2007/2/16, Jos van den Oever <<a href="mailto:jvdoever@gmail.com">jvdoever@gmail.com</a>>:<div><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> 2007/2/15, Mikkel Kamstrup Erlandsen <<a href="mailto:mikkel.kamstrup@gmail.com">mikkel.kamstrup@gmail.com</a>>: > 2007/2/15, Joe Shaw <<a href="mailto:joeshaw@novell.com">joeshaw@novell.com</a>>: > > > Hi, > > > > Jos van den Oever wrote: > > > Shouldnt we start sharing parts of the code required to implement > > > wasabi? I'm mainly thinking about a struct in c representing a parsed > > > wasabi query and functions to de- and serialize these from either the > > > xml or the user language. > > > > > > This should be no more than a couple of files that we can put in an > > > xsd svn. Requirements to make it palletable for most: few deps and > > > written in c. For C++ fans we can add a small wrapper. > > > > > > What do you think? > > > > Definitely.  For the spec to be useful (particularly the query language > > part) we need either a reference implementation or a conformance test > > (preferably both). > > > Agreed. I have been thinking about doing something in Python, but held back > a little to anticipate where a hand was needed (fx. glib bindings). However > several people have already showed interest in glib bindings so I figure I > can do something else (sorry qt guys - you qouldn't like my c++ :-D). > > What requirements would a reference implementation have? I'm thinking > something like 30-50 files with a priori known metadata hardcoded into an > "index" and small wasabi-compliant "search engine" that searches this index. > To go with the bundle there could be a collection of queries with matching > results. Given this it would be trivial to write a Python script that did a > conformance test (on a search engine that had indexed the given files - and > only them). I was thinking about a small c file that implements serializing and deserializing the query language in xml and user language. So no glib and no Qt. That way we can all use it. Starting the complete search is too much. Let's see if we can pull even serializing and deserializing off decently. Emphasis should be on the right struct / class for this.</blockquote><div> I agree - a fast reference implementation in C could be a good idea.  We are going to need a sax parser though. I don't know how big the glib one is - maybe we can rip it out... Well, also if we store (query,result) tuples beforehand, then there wont be any big need for a real service. However a Python script that "unit tests" the service against the known metadata would be useful (or s/Python/Java/ - - where have I seen that before Jos :-D). </div> Cheers, Mikkel </div>