[Debtags-devel] tdb

[Enrico Zini]

--HcAYCG3uE/tztfnV
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline

On Sun, Apr 10, 2005 at 03:48:31PM -0700, Erich Schubert wrote:

> Hi Debtaggers, Hi Enrico,
> Enrico, you wrote that you adopted my TDB code, but had to change some
> of my structures for your needs. Can you give me a short overview of
> the differences, so I can see if they fit my needs, so I could switch
> over to your database format?
> Sorry, I won't have much time to work on debtags right now. :-(

No problem.

I have this:

  Item -> Tags mapping:

    Key: "Idebtags"  ("I" + package name)
    Value: "implemented-in::c++\0suite::debian"  (\0 separated list of
                                                  tag full names)

  Tag -> Items mapping:

    Key: "Tsuite::debian"  ("T" + tag name)
    Value: "debtags\0apt"  (\0 separated list of package names)

  Untagged packages:

    Key: "T"
    Value: (\0 separated list of package names)

That way I can get the tags for a package, the packages that have a tag,
and I can easily get the packages that have a certain tagset by
intersecting the item sets of every tag in the tagset.

I created 3 classes to handle this:

   TDBDiskIndex

      does read and write on disk

   TDBIndexer

      does everything in memory, and writes an index to disk on request
      (used during debtag update: it's much faster working to memory,
      and the resulting tdb file is more compact since it's written once
      in a single pass and never updated)

   TDBReadonlyDiskIndex

      reads from a TDB index on disk, and keeps changes in an tag patch
      (so handy to implement debtags-edit on top of this!  The index is
      owned by root, and debtags-edit only reads it, writing to a
      separate patch file that can be directly submitted to the central
      database!)

I started working in swig bindings for libdebtags, but it seems there's
a lot of work to do (hoping that std::set is iterable at all after
wrapping: so far it doesn't seem to be, I asked in the swig list, hope
for good clues).


Ciao,

Enrico

--
GPG key: 1024D/797EBFAB 2000-12-05 Enrico Zini <enrico@enricozini.org>

--HcAYCG3uE/tztfnV
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: Digital signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)

iD8DBQFCWbnd9LSwzHl+v6sRAqjXAJ9jWsUE9geRMdEF++oJPxlBuSx2eACfbxw7
8G9X3/AT8tFX9LBTUDgpzjs=
=RLIY
-----END PGP SIGNATURE-----

--HcAYCG3uE/tztfnV--