I have two plugin skeletons now but need to fix a lot of details. Right now they are implemented together as one plugin copied from the existing classifier and another
cleaner one with no obvious leftover junk. The latter is the one I wanted to finish developing and it currently imports 2 shared objects that implement classification and visualization functions. They ignore the qiime2 input and output parameters
and right now make use of files. For early testing, I wanted to just copy
the qiime2 semantic types to files and then read those into my c++ code
as this avoids dealing with the python interface and automtically creates debug
information. However, I did want to put them some where suitable for others to
play with should that seem reasonable.What should a plugin developer
do for file usage? I did note you make a lot of use of /tmp which I guess
is a good place but not sure how you generate the names etc.
Another benefit of this may be memory conservation although I’m not
sure I can free an input object after writing it to a temp file.
I just took the code and generated a bunch of svg files showing relaive fits
between my candidates and those apparently from a qiime1 pipeline and
digging into it now. While this is not using qiime2 right now, it is the same stuff
I have called from my plugins.