On Tuesday, August 26, 2003, 8:57:49 AM, jbritt wrote:
> Last night I hacked around on rimport and ri, and found that, with some
> effort, the ri method index file can be updated with new class and
> method names when a new dat file is added.
> It's tedious, though, and of course thiere is no provision for
> distinguishnig among library source (built-in, standard distro, etc).
> For rj, do we want to continue using the ri format and data set as a
> base, or perhaps use the general approach (marshalled data structures
> for fast access) but define our own file and indexing hierarchy?
Part of the rj suite is a program 'rj-mkidx.rb' which is given a
directory, slurps all the data files therein, and spits out an
"index" file. So rj currently has no problem with indexing, and I
have no problem with the data format in general. We could
investigate using subdirectories to model module/class
structures, though. I suspect that would bring some benefit.
Gavin
|
