Strategy behind the new generation CCP4 software library
Author: Martyn Winn
- C-level access to crystallographic functionality.
- Access to Fortran libraries from C is possible (see e.g.
Solomon code), but preferably avoided.
- Support for scripting.
- Scripting is central to ccp4i, ccp4mg and automation efforts.
- Support for legacy code.
- Essential to retain existing users and collaborating developers.
- Conformance with community-accepted data model.
- CCP4 library implements an implicit data model. This model needs
to be made explicit, and should conform as closely as is practical
to the EBI-led data model.
- Unit cell conversions
- CCP4 style
- System-dependent functions.
- CMTZ,CMAP used by 3rd parties to access CCP4 formats
- MMDB used in ccp4mg etc.
- Support of CCP4 suite
- Optional back-ends to clipper
- Provide CCP4 environment to clipper applications
Here is the corresponding Use Case diagram:
Form of library
- The library should compile as a whole to give a libccp4
which supports the current CCP4 release. To support other target uses
listed above, it is necessary to allow partial compilation of subsets
of the library. For example, compilation of CMTZ, CSYM and the low-level
libraries should be sufficient to support 3rd party applications wishing
to access MTZ files.
- Along with the source files, the library depends on a number of
- syminfo.lib holds spacegroup information
- atomsf.lib contains form factors
- cif_mm.dic is the mmCIF dictionary
- crossec.lib currently used by CROSSEC - move to library?
- The library has the following dependencies:
Future dependencies might include:
- libccif - low-level i/o for mmCIF
Python will be used in:
- CCP4 Molecular Graphics project
- CCP4i run scripts, replacing current tcl run scripts in suitable cases
- cctbx if this is to be a dependency
CCP4 is committed to the use of python, but the question of which version
is still open.
Arguments in favour of 1.5.2
Arguments in favour of 2.2.1
- Increased likelihood of user systems having a suitable version of python.
- CCP4mg aims to run with Python 1.5 (Stuart reports that CCP4mg also
runs with 2.1 and 2.2).
- We can migrate to Python2 at a later date.
- Integration with cctbx.
- Some features of 1.5 are deprecated in Python 2 (but will presumably
- Nick Sauter lists the following useful features of Python 2:
- C-speed support for regular expressions
- comparison operators can work on array types
- cross-platform support for pipes
- much better install procedures for third party add-ons
- new language features like iterators and generators
All legacy applications should work with the new libraries via the
Fortran interface. New and/or converted applications include:
- Pete's ccp4mapwish has been converted to use the
cmap part of the new library. The source is currently
in $CPROG/ccp4mapwish_ under the CVS tag "newlib-dev".
- Pete's mtzdiff, mapdiff and pdbdiff.
- Eugene's pdbcur and ncont use CCP4 style functions
as well as MMDB.
- Martyn's reindex uses/will use cmtz and csym functionality.
- Test support of existing CCP4 suite.
- Release CCP4 based on new library.
- Lock APIs at some point.
- Convert other sections of library (see "planned" above).
- Write C applications.
- Write Tcl applications.
- Integrate with clipper applications.
Possible future inclusions:
- Reading of atomsf.lib and crossec.lib (or updated versions
of these files).