Writer emits 'class="empty set"' which is not accepted by published SBGN-ML schemas

[pklemmer]

Following #12 / #13, Glyph.prototype.buildJsObj on develop writes the new name:

// src/libsbgn.js:756 (develop)
attributes.class = (this.class_ === 'source and sink') ? 'empty set' : this.class_;

The lenient reader (src/libsbgn.js:850) is fine and forward-compatible. The problem is the writer: the rename it depends on (#62) is still open, and no published schema has been updated. empty set / emptyset appears zero times in resources/SBGN.xsd or
documentation/differences-to-M2/SBGN-v0.2.xsd on either master or develop; both still enumerate only source and sink.

Effect: every file libsbgn.js writes containing a source/sink glyph is schema-invalid.

$ xmllint --schema SBGN-v0.2.xsd map.sbgn --noout
Element '{http://sbgn.org/libsbgn/0.2}glyph', attribute 'class': [facet 'enumeration']
The value 'empty set' is not an element of the set {..., 'source and sink', ...}

Newt bundles libsbgn.js, and sbgnviz.js correctly maps its internal empty set to source and sink before serialising — which this line then reverses, so the fix upstream of it has no effect. The resulting files are rejected by other SBGN-ML consumers; MINERVA's importer returns HTTP 500 on them.

Maybe revert the writer to emit source and sink until #62 is resolved and a schema enumerating the new value is published.

[ugurdogrusoz]

So, resolving #62 should fix the problem?

[pklemmer]

From what I can tell, yes. I'm no expert in JS but it seems that libsbgn.js writes 'empty set' instead of 'source and sink', but the SBGN-ML schema only takes in 'source and sink' and the Java libSBGN org.sbgn.GlyphClazz does not accept 'empty set' either. So there seems to be a disconnection between libsbgn.js and Java libSBGN, which i think #62 should fix. This is also, I think, why Newt (which uses libsbgn.js) can use 'empty set' during reading, but e.g. cd2sbgn and MINERVA (which use Java libSBGN) fail to read files with class 'empty set'.