On Fri, Mar 27, 2020 at 02:04:18PM -0700, Junio C Hamano wrote:
> Patrick Steinhardt <ps@pks.im> writes:
> 
> > While the strbuf interface already provides functions to read a line
> > into it that completely replaces its current contents, we do not have an
> > interface that allows for appending lines without discarding current
> > contents.
> >
> > Add a new function `strbuf_appendwholeline` that reads a line including
> > its terminating character into a strbuf non-destructively. This is a
> > preparatory step for git-update-ref(1) reading standard input line-wise
> > instead of as a block.
> 
> Hmph.  If we were to gain more uses of this function over time, the
> necessity for an extra strbuf_addbuf() becomes annoying.  I wonder
> if we should be doing this patch the other way around, i.e. move the
> whole implementation, except for the early
> 
> 	if (feof(fp))
> 		return EOF;
> 	strbuf_reset(sb);
> 
> part, and call it strbuf_addwholeline(), and strbuf_getwholeline()
> becomes a thin wrapper around it that resets the incoming buffer
> before calling strbuf_addwholeline().  The logic to determine EOF
> would need to be updated if we go that route (i.e. instead of
> checking sb->len is zero in the !HAVE_GETDELIM side of the
> implementation, we need to see if the number is the same as before)
> but it should be minor.

Unfortunately it's not that easy for the HAVE_GETDELIM case, though, as
we cannot call getdelim(3P) to append to an already populated buffer. So
we'd have to call getdelim(3P) either on a NULL line and append manually
or on an empty line in case the strbuf doesn't have any contents. While
doable, I wanted to keep out this additional complexity for now.

Patrick