view bindings/python/README.txt @ 693:b1723ae7675e

begin work on sampling API

This is motivated by the need to be able to sample with arbitrary feature data
(e.g. from a feature file) against a database, for the JNMR "collections" paper
revisions or possible ISMIR paper revisions.  That bit doesn't work yet, but
the C-ified version of the current functionality (sample db x db and
sample key x db) works to the level of anecdotal tests.

The general approach is to mirror the _query_spec() API, where a whole heap
of knobs and twiddles are available to the user.  Unlike in the _query_spec()
API, not quite all of the knobs make sense (and even fewer are actually
implemented), but the basic idea is the same.

I pity the poor chump who will have to document all this.

README.txt

to install, try:

python setup.py build
python setup.py install

Notes: 
 - a compatible build of audioDB (>=r914) needs to be linkable at runtime
 - currently only the direct C-api exposure layer is visible, so the calling semantics are a bit gross


The actual query call is a bit of a mess, but will be more intuitive from the native python layer (to be written)...
so the python bindings now have a complete path:
	>>import _pyadb
	>>aDB = _pyadb._pyadb_create("test.adb", 0,0,0)
	>>_pyadb._pyadb_status(aDB)
	>>_pyadb._pyadb_insertFromFile(aDB, "someFeats.mfcc12")
		...(add some more data)
	>>result = _pyadb._pyadb_queryFromKey(aDB, "a Key in aDB", [options])
	
	and then result has a nice dict of your results.  



21 September 2009, Ben Fields, [e-mail]