Files and carml-jar

[ktk]

There is a new cli tool for carml available: https://github.com/carml/carml-jar

I could get it to work with XRM carml output but I had to figure out how to do it properly. With the current configuration I can only get it to work with piping the file as a stream and I have to do the XRM source like this: source "". This generates:

<#MapCanton>
    a rr:TriplesMap;

    rml:logicalSource [
        rml:source [
            a carml:Stream;
            carml:streamName "";
            carml:declaresNamespace [
                carml:namespacePrefix "eCH-0071";
                carml:namespaceName "http://www.ech.ch/xmlns/eCH-0071/1"
            ]
        ];
        rml:referenceFormulation ql:XPath;
        rml:iterator "/eCH-0071:nomenclature/cantons/canton"
    ];

I can then run: cat input/eCH0071.xml| java -jar bin/carml-jar-1.0.0-SNAPSHOT-0.4.4.jar map -m src-gen/mapping.carml.ttl

However, I cannot run it directly from a file, I always have to pipe it. So I checked out the carml docs and it says the name can be left out or empty, so that seems to be correct.

So I checked out the ontology and I found stream:url, so I changed the generated output to this (manually):

<#MapCanton>
    a rr:TriplesMap;

    rml:logicalSource [
        rml:source [
            carml:url "eCH0071.xml"; # cannot have a carml:Stream in this case
            carml:declaresNamespace [
                carml:namespacePrefix "eCH-0071";
                carml:namespaceName "http://www.ech.ch/xmlns/eCH-0071/1"
            ]
        ];
        rml:referenceFormulation ql:XPath;
        rml:iterator "/eCH-0071:nomenclature/cantons/canton"
    ];

And then I can run it directly from file:

java -jar bin/carml-jar-1.0.0-SNAPSHOT-0.4.4.jar map -m src-gen -rsl input -of ttl -P

This probably would work in XRM via rml output but I use the XML Namespace extension so I have to stick with carml-syntax.

Not sure how to handle that properly in XRM though, would be nice if both options could be generated.

[mchlrch]

I understand this as follows: carml requires different rml:logicalSource definitions, depending on if either a stream source or a files source should be used.

For stream sources:

    rml:logicalSource [
        rml:source [
            a carml:Stream;
            carml:streamName "";
            ...
        ];
        ...
    ];

For file sources:

    rml:logicalSource [
        rml:source [
            carml:url "eCH0071.xml";
            ...
        ];
        ...
    ];

Using output carml in xrm version 1.2.0 generates the "stream sources" definition.

The new requirement is now, that both variations can be generated from xrm. Is that correct @ktk ?

[ktk]

Correct interpretation.

It would be nice as people can use XRM and create carml output without having to interact with streams that way. For those that do not necessarily use a barnard59 pipeline, that's probably a nice thing to have.

[ktk]

Ran into this again btw. I think for pure carml-cli mode, this feature would be useful. So either have an anonymous (empty) stream or a filename with carml:url