On May 24, 2010, at 1:11 PM, Matthew Flatt wrote: > At Mon, 24 May 2010 11:12:56 -0700, John Clements wrote: >> Is there a good reason why port->lines doesn't close the input port? > > An eof-of-file doesn't necessarily mean an end-of-stream, and error > handling usually requires extra effort to close a port, which is why we > usually use functions like `with-input-form-file'. Probably, though, > it's mostly that port-reading functions don't normally close the port. > >> Perhaps there >> could be an optional argument that allowed this behavior? > > An optional keyword argument seems ok to me.
How about:
(port->lines p #:close-on-eof? #t)
implemented something like this:
extra optional arguments in port.rkt:
...
(define (port->lines [p (current-input-port)] #:line-mode [mode 'any]
#:close-on-eof? [close-on-eof? #f])
(port->x-lines 'port->lines p mode read-line close-on-eof?))
(define (port->bytes-lines [p (current-input-port)] #:line-mode [mode 'any]
#:close-on-eof? [close-on-eof? #f])
(port->x-lines 'port->bytes-lines p mode read-bytes-line close-on-eof?))
...
additional explicit arg in portlines.rkt:
(define (port->x-lines who p mode read-line close-on-eof?)
(unless (input-port? p)
(raise-type-error who "input-port" p))
(check-mode who mode)
(let loop ([l null])
(let ([line (read-line p mode)])
(if (eof-object? line)
(begin0 (reverse l)
(when close-on-eof?
(close-input-port p)))
(loop (cons line l))))))
If this makes sense, I'll add docs (and a test case, if there are existing
rackety test cases for things like this).
N.B.: Nothing here is committed yet, and I'm not proposing this for 5.0.
John
smime.p7s
Description: S/MIME cryptographic signature
_________________________________________________ For list-related administrative tasks: http://list.cs.brown.edu/mailman/listinfo/plt-dev
