peter@ficc.ferranti.com (Peter da Silva) (09/07/90)
In article <17632@haddock.ima.isc.com> karl@kelp.ima.isc.com (Karl Heuer) writes: > Fine; functions that have that property can be declared with a type other than > "bool". Functions that don't have any useful value to return on success can > and should return TRUE. (fgets() was another mistake.) Fgets() has a useful value to return: the number of characters read. A 0 length read from fgets should be an EOF (think about it). For that matter, the whole stdio suite is fubar. But, it's a standard. Perhaps a new library could be designed in alt.lang.cfutures FILE *openf(name, mode); char *name; int mode; /* From fcntl.h */ int readf(fp, buf, len); FILE *fp; char *buf; int len; int readlnf(fp, buf, len); FILE *fp; char *buf; int len; int parsef(fp, fmt, args...) FILE *fp; char *fmt; int readln(buf, len); char *buf; int len; int formatf(fp, fmt, args...); FILE *fp; char *fmt; int formats(buf, len, fmt, args...); char *buf; int len; char *fmt; Fp, when present, always comes first. Buffers are always associated with a length. Consistent rules for the syntax would make things a lot easier to remember. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
stephen@estragon.uchicago.edu (Stephen P Spackman) (09/08/90)
In article <38Q5QG2@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes:
For that matter, the whole stdio suite is fubar. But, it's a standard.
Perhaps a new library could be designed in alt.lang.cfutures
[...]
Another thing that is needed is a function
FILE *pseudof
( int (*producer)(void *handle, char *buf, size_t len)
, int (*consumer)(void *handle, char const *buf, size_t len)
, void *handle
, char const *mode
);
that takes USER-DEFINED buffer readers and writers and makes a "FILE"
object. That way, you gat all the sprintf analogues for free, and you
can fake up portable I/O for nonstandard filesystems and all sorts of
groovy stuff.
That signature needs further thought - maybe it needs a flush/close
function as well, e.g.; but it's not so far away.
stephen p spackman stephen@estragon.uchicago.edu 312.702.3982chris@mimsy.umd.edu (Chris Torek) (09/08/90)
In article <STEPHEN.90Sep7125034@estragon.uchicago.edu> stephen@estragon.uchicago.edu (Stephen P Spackman) writes: >Another thing that is needed is a function > FILE *pseudof > ( int (*producer)(void *handle, char *buf, size_t len) > , int (*consumer)(void *handle, char const *buf, size_t len) > , void *handle > , char const *mode > ); > >that takes USER-DEFINED buffer readers and writers and makes a "FILE" >object. Gee, you mean like this one? :-) FILE *funopen(const void *_cookie, int (*readfn)(void *_cookie, char *_buf, int _n), int (*writefn)(void *_cookie, const char *_buf, int _n), fpos_t (*seekfn)(void *_cookie, fpos_t _off, int _whence), int (*closefn)(void *_cookie)); >That signature needs further thought - maybe it needs a flush/close >function as well, e.g.; but it's not so far away. No flush function is required, but seek and close functions are. No `mode' is required either; this is implicit in the arguments. (At least one of readfn and writefn must be non nil; if both are, the function is a read-and-write. All operations are `binary'; any `text' conversion must be done in the read/write operations.) This is running in the stdio on my 4.3BSD-reno-ish machine. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 405 2750) Domain: chris@cs.umd.edu Path: uunet!mimsy!chris
peter@sugar.hackercorp.com (Peter da Silva) (09/08/90)
In article <STEPHEN.90Sep7125034@estragon.uchicago.edu> stephen@estragon.uchicago.edu (Stephen P Spackman) writes: > Another thing that is needed is a function > FILE *pseudof > ( int (*producer)(void *handle, char *buf, size_t len) > , int (*consumer)(void *handle, char const *buf, size_t len) > , void *handle > , char const *mode > ); That's pretty cool, man. I like it. Except that in my ideal stdio the mode is back to being a binary mask, but that was easy to miss. The mas would have values like: O_RONLY O_RDWR O_WRONLY O_BINARY O_APPEND O_UNBUFFERED etc... > That signature needs further thought - maybe it needs a flush/close > function as well, e.g.; but it's not so far away. Sounds good. -- Peter da Silva. `-_-' <peter@sugar.hackercorp.com>.
roland@ai.mit.edu (Roland McGrath) (09/09/90)
The GNU C Library (when I release it--don't ask) allows user-specified read, write, seek, and close functions that operate on magic cookies (void *). There is also another level of customization that can be made: user-specified input_room and output_room functions. These must understand stdio internals, but are very useful for things like a self-expanding buffer (this is how I implement asprintf, which is sprintf but gives you back a malloc'd string of just the right length). -- Roland McGrath Free Software Foundation, Inc. roland@ai.mit.edu, uunet!ai.mit.edu!roland
scott@nbc1.ge.com (Scott Barman) (09/10/90)
In article <38Q5QG2@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: >Fp, when present, always comes first. Buffers are always associated with >a length. Consistent rules for the syntax would make things a lot easier >to remember. >-- >Peter da Silva. `-_-' Aww Peter, now why did you do that? It made sense!!! :-) -- scott barman NBC Systems Development scott@nbc1.ge.com 30 Rockerfeller Plaza, Room 1615W {philabs,crdgw1}!nbc1!scott New York, NY 10112 +1 212/664-2787 (This does not represent any [un]official opinions of NBC or its affiliates)