a BytesBuf has a "void*buf" and "int len", but no means to set content about irods_rule_engine_plugin_python

[trel]

two additional options...

1. would it be better/easier to have the first NUL of a string of bytestring not be made so important?
   or
2. have the bytesBuf_t always 'self heal' and make it automatically have the correct length based on its content?

[d-w-moore]

bytesBuf_t appears to already do the right thing, but the difficulty we have in the Python RE plugin is the inability to set its content (absent this modification) If we consider these two implementations of a rule to write from a local file into a data object (adapted from here ):

def py_obj_write_1(args,callback,rei):
    dest_file = args[0]
    with open('/tmp/file.tar.gz', 'rb') as file_:
        binarycontent = (file_.read())
        ret_val = callback.msiDataObjOpen(('objPath={}++++openFlags=O_WRONLY''O_CREAT''O_TRUNC').format( dest_file), 0)
        f = ret_val['arguments'][1]
        callback.msiDataObjWrite(f, binarycontent, len(binarycontent))
        callback.msiDataObjClose(f, 0)

def py_obj_write_2(args,callback,rei):
    dest_file = args[0]
    with open('/tmp/file.tar.gz', 'rb') as file_:
        binarycontent = (file_.read())
        b = BytesBuf()
        b.setbuf(binarycontent, len(binarycontent))
        ret_val = callback.msiDataObjOpen(('objPath={}++++openFlags=O_WRONLY''O_CREAT''O_TRUNC').format( dest_file), 0)
        f = ret_val['arguments'][1]
        callback.msiDataObjWrite(f, b, b.len)
        callback.msiDataObjClose(f, 0)

The first attempts the write with a string as the buffer, which Python can handle even with NULs, but the plugin seems to translate it as const char* . So if there's a NUL in the string, then regardless of specifying the length param in the msiDataObjWrite, the whole buffer never makes it into the microservice. The BytesBuf_t is the proper solution, but the Python plugin even in its latest release doesn't actually let you set the content; it only allows you to instantiate a blank object with no content.

[trel]

Right, so setbuf appears to be the answer.

And the __del__ feels correct as well. The tests will let us know if it creates a double-free anywhere, but I'd be surprised.