Re: io:lines() and \0

[Andrew Starks <andrew.starks@...>]

On Friday, February 21, 2014, Cezary H. Noweta <chn@poczta.onet.pl> wrote:
> On 2014-02-21 08:35, René Rebe wrote:
>
> > it does NOT say:
>
> > MAY copy, or clobber, additional bytes after the \n was encountered.
>
> It *does* say this implicitly, i.e. it *does not* say this is
> prohibited. In the simplest case, ,,fgets()'' can memset all an outut
> buffer with zeroes at the beginning of its operation to verify that all
> the buffer is writable. Untouched buffer is present solely in an
> EOF-at-the-beginning case.
>
> {To René: Your patch has one major disadvantage: it assumes that
> the buffer will not be touched beyond terminating null character. In
> general it *is not* true. Lua is sitting on top of C and not on top of
> your implementation of C.}
>
> ,,fgets'' *is not* for reading zeroes (no matter: from text or from
> binary streams). Using ,,fgets'' is ,,implementation detail'' which
> should be mentioned in a documentation at least (or maybe, at most?).
>
> BTW. It is unreasonable, how many can be told about differences between
> text/binary streams, in the world, where 99.99% implementations *does
> not* distinguish between them. :-)
>
> -- best regards
>
> Cezary H. Noweta

What we all think it should do is a factor. What your sub-30 year old trying to script a Lua-extended application by reading in a machine-generated "text" file is another importantant factor. 

This isn't a vote. If it were a vote, mine would be for raising an error. Especially given that you could also read chunks ("*n") at a time and parse the stream in anyway you please, 0's and all.  

Lua's library functions are lightweight bindings around C, for the most part. Radically different behavior than C would bring too much complexity. However, catching behavior that would confuse someone without a masters in CSCI seems to be in line with the role that Lua plays. 

The update to the manual is also an improvement and perhaps it is enough?

-Andrew