Is it possible to unset output-format-type?

[polm]

I would like to specify a custom format string even if the user has an output-format-type specified in their mecabrc. As described in taku910/mecab#41, the way to do this on the command line is to give an argument like -O "". So I tried this:

tagger = MeCab.Tagger('-O ""')

Unfortunately this results in a RuntimeError. I tried escaping the quotes without effect. I guess something about the way the argument string is passed to the Mecab library is different than its own internal argument handling?

For reference, the reason I want to specify a format string is explained in explosion/spaCy#3491.

[zackw]

Unfortunately, the SWIG interface definition file that we use only exposes the C++-level Tagger constructor that takes a single string containing command-line options. That constructor doesn't implement any shell quotation syntax; it blindly splits its input on whitespace. It is therefore not possible to set the -O option's argument to the empty string using this constructor.

There is another constructor that takes an array of command-line options. If we exposed that constructor, you could write something like

tagger = MeCab.Tagger(["-O", ""])

and it would work. Unfortunately, I do not understand SWIG well enough to know how to make that change. If you can figure it out I would be glad to merge your patch. (Please note that both the Tagger and Model classes should be changed.)