BY arq:collation("fi", ?label)then ran it against the Fuseki Docker container using rsparql (from Jena) and the result was this:
$ rsparql --query collation.rq --service http://localhost:9030/skosmos/sparql
-----------------------
| label |
=======================
| "töyhtöhyyppä"@fi |
| "tsahurin kieli"@fi |
| "tšekin kieli"@fi |
| "tulun kieli"@fi |
-----------------------This is not the correct, locale-aware order; tsahurin kieli should be first and töyhtöhyyppä last. So I think that the Fuseki container configuration should be modified to enable support for locale-aware collation.
afs
kinow
Version
main
Feature
The docker image provided with Apache Jena code uses a multi-stage build, where the last layer (used to run it) uses Alpine Linux.
In Skosmos we are upgrading the Docker setup to switch from stian's Jena image to the one included in Apache Jena code, and also to run CICD tests with the Apache Jena image.
We ran into an issue now where we are not able to use multiple locales in PHP due to the missing
localecommand. Looking at the locales available for Java and Jena [^1], I get:I believe we will be able to install extra locales in the Skosmos container, but perhaps that should be included in the Apache Jena Docker image as well, allowing users that rely on other locales (for ARQ query collation, for example) to use the container with the desired locale.
If that sounds like a good idea, we can send a pull request upstream with our solution for the locales.
As for implementation details, it could be i) a fixed list of common locales, ii) just English but with the necessary tools and configuration to easily add more locales, or iii) something more elaborate like trying to download all locales but enable selectively via some env var (no idea if that's actually doable). I think the simplest would be ii), and then in Skosmos we would add an extra step to install finnish, swedish, german, etc.
[^1] Compiled the program from this oracle docs page with
-source 8 -target 8inside the Alpine containerAre you interested in contributing a solution yourself?
Yes