[Freeciv-Dev] Re: [RFC PATCH] init_techs

[Justin Moore <justin@xxxxxxxxxxx>]

> >    I posted a perl-like split a little while back, but got very little
> > response other than, "Boy, that's kind of neat."  Is there interest in
> > someone (or me) rewriting some of the parsing code?  I don't think I have
> > anything better to do this weekend. :)
>
> I would really love to see this.  Some small patches I've been meaning to
> do hang on the lack of proper parsing.

   My eyes hurt.

   I'm in the process of resurrecting my split patch, but I'm learning
how, uhm, "robust" the command line-parsing is.  All the following
commands are equivalent on the server:

set scorelog 1
/set scorelog 1
      /set scorelog 1
!@#$%^&*()/set scorelog 1
/!@#$%^&*()set scorelog 1

   I can see 1, 2, and maybe 3 as being valid.  But is there really any
reason to make the code as complex as it is just so it can handle cases 4
and 5?  Yes, I know, be conservative in what you send and liberal in what
you accept, but I think this may be a bit overboard.  The command-line
parsing could/should probably be implemented this way:

  /* ... */
  strip_leading_whitespace();
  handle_server_char(); /* the '/', can/should keep the if statement */
  split(pattern, input);  /* Get command in one buf, all args in other */
  lookup_command(command); /* Get a callback function */
  command_func(args);
  /* ... */

   If the command they entered was invalid, tell them so.  The command
struct could easily be modified to facilitate a change by associating a
function pointer with it.  Thus:

typedef void (*command_func)(char *args);

struct command {
  char *name;              /* name - will be matched by unique prefix   */
  enum cmdlevel_id level;  /* access level required to use the command  */
  char *synopsis;          /* one or few-line summary of usage */
  char *short_help;        /* one line (about 70 chars) description */
  char *extra_help;        /* extra help information; will be line-wrapped */
  command_func *fptr;      /* Pointer to a callback function */
};

void set_cmd(char *args);
void hard_cmd(char *args);
void wall_cmd(char *args);
/* ... */
static struct command commands[] = {
  /* ... */
  {"set",       ALLOW_CTRL,
   N_("set <option-name> <value>"),
   N_("Set server options."),
   set_cmd
  },
  {"hard",      ALLOW_CTRL,
   /* TRANS: translate text between <> only */
   N_("hard\n"
      "hard <player-name>"),
   N_("Set one or all AI players to 'hard'."),
   N_("With no arguments, sets all AI players to skill level 'hard', and "
      "sets the default level for any new AI players to 'hard'.  With an "
      "argument, sets the skill level for that player only.")
   hard_cmd
  },
  {"wall",      ALLOW_HACK,
   N_("wall <message>"),
   N_("Send message to all connections."),
   N_("For each connected client, pops up a window showing the message "
      "entered.")
   wall_cmd
  },
  /* ... */
}

   Within each *_cmd function it could do all the parsing of per-function
arguments.  The 'set' function could be much, much simpler, too, and allow
for per-option callbacks, too (just an idea).  The init_techs patch (among
others) would become relatively trivial.  The syntax could be

set attribute nation.Greek.techs Railroads,Explosives,etc,etc

   And in the attribute sub-func call:

#define MAX_PARSE_DETAIL 6 /* Can only be x.x.x.x.x.x descriptions */
#define MAX_INIT_TECHS  16 /* Don't want anyone to be too strong */

char *opts[MAX_PARSE_DETAIL];
int nfound;
char *techs[MAX_INIT_TECHS];

nfound = split(".", args[2], opts, MAX_PARSE_DETAIL);
/* Realize we're doing a nation */
/* Get Greek player data */
/* Realize it's tech-related */
nfound = split(",", args[3], techs, MAX_INIT_TECHS);
/* Add all techs */

   Or something like that. :)

   Now my brain hurts.

-jdm

"You don't give blood then take it back again
 We're all deserving of something more"
 - "Grievance"

Department of Computer Science, Duke University, Durham, NC 27708-0129
Email:  justin@xxxxxxxxxxx