Re: [Groff] changed .substring

[Werner LEMBERG]

> The actual groff implementation has
> 
> .ds s abcdefg\"
> .substring s 0 3
> .tm => \*[s]
> 
> => abcd
> 
> Emacs, however, does (from the Xemacs info file):
> 
>           (substring "abcdefg" 0 3)
>                => "abc"
> 
> Python parallels Emacs.
> 
> The difference is that groff follows the notion of "including" for
> the .substring arguments, while the other languages use the notion
> of "before".

Note that both Emacs and Python have a third `parameter' for the
substring operation: `:' in Python and `nil' in Elisp which we don't
have in groff.  You can say `s[0:]' and (substring 0 nil),
respectively, to get the full string.  And in Python you can omit the
colon to access a single element.

> I prefer Werner's implementation from a user's point of view.  But
> then we must be conscious about the fact that .substring is not
> compatible to other languages.

I can live with that. :-)

Even after some thinking I can't find the real advantages of using a
[start,end[ range (i.e., including start but excluding end).  And
there is no way to specify the full range of a string if the second
argument counts from the end: In Python, this would be [0:(-1+1)] =
[0:0] but this is the empty string (not very useful in a substring
operation IMHO).

> If we agree on discarding this kind of compatibility, I have another
> proposal for improvement.  The actual groff implementation allows to
> specify the rest of the string by having `-1' as a third argument,
> e.g.
> 
> .ds s abcdefg\"
> .substring s 2 -1
> .tm => \*[s]
> 
> => cdefg
> 

> So it is not necessary to provide this feature again by the
> construct with the last argument omitted.  We can use the 2-argument
> version of .substring for other purposes, for example, to retrieve
> the character for a single index.
> 
> More exactly, I propose to identify `.substring s n' with 
> `.substring s n n' instead of the actual `.substring s n -1'.

I don't have an opinion here.  For me, both Bernd's suggestion
(resembling Python) and the current implementation (resembling Elisp)
make sense.

What do others think?

Please answer quickly!  I really, really want to release 1.18, but we
have to agree on the new .substring syntax before since I won't change
it again.


    Werner