Technical Information about EMBOSS for Windows

Principles

The seemingly easiest solution for porting EMBOSS from Unix to Windows is to rely on some Unix compatibility layer, such as Cygwin, DJGPP, PW32 or UWIN. However, this approach is only fully justified when porting applications that interact a lot with the Unix OS, or with the user through the X-Window system.

But EMBOSS is much more a computation engine than an interactive device. So it seems more appropriate to build a native Windows EMBOSS: performance should then be optimal, and it should be easier to build Windows-specific software based on the package, e.g. GUIs.

EMBOSS source code modifications

Getting the ajax and nucleus libraries to compile on Windows requires only minor modifications to the source code. Starting from the original 2.7.1 source code, all the alterations I applied are bracketed by #ifdef WIN32...#endif preprocessor directives, so that everything should still compile on Unix.

Changes concern mainly:

Directory handling: Since the Windows directory handling API is quite different from Unix's (opendir(), readdir(), and other functions from <dirent.h>), I had to provide a custom directory handling API (dirent_w32.[hc]). This is drawn directly from PW32 and DJGPP.
EMBOSS configuration: Unix EMBOSS is quite flexible about locating installed files and configuration files (e.g. .embossrc). I have simplified function ajNamInit for the Windows case. Here, a unique environment variable, EMBOSSWIN, points to the EMBOSS directory, where the software should find everything it needs, specially the emboss.ini configuration file.
Windows sockets: Microsoft's socket implementation differs slightly from BSD sockets. I had to take that into account in ajseqdb.c.
USA parsing: I made a simple but critical change to one regex in ajseqread.c so as to allow full Windows file names (e.g. C:\toto\tata) in USAs.
Sorting in embDbiSortFile(): This function calls Unix sort(1) to sort and merge files. I had trouble getting Windows sort behave as embDbiSortFile() expects, so I downloaded UnxUtils, a collection of ports of GNU utilities to Windows. This EMBOSS build uses UnxUtils's sort, which must be located in the EMBOSSWIN directory.
System functions: A few system functions in ajsys.c are excluded from the build.

To the best of my knowledge, the only functionality removed from this build is the ability to launch sequence-producing programs through piped USAs (e.g 'getz ...|').

Graphics library

Version 5.1.0 of the PLPLOT library comes with a Windows driver and a makefile for Visual C++. I've had no trouble compiling it, but substituting it for version 4.99, which is used in Unix EMBOSS, has been a somewhat hazardous process.

I changed the default graphic device from 'x11' to 'win3'.

Source distribution

The source distribution is based on the reference (Unix) source distribution v. 2.10.0, with the following additions:

Source code changes as described above.
PLPLOT v. 5.1.0 source distribution.
Visual C++ 7.1 'solutions' for the ajax and nucleus DLLs in a DLLs directory, and for the individual applications in an apps directory.