Wrapping new tools ================== This section serves as a reference for wrapping new tools as they are added to BEDTools. Example program description --------------------------- Let's assume we would like to wrap a new program, appropriately named `newProgramBed`. Its signature from the command line is `newProgramBed -a -b [options]`, and it accepts `-a stdin` to indicate data is being piped to it:: newProgramBed -a -b [options] Method name ----------- Generally, I've tried to keep method names as similar as possible to BEDTools programs while still being PEP8-compliant. The trailing 'Bed' is usually removed from the program name. So here the name would probably be `new_program`. Define a method in :class:`BedTool` ----------------------------------- Define a method in :class:`BedTool` . . . and *don't add any content to the function body*. This is because the decorator we're about to add will replace the method wholesale; anything that's in the function body will effectively be ignored. :: def new_program(self): pass Add the :func:`_wraps` decorator -------------------------------- This is where most of the work happens. Since most of the work of wrapping BEDTools programs needs to happen every time a new program is wrapped, this work is abstracted out into the :func:`_wraps` function. .. note:: The :func:`_wraps` docstring and source is the best place to learn the details on what it's doing; here we'll focus on using it. Our hypothetical program, `newProgramBed`, takes `-a` as the first input. We'd like to have `-a` implicitly be passed as whatever our :class:`BedTool` already points to, so we use the `implicit='a'` kwarg to :func:`_wraps` here. `newProgramBed` also takes a second input, `-b`. We describe that to the wrapper with the `other='b'` kwarg. Any other keyword args that are used when calling the method will automatically be passed to the program. So if `newProgramBed` has an optional `-s` argument, we don't need to specify that here. When the user passes an `s=True` kwarg, it will be passed automatically to `newProgramBed` as the argument `-s`. If `newProgramBed` does not accept a `-z` argument but the user passes one anyway, we rely on the BEDTools program to do the error-checking of arguments and report any errors back to Python. Here's what the new method looks like so far: :: @_wraps(prog='newProgramBed', implicit='a', other='b') def new_program(self): pass For wrapped programs that expect a genome file or have more complex arguments, see the docstring and source for :func:`_wrap`. Add doctests ------------ While the function body will be replaced wholesale by the decorator, the docstring will be copied to the new function. This is important because it means we can write meaningful documentation and, even more importantly, doctests for this method. Writing a doctest within the method's docstring means it will automatically be found by the test suite. :: @_wraps(prog='newProgramBed', implicit='a', other='b') def new_program(self): """ Converts all features to length of 1. Example usage: >>> a = pybedtools.example_bedtool('a.bed') >>> b = pybedtools.example_bedtool('b.bed') >>> c = a.new_program(b, s=True) >>> print c #+NORMALIZE_WHITESPACE chr1 1 2 chr1 100 101 chr1 150 151 chr1 900 901 """ Add to list of known programs ----------------------------- The last thing to do is to add the new program to the end of the tuple `pybedtools.helpers._prog_names`. This creates rudimentary security by only allowing these programs to be called, and acts as sort of a central registry for programs that have been wrapped. Summary ------- That's it! We now have a method, :meth:`BedTool.new_program`, that wraps a hypothetical `newProgramBed` BEDTools program, will accept any optional args that `newProgramBed` does, will return a new :class:`BedTool` containing the results, *and it's tested*. This new method can be be chained with other :class:`BedTool` instances, used as an iterator or generator, or anything else a normal :class:`BedTool` can do . . . for example:: a = pybedtools.example_bed('a.bed') b = pybedtools.example_bed('b.bed') c = a.new_program(b, s=True).filter(lambda x: x.start < 125).saveas('t.bed', trackline='track name="one-bp features"') .. _decorator: http://www.python.org/dev/peps/pep-0318/