StreamDocumentSource buffer can be big about owlapi HOT 2 CLOSED

Yeah, it's because of the multiple parsing attempts... the stream might need to be read more than once. It might be better if an ontology loader was instantiated for each load and then this could decide whether it needed to do buffering. If only one parser factory is registered (RDF/XML say) then the buffering is unnecessary.

from owlapi.

An alternative would be to buffer only what is actually read and enable rewinding the stream, so reading from the actual stream would only happen once. The first bytes read might also be discarded after all parsers have examined them - although this is harder to get right.

Should still be faster than the current implementation, as most parsers will fail very early while reading the input - now no parser will be able to even start before the whole stream has been read. Especially bad if someone passes a big file that does not match any parser :-P

There is a possibility to start parsing with all parsers at once and only read the next chunk of input once all parsers have consumed the previous chunk or failed to parse, but this is more complex to implement and for small ontologies might cost a lot - all the parsers going at once might take longer than just loading the whole input and trying them one at a time.

from owlapi.