Set IDs are lost during writing

[minhlab]

I find this behaviour counter-intuitive: if you read a corpus (using Corpus.from_file) and write it out right away (using write_to_file), all set IDs are lost, i.e. the last column contains only minus signs.

I attached a test script and sample document.

test.zip