subtle "algorithm" definition issue.

[Matthias Andree]

On Sun, 03 Nov 2002, David Relson wrote:

> The statement in bogofilter.h wasn't intended to declare the variable.  It 
> probably should be the enum statement in your change.  However, when I do 
> that, I see that global variable "algorithm" is used in programs 
> bogofilter, bogoutil, configtest, and debugtest.  Bogofilter has at least 
> one other global variable that I see declared in the main routines of 
> several programs, i.e. run_type.

Well, yes.

> This, and the implementation of config.c, leads to the broader questions:
> 
>         Where do global variables belong?
>         Where do variables for configuration options belong?

setup_lists does not need the algorithm. It's only ever called from
main.c, and I have main.c now pass in the good/bad weights, with the
latter fixed to 1.0 and the former to whatever GOOD_BIAS expands to. (We
might consider making GOOD_BIAS a function though.)

That way, the global variable which does definitely not belong into
wordhash.? can be kept out of this.

> The answer that comes to mind first is:
> 
>         create a globals.c to hold run_type, algorithm, etc.
>         use config.c as the home for configuration options.
> 
> I suspect some variables fit both categories (config and global).  I doubt 
> there'll be any code in globals.c, so maybe it shouldn't even 
> exist.  Perhaps we should put them all in config.c.

I believe we have a problem with the interfaces, not with where the
variables are. We should review all interfaces and make sure they are
complete -- global variables usually give the hint that some interface
is not powerful enough.

Passing in a handle for the configuration should be fair enough.

-- 
Matthias Andree