Re: Loading database from string (C API)

On 12/12/2013 01:08 PM, Edd Barrett wrote:
> Hello again,
>
> On Wed, Dec 11, 2013 at 03:59:09PM +0100, Jan Wielemaker wrote:
>> On 12/11/2013 03:41 PM, Edd Barrett wrote:
>>
>> Search for blob.  Basically, if PL_term_type() gives PL_ATOM, it
>> may be a blob or a real atom.  Use PL_is_atom() to find out.
>>
>
> With this workaround applied I am now at the point where I am calling
> load_files/2, as suggested earlier in this thread. Sadly something
> is amiss; ideally I would have a C program to demonstrate, but this
> would be a fairly large piece of code and I would have to translate it
> from Python CFFI to C. Before I go there let's see if anything obvious
> springs out.
>
> So I have constructed the predicate and arguments needed to call
> load_files(pl_source, [stream(pl_stream)]), where pl_source is an atom to
> identify the code in the stream (as described in the docs) and pl_stream is
> the stream opened using the open_stream C function that we have already
> discussed. For testing purposes I have passed the theory "f(1)." to the
> function.
>
> The call to load_files/2 (in Python/CFFI) looks like this:
> pl_qid = self.libpl.PL_open_query(self.ffi.NULL, self.libpl.PL_CATCH_Q_CATCH_EXCEPTION, pl_pred, pl_args)
>
> pl_pred is a the result of a PL_predicate() call for load_files of arity
> 2. pl_args is an 'array'(?) of term refs as given by PL_new_term_refs().
> To check these are good I can call PL_get_chars with CVT_WRITE_CANONICAL:
>
> Arg 1: swithon_theory -- arbitrary name chosen for the stream
> Arg 2: '.'(stream(<stream>(0x18ebfd824c00)),[]) -- list with one element, the stream
>
> Looks OK. The qid returned looks valid too. The call to open_query is
> followed by: pl_sol = self.libpl.PL_next_solution(pl_qid);
>
> At this point we use PL_exception() to check for an exception etc.
>
> The problem is execution never gets past the call to PL_next_solution.
> The interpreter seems to be stuck in a loop printing:
>
> ---8<---
> ERROR: Syntax error: Unexpected end of clause
> ERROR:
> ERROR: ** here **
> ERROR: .
> ERROR: Syntax error: Unexpected end of clause
> ERROR:
> ERROR: ** here **
> ERROR: .
> ERROR: Syntax error: Unexpected end of clause
> ...
> ---8<---
>
> Stuff I have tried:
>   * Remove dot on the end of the theory string -- same error.
>   * Varying the theory string in other ways -- no change.
>   * Try different PL_Q_* -- no difference.
>   * Change character encoding on the stream -- same.
>
> Is that expected behaviour?

I guess not.  I'd replace the call to load_files/2 by something else and 
then dump the content of the stream.  E.g.

test_files(_Id, [stream(Stream)]) :-
	copy_stream_data(Stream, user_output),
	flush_output(user_output).

That should print f(1). to the user output.  Most likely it will not
do that.  The precise output might give you a hint what is wrong.

I'm quite sure that load_files/2 this way is ok.  You can of course
also test that by loading a real file from a stream by opening the
file yourself.

	Success --- Jan