I think it'd be better to use Parsec to parse .authorspellings, rather
than an ad-hoc Perl-style use of primitives "dropWhile isSpace".

We already require Parsec for other parts of Darcs, so it would not
add a new dependency.

On Sat, Mar 14, 2009 at 07:22:11AM -0000, Trent Buck wrote:
> I think it'd be better to use Parsec to parse .authorspellings,
> rather than an ad-hoc Perl-style use of primitives "dropWhile
> isSpace".

While I'm on the subject, I note that currently

- only Unix-style EOLs are supported;

- comments cannot start with whitespace;

- canonical names cannot contain a comma (e.g. "Chuck Jones,
  Jr. <chuck@pobox.com>");

- regexps cannot contain commas; and

- regexps cannot begin or end in [[:space:]].

The last point is not so bad, since you can just prefix a leading
space with a backslash (but not a trailing space).  The point about ",
Jr." worries me, though.

Since regex already has foo|bar I suggest only supporting one regexp.

I also don't see the point of having canonical names *without* a
regexp.

I've submitted patch (#94) to address majority of issues raised here:
 1) I reimplemented parser using parsec.
 2) I added escape mechanism, so commas can be now parts of names and regexps
 3) because ws are stripped i don't think there is any EOL issue 
 4) to start/end regexp with space, just use class syntax: [ ]
 5) comma separated list is in my opinion better than one regexp because:
   - alternatives are clumsy
   - many people just lists multiple e-mail addresses - it is more natural write
them down as a comma separated list than encode them in one big regexp
   - changing format is not backward compatible - all users would have to change
their .authorspelling files

There are few issues left:
 - canonical email address is now mandatory (previous doc suggested that, but
implementation did not require that)
 - malformed regexps are not handled nicely 
 - the matching process is sort of error-prone, people can accidentally take
over patches that don't belong to them. It is not unrealistic that two people
will have regexp that will match the same authorstring.

The following patch updated the status of issue1392 to be resolved:

* resolve issue1392: use parsec to parse .authorspelling 
Ignore-this: 491f823882e9e2c9a49ad303ee6fb4ba
.authorspelling file is now parsed using parsec.
Parser reports errors (only affected line is discarded).
Added escaping of commas.

The following patch updated the status of issue1392 to be resolved:

* resolve issue1392: use parsec to parse .authorspelling 
Ignore-this: 491f823882e9e2c9a49ad303ee6fb4ba
.authorspelling file is now parsed using parsec.
Parser reports errors (only affected line is discarded).
Added escaping of commas.