[racket] How to invoke a blocking foreign function without blocking whole racket?
On Wed, Jan 30, 2013 at 3:06 PM, Haiwei Zhou <highfly22 at gmail.com> wrote:
> Cool! The open-fd-input-port is a great function.
>
Yes, indeed. Maybe something like this would be a useful addition to Racket.
> Inotify cannot handle subtrees. I will try to make it more convenience.
>
FWIW, the code I gave above can add watches recursively (see around line
252),
but it does not (yet) add watches to directories that are added in watched
directory.
It can also follow symlinks.
> I plan to make a tool like guard - https://github.com/guard/guard
>
It's useful to run unit tests while some files are changed.
>
That'd be nice indeed.
Laurent
>
> Thanks,
> Haiwei
>
>
> On 30 January 2013 20:52, Laurent <laurent.orseau at gmail.com> wrote:
>
>> For the record, Kevin Tew's function makes it easy to use:
>> https://github.com/kazzmir/x11-racket/blob/master/fd.rkt
>>
>> For example, that's what I use in my own version of libinotify FFI:
>> https://github.com/Metaxal/linux-tools/blob/master/inotify.rkt
>>
>> Laurent
>>
>>
>> On Wed, Jan 30, 2013 at 12:28 PM, Tim Brown <tim at timb.net> wrote:
>>
>>> Resend... original (and previous resend!) sent from an address not
>>> valid for the list.
>>>
>>> ---------- Forwarded message ----------
>>> From: Tim Brown <tim.brown at timb.net>
>>> Date: 30 January 2013 10:01
>>> Subject: Re: [racket] How to invoke a blocking foreign function
>>> without blocking whole racket?
>>> To: Haiwei Zhou <highfly22 at gmail.com>
>>> Cc: users at racket-lang.org
>>>
>>>
>>> There are internal FFI calls: scheme_make_fd_input_port and
>>> scheme_make_fd_output_port, documented in
>>>
>>> http://docs.racket-lang.org/inside/Ports_and_the_Filesystem.html?q=to%20port#(idx._(gentag._480._(lib._scribblings/inside/inside..scrbl)))<http://docs.racket-lang.org/inside/Ports_and_the_Filesystem.html?q=to%20port#%28idx._%28gentag._480._%28lib._scribblings/inside/inside..scrbl%29%29%29>
>>>
>>> If you would normally be able to poll (or select) from the port, then
>>> a little bit of work will allow you to tie your IO closer to the
>>> racket events system.
>>>
>>> There are more involved variants, I've mostly only needed this level
>>> of access at the racket level; servicing the known-to-be ready fd is
>>> done by your FFI library.
>>>
>>> Tim
>>>
>>> On 30 Jan 2013 09:44, "Haiwei Zhou" <highfly22 at gmail.com> wrote:
>>> >
>>> > Finally, I figured out the callback from this letter:
>>> http://lists.racket-lang.org/users/archive/2010-July/040409.html . I
>>> wish I red this letter early.
>>> >
>>> > After realized that the foreign function blocks the caller thread, I
>>> used dynamic-place to create worker thread. The code is here :
>>> https://github.com/highfly22/alert/blob/master/inotify.rkt
>>> >
>>> > Racket is amazing simple after you know the bloody details.
>>> >
>>> > Haiwei
>>> >
>>> >
>>> > On 30 January 2013 14:23, Haiwei Zhou <highfly22 at gmail.com> wrote:
>>> >>
>>> >> Hi,
>>> >>
>>> >> I try to wrap inotify API. After adding a watcher, I try to read
>>> events from file descriptor. But the racket thread is blocking until the
>>> read operation is done. Here is the definition of read.
>>> >>
>>> >> (define _read (get-ffi-obj "read" libc (_fun #:async-apply (lambda
>>> (f) (f))
>>> >> #:save-errno 'posix
>>> >> _fd_t
>>> >> (output : (_bytes o size))
>>> >> (size : _uint32)
>>> >> -> (r : _int32)
>>> >> -> (values r saved-errno
>>> output))))
>>> >>
>>> >> I am confused by the term callback. Is the callback called in the
>>> racket thread or another OS thread?
>>> >>
>>> >> Thanks,
>>> >> Haiwei
>>> >
>>> >
>>> >
>>> > ____________________
>>> > Racket Users list:
>>> > http://lists.racket-lang.org/users
>>> >
>>> ____________________
>>> Racket Users list:
>>> http://lists.racket-lang.org/users
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.racket-lang.org/users/archive/attachments/20130130/13232ca1/attachment.html>