hosted by CEDAR HepForge

ideas for the future

Back to WikiStart.

a new hierarchic HDF backend

  • Priority: 8/10
  • Status: contacted NCSA developers, they wanna help
  • Effort: 5+ person*days
  • Assigned to: GR

And especially one that allows us to map the hierarchical structure of a Tuple to an equivalent one in the HDF5 format, which was especially designed to take care about those things. (Nesting tables in tables)

Currently data to HDF is serialised into a flat table structure.

Streamable Xml Reader

  • Priority 6/10
  • Status: waiting for libxml++ to arrive at 2.17
  • Effort: 1 person * hour
  • Assigned to: WW

The XmlReader can currently only read files, not single lines. This should be changed. Otherwise .xml.gz files cannot be read.

Code Cleanup

  • Priority: 6/10
  • Status: 75 % done
  • Effort: 1 person*days
  • Assigned to: WW

Remove obsolete classes, obsolete functionality.

automatic object mapping

  • Priority: 5/10
  • Status: feasible concept
  • Effort: 8+ person*days
  • Assigned to: GR?

As mentioned above, there is the possibility to use gccxml or sth. like root-reflex for parsing of class data structure, that would make it possible to automagically map a class data structure to a Tuple. See also boost serialiser, if there is anything useful.

read ROOT objects

  • Priority: 4/10
  • Status: feasible concept
  • Effort: 1-3 person*days
  • Assigned to: WW?

Any ROOT TObject should be mappable onto a harvest::Tuple. Implement this, in the RootReader?.

Create scientific linux packages

  • Priority: 4/10
  • Status: First version exists.
  • Effort: 1-5 person*days
  • Assigned to: WW?

Create gentoo ebuild files

  • Priority: 1/10
  • Status: Considering
  • Effort: 1-5 person*days
  • Assigned to: WW?

reworking the TCP socket handling for the <tt>dhtp://</tt>

  • Priority: 2/10
  • Status: ...considering
  • Effort: 5-10 person*days
  • Assigned to: GR

Since the socket communication is a somewhat 'rough hack' it is desirable to put it on a more solid base.

binary writer / reader

  • Priority: 2/10
  • Status: ...considering
  • Effort: 1-3 person*days
  • Assigned to: GR

We want to find some good solution for creating a stream of type-informed binary data from the content of a Tuple, since it could benefit the TCP transmission handling.

The paradigm of the data harvester being "glue code only" stands against this idea, unless we can use e.g. the boost library.

Currently TCP streaming is done via a gzipped text stream. There is some promising library implemetation for this problems.

SQL Backend

  • Priority: 2/10
  • Status: considering
  • Effort: 2-4 person*days
  • Assigned to: WW

An SQL backend (reader and writer) is conceivable. Note that an sqlite backend exists already.

Spreadsheet backend

  • Priority: 1/10
  • Status: considering
  • Effort: 2-4 person*days
  • Assigned to: WW

OpenOffice? spreadsheets should be interfacable as well ...